Add cache for when rollbacking on edge adding and removing

* Add AUTH system tx deltas
* Add auth data RPC and handlers
* Support multiple system deltas in a single transaction
* Added e2e test
* Bugfix: KVStore segfault after move

---------

Co-authored-by: Gareth Lloyd <gareth.lloyd@memgraph.io>

ADRs/003_rocksdb.md

@@ -0,0 +1,38 @@
+# RocksDB ADR
+
+**Author**
+Marko Budiselic (github.com/gitbuda)
+
+**Status**
+ACCEPTED
+
+**Date**
+January 23, 2024
+
+**Problem**
+
+Interacting with data (reads and writes) on disk in a concurrent, safe, and
+fast way is a challenging task. Implementing all low-level primitives to
+interact with various disk hardware efficiently consumes significant
+engineering people. Whenever Memgraph has to store data on disk (or any
+other colder than RAM storage system), the problem is how to do that in the
+least amount of development time while satisfying all functional requirements
+(often performance).
+
+**Criteria**
+
+- working efficiently in a highly concurrent environment
+- easy integration with Memgraph's C++ codebase
+- providing low-level key-value API
+- heavily tested in production environments
+- providing abstractions for the storage hardware (even for cloud-based
+  storages like S3)
+
+**Decision**
+
+There are a few robust key-value stores, but finding one that is
+production-ready and compatible with Memgraph's C++ codebase is challenging.
+**We select [RocksDB](https://github.com/facebook/rocksdb)** because it
+delivers robust API to manage data on disk; it's battle-tested in many
+production environments (many databases systems are embedding RocksDB), and
+it's the most compatible one.

src/CMakeLists.txt

@@ -22,8 +22,10 @@ add_subdirectory(dbms)
 add_subdirectory(flags)
 add_subdirectory(distributed)
 add_subdirectory(replication)
+add_subdirectory(replication_handler)
 add_subdirectory(coordination)
 add_subdirectory(replication_coordination_glue)
+add_subdirectory(system)
 
 string(TOLOWER ${CMAKE_BUILD_TYPE} lower_build_type)
 
@@ -43,10 +45,10 @@ set(mg_single_node_v2_sources
 add_executable(memgraph ${mg_single_node_v2_sources})
 target_include_directories(memgraph PUBLIC ${CMAKE_SOURCE_DIR}/include)
 target_link_libraries(memgraph stdc++fs Threads::Threads
-        mg-telemetry mg-communication mg-communication-metrics mg-memory mg-utils mg-license mg-settings mg-glue mg-flags)
+        mg-telemetry mg-communication mg-communication-metrics mg-memory mg-utils mg-license mg-settings mg-glue mg-flags mg::system mg::replication_handler)
 
 # NOTE: `include/mg_procedure.syms` describes a pattern match for symbols which
-# should be dynamically exported, so that `dlopen` can correctly link the
+# should be dynamically exported, so that `dlopen` can correctly link th
 # symbols in custom procedure module libraries.
 target_link_libraries(memgraph "-Wl,--dynamic-list=${CMAKE_SOURCE_DIR}/include/mg_procedure.syms")
 set_target_properties(memgraph PROPERTIES

src/auth/CMakeLists.txt

@@ -2,7 +2,9 @@ set(auth_src_files
     auth.cpp
     crypto.cpp
     models.cpp
-    module.cpp)
+    module.cpp
+    rpc.cpp
+    replication_handlers.cpp)
 
 find_package(Seccomp REQUIRED)
 find_package(fmt REQUIRED)
@@ -11,7 +13,7 @@ find_package(gflags REQUIRED)
 
 add_library(mg-auth STATIC ${auth_src_files})
 target_link_libraries(mg-auth json libbcrypt gflags fmt::fmt)
-target_link_libraries(mg-auth mg-utils mg-kvstore mg-license )
+target_link_libraries(mg-auth mg-utils mg-kvstore mg-license mg::system mg-replication)
 
 target_link_libraries(mg-auth ${Seccomp_LIBRARIES})
 target_include_directories(mg-auth SYSTEM PRIVATE ${Seccomp_INCLUDE_DIRS})

src/auth/auth.cpp

@@ -9,13 +9,16 @@
 #include "auth/auth.hpp"
 
 #include <iostream>
+#include <optional>
 #include <utility>
 
 #include <fmt/format.h>
 
 #include "auth/crypto.hpp"
 #include "auth/exceptions.hpp"
+#include "auth/rpc.hpp"
 #include "license/license.hpp"
+#include "system/transaction.hpp"
 #include "utils/flag_validation.hpp"
 #include "utils/message.hpp"
 #include "utils/settings.hpp"
@@ -41,12 +44,84 @@ DEFINE_VALIDATED_int32(auth_module_timeout_ms, 10000,
                        FLAG_IN_RANGE(100, 1800000));
 
 namespace memgraph::auth {
+
+namespace {
+#ifdef MG_ENTERPRISE
+/**
+ * REPLICATION SYSTEM ACTION IMPLEMENTATIONS
+ */
+struct UpdateAuthData : memgraph::system::ISystemAction {
+  explicit UpdateAuthData(User user) : user_{std::move(user)}, role_{std::nullopt} {}
+  explicit UpdateAuthData(Role role) : user_{std::nullopt}, role_{std::move(role)} {}
+
+  void DoDurability() override { /* Done during Auth execution */
+  }
+
+  bool DoReplication(replication::ReplicationClient &client, replication::ReplicationEpoch const &epoch,
+                     memgraph::system::Transaction const &txn) const override {
+    auto check_response = [](const replication::UpdateAuthDataRes &response) { return response.success; };
+    if (user_) {
+      return client.SteamAndFinalizeDelta<replication::UpdateAuthDataRpc>(
+          check_response, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(), *user_);
+    }
+    if (role_) {
+      return client.SteamAndFinalizeDelta<replication::UpdateAuthDataRpc>(
+          check_response, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(), *role_);
+    }
+    // Should never get here
+    MG_ASSERT(false, "Trying to update auth data that is not a user nor a role");
+    return {};
+  }
+
+  void PostReplication(replication::RoleMainData &mainData) const override {}
+
+ private:
+  std::optional<User> user_;
+  std::optional<Role> role_;
+};
+
+struct DropAuthData : memgraph::system::ISystemAction {
+  enum class AuthDataType { USER, ROLE };
+
+  explicit DropAuthData(AuthDataType type, std::string_view name) : type_{type}, name_{name} {}
+
+  void DoDurability() override { /* Done during Auth execution */
+  }
+
+  bool DoReplication(replication::ReplicationClient &client, replication::ReplicationEpoch const &epoch,
+                     memgraph::system::Transaction const &txn) const override {
+    auto check_response = [](const replication::DropAuthDataRes &response) { return response.success; };
+
+    memgraph::replication::DropAuthDataReq::DataType type{};
+    switch (type_) {
+      case AuthDataType::USER:
+        type = memgraph::replication::DropAuthDataReq::DataType::USER;
+        break;
+      case AuthDataType::ROLE:
+        type = memgraph::replication::DropAuthDataReq::DataType::ROLE;
+        break;
+    }
+    return client.SteamAndFinalizeDelta<replication::DropAuthDataRpc>(
+        check_response, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(), type, name_);
+  }
+  void PostReplication(replication::RoleMainData &mainData) const override {}
+
+ private:
+  AuthDataType type_;
+  std::string name_;
+};
+#endif
+
+/**
+ * CONSTANTS
+ */
 const std::string kUserPrefix = "user:";
 const std::string kRolePrefix = "role:";
 const std::string kLinkPrefix = "link:";
 const std::string kVersion = "version";
 
 static constexpr auto kVersionV1 = "V1";
+}  // namespace
 
 /**
  * All data stored in the `Auth` storage is stored in an underlying
@@ -148,6 +223,12 @@ std::optional<User> Auth::Authenticate(const std::string &username, const std::s
     // Authenticate the user.
     if (!is_authenticated) return std::nullopt;
 
+    /**
+     * TODO
+     * The auth module should not update auth data.
+     * There is now way to replicate it and we should not be storing sensitive data if we don't have to.
+     */
+
     // Find or create the user and return it.
     auto user = GetUser(username);
     if (!user) {
@@ -240,7 +321,7 @@ std::optional<User> Auth::GetUser(const std::string &username_orig) const {
   return user;
 }
 
-void Auth::SaveUser(const User &user) {
+void Auth::SaveUser(const User &user, system::Transaction *system_tx) {
   bool success = false;
   if (const auto *role = user.role(); role != nullptr) {
     success = storage_.PutMultiple(
@@ -252,6 +333,12 @@ void Auth::SaveUser(const User &user) {
   if (!success) {
     throw AuthException("Couldn't save user '{}'!", user.username());
   }
+  // All changes to the user end up calling this function, so no need to add a delta anywhere else
+  if (system_tx) {
+#ifdef MG_ENTERPRISE
+    system_tx->AddAction<UpdateAuthData>(user);
+#endif
+  }
 }
 
 void Auth::UpdatePassword(auth::User &user, const std::optional<std::string> &password) {
@@ -284,7 +371,8 @@ void Auth::UpdatePassword(auth::User &user, const std::optional<std::string> &pa
   user.UpdatePassword(password);
 }
 
-std::optional<User> Auth::AddUser(const std::string &username, const std::optional<std::string> &password) {
+std::optional<User> Auth::AddUser(const std::string &username, const std::optional<std::string> &password,
+                                  system::Transaction *system_tx) {
   if (!NameRegexMatch(username)) {
     throw AuthException("Invalid user name.");
   }
@@ -294,17 +382,23 @@ std::optional<User> Auth::AddUser(const std::string &username, const std::option
   if (existing_role) return std::nullopt;
   auto new_user = User(username);
   UpdatePassword(new_user, password);
-  SaveUser(new_user);
+  SaveUser(new_user, system_tx);
   return new_user;
 }
 
-bool Auth::RemoveUser(const std::string &username_orig) {
+bool Auth::RemoveUser(const std::string &username_orig, system::Transaction *system_tx) {
   auto username = utils::ToLowerCase(username_orig);
   if (!storage_.Get(kUserPrefix + username)) return false;
   std::vector<std::string> keys({kLinkPrefix + username, kUserPrefix + username});
   if (!storage_.DeleteMultiple(keys)) {
     throw AuthException("Couldn't remove user '{}'!", username);
   }
+  // Handling drop user delta
+  if (system_tx) {
+#ifdef MG_ENTERPRISE
+    system_tx->AddAction<DropAuthData>(DropAuthData::AuthDataType::USER, username);
+#endif
+  }
   return true;
 }
 
@@ -321,6 +415,19 @@ std::vector<auth::User> Auth::AllUsers() const {
   return ret;
 }
 
+std::vector<std::string> Auth::AllUsernames() const {
+  std::vector<std::string> ret;
+  for (auto it = storage_.begin(kUserPrefix); it != storage_.end(kUserPrefix); ++it) {
+    auto username = it->first.substr(kUserPrefix.size());
+    if (username != utils::ToLowerCase(username)) continue;
+    auto user = GetUser(username);
+    if (user) {
+      ret.push_back(username);
+    }
+  }
+  return ret;
+}
+
 bool Auth::HasUsers() const { return storage_.begin(kUserPrefix) != storage_.end(kUserPrefix); }
 
 std::optional<Role> Auth::GetRole(const std::string &rolename_orig) const {
@@ -338,24 +445,30 @@ std::optional<Role> Auth::GetRole(const std::string &rolename_orig) const {
   return Role::Deserialize(data);
 }
 
-void Auth::SaveRole(const Role &role) {
+void Auth::SaveRole(const Role &role, system::Transaction *system_tx) {
   if (!storage_.Put(kRolePrefix + role.rolename(), role.Serialize().dump())) {
     throw AuthException("Couldn't save role '{}'!", role.rolename());
   }
+  // All changes to the role end up calling this function, so no need to add a delta anywhere else
+  if (system_tx) {
+#ifdef MG_ENTERPRISE
+    system_tx->AddAction<UpdateAuthData>(role);
+#endif
+  }
 }
 
-std::optional<Role> Auth::AddRole(const std::string &rolename) {
+std::optional<Role> Auth::AddRole(const std::string &rolename, system::Transaction *system_tx) {
   if (!NameRegexMatch(rolename)) {
     throw AuthException("Invalid role name.");
   }
   if (auto existing_role = GetRole(rolename)) return std::nullopt;
   if (auto existing_user = GetUser(rolename)) return std::nullopt;
   auto new_role = Role(rolename);
-  SaveRole(new_role);
+  SaveRole(new_role, system_tx);
   return new_role;
 }
 
-bool Auth::RemoveRole(const std::string &rolename_orig) {
+bool Auth::RemoveRole(const std::string &rolename_orig, system::Transaction *system_tx) {
   auto rolename = utils::ToLowerCase(rolename_orig);
   if (!storage_.Get(kRolePrefix + rolename)) return false;
   std::vector<std::string> keys;
@@ -368,6 +481,12 @@ bool Auth::RemoveRole(const std::string &rolename_orig) {
   if (!storage_.DeleteMultiple(keys)) {
     throw AuthException("Couldn't remove role '{}'!", rolename);
   }
+  // Handling drop role delta
+  if (system_tx) {
+#ifdef MG_ENTERPRISE
+    system_tx->AddAction<DropAuthData>(DropAuthData::AuthDataType::ROLE, rolename);
+#endif
+  }
   return true;
 }
 
@@ -385,6 +504,18 @@ std::vector<auth::Role> Auth::AllRoles() const {
   return ret;
 }
 
+std::vector<std::string> Auth::AllRolenames() const {
+  std::vector<std::string> ret;
+  for (auto it = storage_.begin(kRolePrefix); it != storage_.end(kRolePrefix); ++it) {
+    auto rolename = it->first.substr(kRolePrefix.size());
+    if (rolename != utils::ToLowerCase(rolename)) continue;
+    if (auto role = GetRole(rolename)) {
+      ret.push_back(rolename);
+    }
+  }
+  return ret;
+}
+
 std::vector<auth::User> Auth::AllUsersForRole(const std::string &rolename_orig) const {
   const auto rolename = utils::ToLowerCase(rolename_orig);
   std::vector<auth::User> ret;
@@ -404,48 +535,48 @@ std::vector<auth::User> Auth::AllUsersForRole(const std::string &rolename_orig)
 }
 
 #ifdef MG_ENTERPRISE
-bool Auth::GrantDatabaseToUser(const std::string &db, const std::string &name) {
+bool Auth::GrantDatabaseToUser(const std::string &db, const std::string &name, system::Transaction *system_tx) {
   if (auto user = GetUser(name)) {
     if (db == kAllDatabases) {
       user->db_access().GrantAll();
     } else {
       user->db_access().Add(db);
     }
-    SaveUser(*user);
+    SaveUser(*user, system_tx);
     return true;
   }
   return false;
 }
 
-bool Auth::RevokeDatabaseFromUser(const std::string &db, const std::string &name) {
+bool Auth::RevokeDatabaseFromUser(const std::string &db, const std::string &name, system::Transaction *system_tx) {
   if (auto user = GetUser(name)) {
     if (db == kAllDatabases) {
       user->db_access().DenyAll();
     } else {
       user->db_access().Remove(db);
     }
-    SaveUser(*user);
+    SaveUser(*user, system_tx);
     return true;
   }
   return false;
 }
 
-void Auth::DeleteDatabase(const std::string &db) {
+void Auth::DeleteDatabase(const std::string &db, system::Transaction *system_tx) {
   for (auto it = storage_.begin(kUserPrefix); it != storage_.end(kUserPrefix); ++it) {
     auto username = it->first.substr(kUserPrefix.size());
     if (auto user = GetUser(username)) {
       user->db_access().Delete(db);
-      SaveUser(*user);
+      SaveUser(*user, system_tx);
     }
   }
 }
 
-bool Auth::SetMainDatabase(std::string_view db, const std::string &name) {
+bool Auth::SetMainDatabase(std::string_view db, const std::string &name, system::Transaction *system_tx) {
   if (auto user = GetUser(name)) {
     if (!user->db_access().SetDefault(db)) {
       throw AuthException("Couldn't set default database '{}' for user '{}'!", db, name);
     }
-    SaveUser(*user);
+    SaveUser(*user, system_tx);
     return true;
   }
   return false;

src/auth/auth.hpp

@@ -18,10 +18,15 @@
 #include "auth/module.hpp"
 #include "glue/auth_global.hpp"
 #include "kvstore/kvstore.hpp"
+#include "system/action.hpp"
 #include "utils/settings.hpp"
+#include "utils/synchronized.hpp"
 
 namespace memgraph::auth {
 
+class Auth;
+using SynchedAuth = memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock>;
+
 static const constexpr char *const kAllDatabases = "*";
 
 /**
@@ -68,6 +73,13 @@ class Auth final {
     config_ = std::move(config);
   }
 
+  /**
+   * @brief
+   *
+   * @return Config
+   */
+  Config GetConfig() const { return config_; }
+
   /**
    * Authenticates a user using his username and password.
    *
@@ -96,7 +108,7 @@ class Auth final {
    *
    * @throw AuthException if unable to save the user.
    */
-  void SaveUser(const User &user);
+  void SaveUser(const User &user, system::Transaction *system_tx = nullptr);
 
   /**
    * Creates a user if the user doesn't exist.
@@ -107,7 +119,8 @@ class Auth final {
    * @return a user when the user is created, nullopt if the user exists
    * @throw AuthException if unable to save the user.
    */
-  std::optional<User> AddUser(const std::string &username, const std::optional<std::string> &password = std::nullopt);
+  std::optional<User> AddUser(const std::string &username, const std::optional<std::string> &password = std::nullopt,
+                              system::Transaction *system_tx = nullptr);
 
   /**
    * Removes a user from the storage.
@@ -118,7 +131,7 @@ class Auth final {
    *         doesn't exist
    * @throw AuthException if unable to remove the user.
    */
-  bool RemoveUser(const std::string &username);
+  bool RemoveUser(const std::string &username, system::Transaction *system_tx = nullptr);
 
   /**
    * @brief
@@ -136,6 +149,13 @@ class Auth final {
    */
   std::vector<User> AllUsers() const;
 
+  /**
+   * @brief
+   *
+   * @return std::vector<std::string>
+   */
+  std::vector<std::string> AllUsernames() const;
+
   /**
    * Returns whether there are users in the storage.
    *
@@ -160,7 +180,7 @@ class Auth final {
    *
    * @throw AuthException if unable to save the role.
    */
-  void SaveRole(const Role &role);
+  void SaveRole(const Role &role, system::Transaction *system_tx = nullptr);
 
   /**
    * Creates a role if the role doesn't exist.
@@ -170,7 +190,7 @@ class Auth final {
    * @return a role when the role is created, nullopt if the role exists
    * @throw AuthException if unable to save the role.
    */
-  std::optional<Role> AddRole(const std::string &rolename);
+  std::optional<Role> AddRole(const std::string &rolename, system::Transaction *system_tx = nullptr);
 
   /**
    * Removes a role from the storage.
@@ -181,7 +201,7 @@ class Auth final {
    *         doesn't exist
    * @throw AuthException if unable to remove the role.
    */
-  bool RemoveRole(const std::string &rolename);
+  bool RemoveRole(const std::string &rolename, system::Transaction *system_tx = nullptr);
 
   /**
    * Gets all roles from the storage.
@@ -191,6 +211,13 @@ class Auth final {
    */
   std::vector<Role> AllRoles() const;
 
+  /**
+   * @brief
+   *
+   * @return std::vector<std::string>
+   */
+  std::vector<std::string> AllRolenames() const;
+
   /**
    * Gets all users for a role from the storage.
    *
@@ -210,7 +237,7 @@ class Auth final {
    * @return true on success
    * @throw AuthException if unable to find or update the user
    */
-  bool RevokeDatabaseFromUser(const std::string &db, const std::string &name);
+  bool RevokeDatabaseFromUser(const std::string &db, const std::string &name, system::Transaction *system_tx = nullptr);
 
   /**
    * @brief Grant access to individual database for a user.
@@ -220,7 +247,7 @@ class Auth final {
    * @return true on success
    * @throw AuthException if unable to find or update the user
    */
-  bool GrantDatabaseToUser(const std::string &db, const std::string &name);
+  bool GrantDatabaseToUser(const std::string &db, const std::string &name, system::Transaction *system_tx = nullptr);
 
   /**
    * @brief Delete a database from all users.
@@ -228,7 +255,7 @@ class Auth final {
    * @param db name of the database to delete
    * @throw AuthException if unable to read data
    */
-  void DeleteDatabase(const std::string &db);
+  void DeleteDatabase(const std::string &db, system::Transaction *system_tx = nullptr);
 
   /**
    * @brief Set main database for an individual user.
@@ -238,7 +265,7 @@ class Auth final {
    * @return true on success
    * @throw AuthException if unable to find or update the user
    */
-  bool SetMainDatabase(std::string_view db, const std::string &name);
+  bool SetMainDatabase(std::string_view db, const std::string &name, system::Transaction *system_tx = nullptr);
 #endif
 
  private:

src/auth/models.cpp

@@ -611,27 +611,49 @@ Permissions User::GetPermissions() const {
 
 #ifdef MG_ENTERPRISE
 FineGrainedAccessPermissions User::GetFineGrainedAccessLabelPermissions() const {
+  return Merge(GetUserFineGrainedAccessLabelPermissions(), GetRoleFineGrainedAccessLabelPermissions());
+}
+
+FineGrainedAccessPermissions User::GetFineGrainedAccessEdgeTypePermissions() const {
+  return Merge(GetUserFineGrainedAccessEdgeTypePermissions(), GetRoleFineGrainedAccessEdgeTypePermissions());
+}
+
+FineGrainedAccessPermissions User::GetUserFineGrainedAccessEdgeTypePermissions() const {
   if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
     return FineGrainedAccessPermissions{};
   }
 
-  if (role_) {
-    return Merge(role()->fine_grained_access_handler().label_permissions(),
-                 fine_grained_access_handler_.label_permissions());
+  return fine_grained_access_handler_.edge_type_permissions();
+}
+
+FineGrainedAccessPermissions User::GetUserFineGrainedAccessLabelPermissions() const {
+  if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
+    return FineGrainedAccessPermissions{};
   }
 
   return fine_grained_access_handler_.label_permissions();
 }
 
-FineGrainedAccessPermissions User::GetFineGrainedAccessEdgeTypePermissions() const {
+FineGrainedAccessPermissions User::GetRoleFineGrainedAccessEdgeTypePermissions() const {
   if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
     return FineGrainedAccessPermissions{};
   }
+
   if (role_) {
-    return Merge(role()->fine_grained_access_handler().edge_type_permissions(),
-                 fine_grained_access_handler_.edge_type_permissions());
+    return role()->fine_grained_access_handler().edge_type_permissions();
   }
-  return fine_grained_access_handler_.edge_type_permissions();
+  return FineGrainedAccessPermissions{};
+}
+
+FineGrainedAccessPermissions User::GetRoleFineGrainedAccessLabelPermissions() const {
+  if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
+    return FineGrainedAccessPermissions{};
+  }
+
+  if (role_) {
+    return role()->fine_grained_access_handler().label_permissions();
+  }
+  return FineGrainedAccessPermissions{};
 }
 #endif
 

src/auth/models.hpp

@@ -207,6 +207,8 @@ bool operator==(const FineGrainedAccessHandler &first, const FineGrainedAccessHa
 
 class Role final {
  public:
+  Role() = default;
+
   explicit Role(const std::string &rolename);
   Role(const std::string &rolename, const Permissions &permissions);
 #ifdef MG_ENTERPRISE
@@ -369,6 +371,10 @@ class User final {
 #ifdef MG_ENTERPRISE
   FineGrainedAccessPermissions GetFineGrainedAccessLabelPermissions() const;
   FineGrainedAccessPermissions GetFineGrainedAccessEdgeTypePermissions() const;
+  FineGrainedAccessPermissions GetUserFineGrainedAccessLabelPermissions() const;
+  FineGrainedAccessPermissions GetUserFineGrainedAccessEdgeTypePermissions() const;
+  FineGrainedAccessPermissions GetRoleFineGrainedAccessLabelPermissions() const;
+  FineGrainedAccessPermissions GetRoleFineGrainedAccessEdgeTypePermissions() const;
   const FineGrainedAccessHandler &fine_grained_access_handler() const;
   FineGrainedAccessHandler &fine_grained_access_handler();
 #endif

src/auth/replication_handlers.cpp

@@ -0,0 +1,170 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "auth/replication_handlers.hpp"
+
+#include "auth/auth.hpp"
+#include "auth/rpc.hpp"
+#include "license/license.hpp"
+
+namespace memgraph::auth {
+
+#ifdef MG_ENTERPRISE
+void UpdateAuthDataHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access, auth::SynchedAuth &auth,
+                           slk::Reader *req_reader, slk::Builder *res_builder) {
+  replication::UpdateAuthDataReq req;
+  memgraph::slk::Load(&req, req_reader);
+
+  using memgraph::replication::UpdateAuthDataRes;
+  UpdateAuthDataRes res(false);
+
+  // Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
+  //       of the set of databases. Hence no history exists to maintain regarding epoch change.
+  //       If MAIN has changed we need to check this new group_timestamp is consistent with
+  //       what we have so far.
+
+  if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
+    spdlog::debug("UpdateAuthDataHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
+                  system_state_access.LastCommitedTS());
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  try {
+    // Update
+    if (req.user) auth->SaveUser(*req.user);
+    if (req.role) auth->SaveRole(*req.role);
+    // Success
+    system_state_access.SetLastCommitedTS(req.new_group_timestamp);
+    res = UpdateAuthDataRes(true);
+    spdlog::debug("UpdateAuthDataHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
+  } catch (const auth::AuthException & /* not used */) {
+    // Failure
+  }
+
+  memgraph::slk::Save(res, res_builder);
+}
+
+void DropAuthDataHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access, auth::SynchedAuth &auth,
+                         slk::Reader *req_reader, slk::Builder *res_builder) {
+  replication::DropAuthDataReq req;
+  memgraph::slk::Load(&req, req_reader);
+
+  using memgraph::replication::DropAuthDataRes;
+  DropAuthDataRes res(false);
+
+  // Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
+  //       of the set of databases. Hence no history exists to maintain regarding epoch change.
+  //       If MAIN has changed we need to check this new group_timestamp is consistent with
+  //       what we have so far.
+
+  if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
+    spdlog::debug("DropAuthDataHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
+                  system_state_access.LastCommitedTS());
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  try {
+    // Remove
+    switch (req.type) {
+      case replication::DropAuthDataReq::DataType::USER:
+        auth->RemoveUser(req.name);
+        break;
+      case replication::DropAuthDataReq::DataType::ROLE:
+        auth->RemoveRole(req.name);
+        break;
+    }
+    // Success
+    system_state_access.SetLastCommitedTS(req.new_group_timestamp);
+    res = DropAuthDataRes(true);
+    spdlog::debug("DropAuthDataHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
+  } catch (const auth::AuthException & /* not used */) {
+    // Failure
+  }
+
+  memgraph::slk::Save(res, res_builder);
+}
+
+bool SystemRecoveryHandler(auth::SynchedAuth &auth, auth::Auth::Config auth_config,
+                           const std::vector<auth::User> &users, const std::vector<auth::Role> &roles) {
+  return auth.WithLock([&](auto &locked_auth) {
+    // Update config
+    locked_auth.SetConfig(std::move(auth_config));
+    // Get all current users
+    auto old_users = locked_auth.AllUsernames();
+    // Save incoming users
+    for (const auto &user : users) {
+      // Missing users
+      try {
+        locked_auth.SaveUser(user);
+      } catch (const auth::AuthException &) {
+        spdlog::debug("SystemRecoveryHandler: Failed to save user");
+        return false;
+      }
+      const auto it = std::find(old_users.begin(), old_users.end(), user.username());
+      if (it != old_users.end()) old_users.erase(it);
+    }
+    // Delete all the leftover users
+    for (const auto &user : old_users) {
+      if (!locked_auth.RemoveUser(user)) {
+        spdlog::debug("SystemRecoveryHandler: Failed to remove user \"{}\".", user);
+        return false;
+      }
+    }
+
+    // Roles are only supported with a license
+    if (license::global_license_checker.IsEnterpriseValidFast()) {
+      // Get all current roles
+      auto old_roles = locked_auth.AllRolenames();
+      // Save incoming users
+      for (const auto &role : roles) {
+        // Missing users
+        try {
+          locked_auth.SaveRole(role);
+        } catch (const auth::AuthException &) {
+          spdlog::debug("SystemRecoveryHandler: Failed to save user");
+          return false;
+        }
+        const auto it = std::find(old_roles.begin(), old_roles.end(), role.rolename());
+        if (it != old_roles.end()) old_roles.erase(it);
+      }
+      // Delete all the leftover users
+      for (const auto &role : old_roles) {
+        if (!locked_auth.RemoveRole(role)) {
+          spdlog::debug("SystemRecoveryHandler: Failed to remove user \"{}\".", role);
+          return false;
+        }
+      }
+    }
+
+    // Success
+    return true;
+  });
+}
+
+void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
+              auth::SynchedAuth &auth) {
+  // NOTE: Register even without license as the user could add a license at run-time
+  data.server->rpc_server_.Register<replication::UpdateAuthDataRpc>(
+      [system_state_access, &auth](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received UpdateAuthDataRpc");
+        UpdateAuthDataHandler(system_state_access, auth, req_reader, res_builder);
+      });
+  data.server->rpc_server_.Register<replication::DropAuthDataRpc>(
+      [system_state_access, &auth](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received DropAuthDataRpc");
+        DropAuthDataHandler(system_state_access, auth, req_reader, res_builder);
+      });
+}
+#endif
+
+}  // namespace memgraph::auth

src/auth/replication_handlers.hpp

@@ -0,0 +1,31 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include "auth/auth.hpp"
+#include "replication/state.hpp"
+#include "slk/streams.hpp"
+#include "system/state.hpp"
+
+namespace memgraph::auth {
+#ifdef MG_ENTERPRISE
+void UpdateAuthDataHandler(system::ReplicaHandlerAccessToState &system_state_access, auth::SynchedAuth &auth,
+                           slk::Reader *req_reader, slk::Builder *res_builder);
+void DropAuthDataHandler(system::ReplicaHandlerAccessToState &system_state_access, auth::SynchedAuth &auth,
+                         slk::Reader *req_reader, slk::Builder *res_builder);
+
+bool SystemRecoveryHandler(auth::SynchedAuth &auth, auth::Auth::Config auth_config,
+                           const std::vector<auth::User> &users, const std::vector<auth::Role> &roles);
+void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
+              auth::SynchedAuth &auth);
+#endif
+}  // namespace memgraph::auth

src/auth/rpc.cpp

@@ -0,0 +1,178 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "auth/rpc.hpp"
+
+#include <json/json.hpp>
+#include "auth/auth.hpp"
+#include "slk/serialization.hpp"
+#include "slk/streams.hpp"
+#include "utils/enum.hpp"
+
+namespace memgraph::slk {
+
+// Serialize code for auth::Role
+void Save(const auth::Role &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.Serialize().dump(), builder);
+}
+namespace {
+auth::Role LoadAuthRole(memgraph::slk::Reader *reader) {
+  std::string tmp;
+  memgraph::slk::Load(&tmp, reader);
+  const auto json = nlohmann::json::parse(tmp);
+  return memgraph::auth::Role::Deserialize(json);
+}
+}  // namespace
+// Deserialize code for auth::Role
+void Load(auth::Role *self, memgraph::slk::Reader *reader) { *self = LoadAuthRole(reader); }
+// Special case for optional<Role>
+template <>
+inline void Load<auth::Role>(std::optional<auth::Role> *obj, Reader *reader) {
+  bool exists = false;
+  Load(&exists, reader);
+  if (exists) {
+    obj->emplace(LoadAuthRole(reader));
+  } else {
+    *obj = std::nullopt;
+  }
+}
+
+// Serialize code for auth::User
+void Save(const auth::User &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.Serialize().dump(), builder);
+  std::optional<auth::Role> role{};
+  if (const auto *role_ptr = self.role(); role_ptr) {
+    role.emplace(*role_ptr);
+  }
+  memgraph::slk::Save(role, builder);
+}
+// Deserialize code for auth::User
+void Load(auth::User *self, memgraph::slk::Reader *reader) {
+  std::string tmp;
+  memgraph::slk::Load(&tmp, reader);
+  const auto json = nlohmann::json::parse(tmp);
+  *self = memgraph::auth::User::Deserialize(json);
+  std::optional<auth::Role> role{};
+  memgraph::slk::Load(&role, reader);
+  if (role)
+    self->SetRole(*role);
+  else
+    self->ClearRole();
+}
+
+// Serialize code for auth::Auth::Config
+void Save(const auth::Auth::Config &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.name_regex_str, builder);
+  memgraph::slk::Save(self.password_regex_str, builder);
+  memgraph::slk::Save(self.password_permit_null, builder);
+}
+// Deserialize code for auth::Auth::Config
+void Load(auth::Auth::Config *self, memgraph::slk::Reader *reader) {
+  std::string name_regex_str{};
+  std::string password_regex_str{};
+  bool password_permit_null{};
+
+  memgraph::slk::Load(&name_regex_str, reader);
+  memgraph::slk::Load(&password_regex_str, reader);
+  memgraph::slk::Load(&password_permit_null, reader);
+
+  *self = auth::Auth::Config{std::move(name_regex_str), std::move(password_regex_str), password_permit_null};
+}
+
+// Serialize code for UpdateAuthDataReq
+void Save(const memgraph::replication::UpdateAuthDataReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.epoch_id, builder);
+  memgraph::slk::Save(self.expected_group_timestamp, builder);
+  memgraph::slk::Save(self.new_group_timestamp, builder);
+  memgraph::slk::Save(self.user, builder);
+  memgraph::slk::Save(self.role, builder);
+}
+void Load(memgraph::replication::UpdateAuthDataReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->epoch_id, reader);
+  memgraph::slk::Load(&self->expected_group_timestamp, reader);
+  memgraph::slk::Load(&self->new_group_timestamp, reader);
+  memgraph::slk::Load(&self->user, reader);
+  memgraph::slk::Load(&self->role, reader);
+}
+
+// Serialize code for UpdateAuthDataRes
+void Save(const memgraph::replication::UpdateAuthDataRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.success, builder);
+}
+void Load(memgraph::replication::UpdateAuthDataRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->success, reader);
+}
+
+// Serialize code for DropAuthDataReq
+void Save(const memgraph::replication::DropAuthDataReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.epoch_id, builder);
+  memgraph::slk::Save(self.expected_group_timestamp, builder);
+  memgraph::slk::Save(self.new_group_timestamp, builder);
+  memgraph::slk::Save(utils::EnumToNum<2, uint8_t>(self.type), builder);
+  memgraph::slk::Save(self.name, builder);
+}
+void Load(memgraph::replication::DropAuthDataReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->epoch_id, reader);
+  memgraph::slk::Load(&self->expected_group_timestamp, reader);
+  memgraph::slk::Load(&self->new_group_timestamp, reader);
+  uint8_t type_tmp = 0;
+  memgraph::slk::Load(&type_tmp, reader);
+  if (!utils::NumToEnum<2>(type_tmp, self->type)) {
+    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
+  }
+  memgraph::slk::Load(&self->name, reader);
+}
+
+// Serialize code for DropAuthDataRes
+void Save(const memgraph::replication::DropAuthDataRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.success, builder);
+}
+void Load(memgraph::replication::DropAuthDataRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->success, reader);
+}
+
+}  // namespace memgraph::slk
+
+namespace memgraph::replication {
+
+constexpr utils::TypeInfo UpdateAuthDataReq::kType{utils::TypeId::REP_UPDATE_AUTH_DATA_REQ, "UpdateAuthDataReq",
+                                                   nullptr};
+
+constexpr utils::TypeInfo UpdateAuthDataRes::kType{utils::TypeId::REP_UPDATE_AUTH_DATA_RES, "UpdateAuthDataRes",
+                                                   nullptr};
+
+constexpr utils::TypeInfo DropAuthDataReq::kType{utils::TypeId::REP_DROP_AUTH_DATA_REQ, "DropAuthDataReq", nullptr};
+
+constexpr utils::TypeInfo DropAuthDataRes::kType{utils::TypeId::REP_DROP_AUTH_DATA_RES, "DropAuthDataRes", nullptr};
+
+void UpdateAuthDataReq::Save(const UpdateAuthDataReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void UpdateAuthDataReq::Load(UpdateAuthDataReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+void UpdateAuthDataRes::Save(const UpdateAuthDataRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void UpdateAuthDataRes::Load(UpdateAuthDataRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+
+void DropAuthDataReq::Save(const DropAuthDataReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void DropAuthDataReq::Load(DropAuthDataReq *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
+void DropAuthDataRes::Save(const DropAuthDataRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void DropAuthDataRes::Load(DropAuthDataRes *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
+
+}  // namespace memgraph::replication

src/auth/rpc.hpp

@@ -0,0 +1,119 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include <optional>
+
+#include "auth/auth.hpp"
+#include "auth/models.hpp"
+#include "rpc/messages.hpp"
+#include "slk/streams.hpp"
+
+namespace memgraph::replication {
+
+struct UpdateAuthDataReq {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(UpdateAuthDataReq *self, memgraph::slk::Reader *reader);
+  static void Save(const UpdateAuthDataReq &self, memgraph::slk::Builder *builder);
+  UpdateAuthDataReq() = default;
+  UpdateAuthDataReq(std::string epoch_id, uint64_t expected_ts, uint64_t new_ts, auth::User user)
+      : epoch_id{std::move(epoch_id)},
+        expected_group_timestamp{expected_ts},
+        new_group_timestamp{new_ts},
+        user{std::move(user)} {}
+  UpdateAuthDataReq(std::string epoch_id, uint64_t expected_ts, uint64_t new_ts, auth::Role role)
+      : epoch_id{std::move(epoch_id)},
+        expected_group_timestamp{expected_ts},
+        new_group_timestamp{new_ts},
+        role{std::move(role)} {}
+
+  std::string epoch_id;
+  uint64_t expected_group_timestamp;
+  uint64_t new_group_timestamp;
+  std::optional<auth::User> user;
+  std::optional<auth::Role> role;
+};
+
+struct UpdateAuthDataRes {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(UpdateAuthDataRes *self, memgraph::slk::Reader *reader);
+  static void Save(const UpdateAuthDataRes &self, memgraph::slk::Builder *builder);
+  UpdateAuthDataRes() = default;
+  explicit UpdateAuthDataRes(bool success) : success{success} {}
+
+  bool success;
+};
+
+using UpdateAuthDataRpc = rpc::RequestResponse<UpdateAuthDataReq, UpdateAuthDataRes>;
+
+struct DropAuthDataReq {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(DropAuthDataReq *self, memgraph::slk::Reader *reader);
+  static void Save(const DropAuthDataReq &self, memgraph::slk::Builder *builder);
+  DropAuthDataReq() = default;
+
+  enum class DataType { USER, ROLE };
+
+  DropAuthDataReq(std::string epoch_id, uint64_t expected_ts, uint64_t new_ts, DataType type, std::string_view name)
+      : epoch_id{std::move(epoch_id)},
+        expected_group_timestamp{expected_ts},
+        new_group_timestamp{new_ts},
+        type{type},
+        name{name} {}
+
+  std::string epoch_id;
+  uint64_t expected_group_timestamp;
+  uint64_t new_group_timestamp;
+  DataType type;
+  std::string name;
+};
+
+struct DropAuthDataRes {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(DropAuthDataRes *self, memgraph::slk::Reader *reader);
+  static void Save(const DropAuthDataRes &self, memgraph::slk::Builder *builder);
+  DropAuthDataRes() = default;
+  explicit DropAuthDataRes(bool success) : success{success} {}
+
+  bool success;
+};
+
+using DropAuthDataRpc = rpc::RequestResponse<DropAuthDataReq, DropAuthDataRes>;
+
+}  // namespace memgraph::replication
+
+namespace memgraph::slk {
+
+void Save(const auth::Role &self, memgraph::slk::Builder *builder);
+void Load(auth::Role *self, memgraph::slk::Reader *reader);
+void Save(const auth::User &self, memgraph::slk::Builder *builder);
+void Load(auth::User *self, memgraph::slk::Reader *reader);
+void Save(const auth::Auth::Config &self, memgraph::slk::Builder *builder);
+void Load(auth::Auth::Config *self, memgraph::slk::Reader *reader);
+
+void Save(const memgraph::replication::UpdateAuthDataRes &self, memgraph::slk::Builder *builder);
+void Load(memgraph::replication::UpdateAuthDataRes *self, memgraph::slk::Reader *reader);
+void Save(const memgraph::replication::UpdateAuthDataReq & /*self*/, memgraph::slk::Builder * /*builder*/);
+void Load(memgraph::replication::UpdateAuthDataReq * /*self*/, memgraph::slk::Reader * /*reader*/);
+void Save(const memgraph::replication::DropAuthDataRes &self, memgraph::slk::Builder *builder);
+void Load(memgraph::replication::DropAuthDataRes *self, memgraph::slk::Reader *reader);
+void Save(const memgraph::replication::DropAuthDataReq & /*self*/, memgraph::slk::Builder * /*builder*/);
+void Load(memgraph::replication::DropAuthDataReq * /*self*/, memgraph::slk::Reader * /*reader*/);
+}  // namespace memgraph::slk

src/communication/websocket/auth.hpp

@@ -1,4 +1,4 @@
-// Copyright 2022 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -14,8 +14,6 @@
 #include <string>
 
 #include "auth/auth.hpp"
-#include "utils/spin_lock.hpp"
-#include "utils/synchronized.hpp"
 
 namespace memgraph::communication::websocket {
 
@@ -30,7 +28,7 @@ class AuthenticationInterface {
 
 class SafeAuth : public AuthenticationInterface {
  public:
-  explicit SafeAuth(utils::Synchronized<auth::Auth, utils::WritePrioritizedRWLock> *auth) : auth_{auth} {}
+  explicit SafeAuth(auth::SynchedAuth *auth) : auth_{auth} {}
 
   bool Authenticate(const std::string &username, const std::string &password) const override;
 
@@ -39,6 +37,6 @@ class SafeAuth : public AuthenticationInterface {
   bool HasAnyUsers() const override;
 
  private:
-  utils::Synchronized<auth::Auth, utils::WritePrioritizedRWLock> *auth_;
+  auth::SynchedAuth *auth_;
 };
 }  // namespace memgraph::communication::websocket

src/coordination/CMakeLists.txt

@@ -12,8 +12,8 @@ target_sources(mg-coordination
         include/coordination/coordinator_slk.hpp
         include/coordination/coordinator_data.hpp
         include/coordination/constants.hpp
-        include/coordination/failover_status.hpp
         include/coordination/coordinator_cluster_config.hpp
+        include/coordination/coordinator_handlers.hpp
 
         PRIVATE
         coordinator_client.cpp
@@ -21,9 +21,11 @@ target_sources(mg-coordination
         coordinator_rpc.cpp
         coordinator_server.cpp
         coordinator_data.cpp
+        coordinator_instance.cpp
+        coordinator_handlers.cpp
 )
 target_include_directories(mg-coordination PUBLIC include)
 
 target_link_libraries(mg-coordination
-    PUBLIC mg::utils mg::rpc mg::slk mg::io mg::repl_coord_glue lib::rangev3 nuraft
+    PUBLIC mg::utils mg::rpc mg::slk mg::io mg::repl_coord_glue lib::rangev3 nuraft mg-replication_handler
 )

src/coordination/coordinator_client.cpp

@@ -20,7 +20,7 @@
 namespace memgraph::coordination {
 
 namespace {
-auto CreateClientContext(const memgraph::coordination::CoordinatorClientConfig &config)
+auto CreateClientContext(memgraph::coordination::CoordinatorClientConfig const &config)
     -> communication::ClientContext {
   return (config.ssl) ? communication::ClientContext{config.ssl->key_file, config.ssl->cert_file}
                       : communication::ClientContext{};
@@ -45,38 +45,33 @@ void CoordinatorClient::StartFrequentCheck() {
             "Health check frequency must be greater than 0");
 
   instance_checker_.Run(
-      "Coord checker", config_.health_check_frequency_sec, [this, instance_name = config_.instance_name] {
+      config_.instance_name, config_.health_check_frequency_sec, [this, instance_name = config_.instance_name] {
         try {
           spdlog::trace("Sending frequent heartbeat to machine {} on {}", instance_name,
                         rpc_client_.Endpoint().SocketAddress());
-          auto stream{rpc_client_.Stream<memgraph::replication_coordination_glue::FrequentHeartbeatRpc>()};
-          stream.AwaitResponse();
+          {  // NOTE: This is intentionally scoped so that stream lock could get released.
+            auto stream{rpc_client_.Stream<memgraph::replication_coordination_glue::FrequentHeartbeatRpc>()};
+            stream.AwaitResponse();
+          }
           succ_cb_(coord_data_, instance_name);
-        } catch (const rpc::RpcFailedException &) {
+        } catch (rpc::RpcFailedException const &) {
           fail_cb_(coord_data_, instance_name);
         }
       });
 }
 
 void CoordinatorClient::StopFrequentCheck() { instance_checker_.Stop(); }
-
 void CoordinatorClient::PauseFrequentCheck() { instance_checker_.Pause(); }
 void CoordinatorClient::ResumeFrequentCheck() { instance_checker_.Resume(); }
 
-auto CoordinatorClient::SetSuccCallback(HealthCheckCallback succ_cb) -> void { succ_cb_ = std::move(succ_cb); }
-auto CoordinatorClient::SetFailCallback(HealthCheckCallback fail_cb) -> void { fail_cb_ = std::move(fail_cb); }
-
-auto CoordinatorClient::ReplicationClientInfo() const -> const CoordinatorClientConfig::ReplicationClientInfo & {
-  return config_.replication_client_info;
+auto CoordinatorClient::SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void {
+  succ_cb_ = std::move(succ_cb);
+  fail_cb_ = std::move(fail_cb);
 }
 
-auto CoordinatorClient::ResetReplicationClientInfo() -> void {
-  // TODO (antoniofilipovic) Sync with Andi on this one
-  // config_.replication_client_info.reset();
-}
+auto CoordinatorClient::ReplicationClientInfo() const -> ReplClientInfo { return config_.replication_client_info; }
 
-auto CoordinatorClient::SendPromoteReplicaToMainRpc(
-    std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info) const -> bool {
+auto CoordinatorClient::SendPromoteReplicaToMainRpc(ReplicationClientsInfo replication_clients_info) const -> bool {
   try {
     auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(std::move(replication_clients_info))};
     if (!stream.AwaitResponse().success) {
@@ -84,23 +79,24 @@ auto CoordinatorClient::SendPromoteReplicaToMainRpc(
       return false;
     }
     return true;
-  } catch (const rpc::RpcFailedException &) {
+  } catch (rpc::RpcFailedException const &) {
     spdlog::error("RPC error occurred while sending failover RPC!");
   }
   return false;
 }
 
-auto CoordinatorClient::SendSetToReplicaRpc(CoordinatorClient::ReplClientInfo replication_client_info) const -> bool {
+auto CoordinatorClient::DemoteToReplica() const -> bool {
+  auto const &instance_name = config_.instance_name;
   try {
-    auto stream{rpc_client_.Stream<SetMainToReplicaRpc>(std::move(replication_client_info))};
+    auto stream{rpc_client_.Stream<DemoteMainToReplicaRpc>(config_.replication_client_info)};
     if (!stream.AwaitResponse().success) {
-      spdlog::error("Failed to set main to replica!");
+      spdlog::error("Failed to receive successful RPC response for setting instance {} to replica!", instance_name);
       return false;
     }
     spdlog::info("Sent request RPC from coordinator to instance to set it as replica!");
     return true;
-  } catch (const rpc::RpcFailedException &) {
-    spdlog::error("Failed to send failover RPC from coordinator to new main!");
+  } catch (rpc::RpcFailedException const &) {
+    spdlog::error("Failed to set instance {} to replica!", instance_name);
   }
   return false;
 }

src/coordination/coordinator_data.cpp

@@ -25,7 +25,7 @@ CoordinatorData::CoordinatorData() {
   auto find_instance = [](CoordinatorData *coord_data, std::string_view instance_name) -> CoordinatorInstance & {
     auto instance = std::ranges::find_if(
         coord_data->registered_instances_,
-        [instance_name](const CoordinatorInstance &instance) { return instance.InstanceName() == instance_name; });
+        [instance_name](CoordinatorInstance const &instance) { return instance.InstanceName() == instance_name; });
 
     MG_ASSERT(instance != coord_data->registered_instances_.end(), "Instance {} not found during callback!",
               instance_name);
@@ -35,105 +35,94 @@ CoordinatorData::CoordinatorData() {
   replica_succ_cb_ = [find_instance](CoordinatorData *coord_data, std::string_view instance_name) -> void {
     auto lock = std::lock_guard{coord_data->coord_data_lock_};
     spdlog::trace("Instance {} performing replica successful callback", instance_name);
-    auto &instance = find_instance(coord_data, instance_name);
-    MG_ASSERT(instance.IsReplica(), "Instance {} is not a replica!", instance_name);
-    instance.UpdateLastResponseTime();
+    find_instance(coord_data, instance_name).OnSuccessPing();
   };
 
   replica_fail_cb_ = [find_instance](CoordinatorData *coord_data, std::string_view instance_name) -> void {
     auto lock = std::lock_guard{coord_data->coord_data_lock_};
     spdlog::trace("Instance {} performing replica failure callback", instance_name);
-    auto &instance = find_instance(coord_data, instance_name);
-    MG_ASSERT(instance.IsReplica(), "Instance {} is not a replica!", instance_name);
-    instance.UpdateInstanceStatus();
+    find_instance(coord_data, instance_name).OnFailPing();
   };
 
   main_succ_cb_ = [find_instance](CoordinatorData *coord_data, std::string_view instance_name) -> void {
     auto lock = std::lock_guard{coord_data->coord_data_lock_};
     spdlog::trace("Instance {} performing main successful callback", instance_name);
+
     auto &instance = find_instance(coord_data, instance_name);
-    MG_ASSERT(instance.IsMain(), "Instance {} is not a main!", instance_name);
-    instance.UpdateLastResponseTime();
+
+    if (instance.IsAlive() || !coord_data->ClusterHasAliveMain_()) {
+      instance.OnSuccessPing();
+      return;
+    }
+
+    bool const demoted = instance.DemoteToReplica(coord_data->replica_succ_cb_, coord_data->replica_fail_cb_);
+    if (demoted) {
+      instance.OnSuccessPing();
+      spdlog::info("Instance {} demoted to replica", instance_name);
+    } else {
+      spdlog::error("Instance {} failed to become replica", instance_name);
+    }
   };
 
-  main_fail_cb_ = [this, find_instance](CoordinatorData *coord_data, std::string_view instance_name) -> void {
+  main_fail_cb_ = [find_instance](CoordinatorData *coord_data, std::string_view instance_name) -> void {
     auto lock = std::lock_guard{coord_data->coord_data_lock_};
     spdlog::trace("Instance {} performing main failure callback", instance_name);
-    auto &instance = find_instance(coord_data, instance_name);
-    MG_ASSERT(instance.IsMain(), "Instance {} is not a main!", instance_name);
-    if (bool main_alive = instance.UpdateInstanceStatus(); !main_alive) {
-      spdlog::info("Main instance {} is not alive, starting automatic failover", instance_name);
-      switch (auto failover_status = DoFailover(); failover_status) {
-        using enum DoFailoverStatus;
-        case ALL_REPLICAS_DOWN:
-          spdlog::warn("Failover aborted since all replicas are down!");
-          break;
-        case MAIN_ALIVE:
-          spdlog::warn("Failover aborted since main is alive!");
-          break;
-        case RPC_FAILED:
-          spdlog::warn("Failover aborted since promoting replica to main failed!");
-          break;
-        case SUCCESS:
-          break;
-      }
+    find_instance(coord_data, instance_name).OnFailPing();
+
+    if (!coord_data->ClusterHasAliveMain_()) {
+      spdlog::info("Cluster without main instance, trying automatic failover");
+      coord_data->TryFailover();
     }
   };
 }
 
-auto CoordinatorData::DoFailover() -> DoFailoverStatus {
-  using ReplicationClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
+auto CoordinatorData::ClusterHasAliveMain_() const -> bool {
+  auto const alive_main = [](CoordinatorInstance const &instance) { return instance.IsMain() && instance.IsAlive(); };
+  return std::ranges::any_of(registered_instances_, alive_main);
+}
 
+auto CoordinatorData::TryFailover() -> void {
   auto replica_instances = registered_instances_ | ranges::views::filter(&CoordinatorInstance::IsReplica);
 
   auto chosen_replica_instance = std::ranges::find_if(replica_instances, &CoordinatorInstance::IsAlive);
   if (chosen_replica_instance == replica_instances.end()) {
-    return DoFailoverStatus::ALL_REPLICAS_DOWN;
+    spdlog::warn("Failover failed since all replicas are down!");
+    return;
   }
 
-  chosen_replica_instance->PrepareForFailover();
+  chosen_replica_instance->PauseFrequentCheck();
+  utils::OnScopeExit scope_exit{[&chosen_replica_instance] { chosen_replica_instance->ResumeFrequentCheck(); }};
 
-  std::vector<ReplicationClientInfo> repl_clients_info;
+  std::vector<ReplClientInfo> repl_clients_info;
   repl_clients_info.reserve(std::ranges::distance(replica_instances));
 
-  auto const not_chosen_replica_instance = [&chosen_replica_instance](const CoordinatorInstance &instance) {
+  auto const not_chosen_replica_instance = [&chosen_replica_instance](CoordinatorInstance const &instance) {
     return instance != *chosen_replica_instance;
   };
-  auto const not_main = [](const CoordinatorInstance &instance) { return !instance.IsMain(); };
 
-  // TODO (antoniofilipovic): Should we send also data on old MAIN???
-  // TODO: (andi) Don't send replicas which aren't alive
-  for (const auto &unchosen_replica_instance :
-       replica_instances | ranges::views::filter(not_chosen_replica_instance) | ranges::views::filter(not_main)) {
-    repl_clients_info.emplace_back(unchosen_replica_instance.client_.ReplicationClientInfo());
+  std::ranges::transform(registered_instances_ | ranges::views::filter(not_chosen_replica_instance),
+                         std::back_inserter(repl_clients_info),
+                         [](const CoordinatorInstance &instance) { return instance.ReplicationClientInfo(); });
+
+  if (!chosen_replica_instance->PromoteToMain(std::move(repl_clients_info), main_succ_cb_, main_fail_cb_)) {
+    spdlog::warn("Failover failed since promoting replica to main failed!");
+    return;
   }
-
-  if (!chosen_replica_instance->client_.SendPromoteReplicaToMainRpc(std::move(repl_clients_info))) {
-    chosen_replica_instance->RestoreAfterFailedFailover();
-    return DoFailoverStatus::RPC_FAILED;
-  }
-
-  auto old_main = std::ranges::find_if(registered_instances_, &CoordinatorInstance::IsMain);
-  // TODO: (andi) For performing restoration we will have to improve this
-  old_main->client_.PauseFrequentCheck();
-
-  chosen_replica_instance->PostFailover(main_succ_cb_, main_fail_cb_);
-
-  return DoFailoverStatus::SUCCESS;
+  spdlog::info("Failover successful! Instance {} promoted to main.", chosen_replica_instance->InstanceName());
 }
 
 auto CoordinatorData::ShowInstances() const -> std::vector<CoordinatorInstanceStatus> {
   std::vector<CoordinatorInstanceStatus> instances_status;
   instances_status.reserve(registered_instances_.size());
 
-  auto const stringify_repl_role = [](const CoordinatorInstance &instance) -> std::string {
-    if (!instance.IsAlive()) return "";
+  auto const stringify_repl_role = [](CoordinatorInstance const &instance) -> std::string {
+    if (!instance.IsAlive()) return "unknown";
     if (instance.IsMain()) return "main";
     return "replica";
   };
 
   auto const instance_to_status =
-      [&stringify_repl_role](const CoordinatorInstance &instance) -> CoordinatorInstanceStatus {
+      [&stringify_repl_role](CoordinatorInstance const &instance) -> CoordinatorInstanceStatus {
     return {.instance_name = instance.InstanceName(),
             .socket_address = instance.SocketAddress(),
             .replication_role = stringify_repl_role(instance),
@@ -151,70 +140,59 @@ auto CoordinatorData::ShowInstances() const -> std::vector<CoordinatorInstanceSt
 auto CoordinatorData::SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus {
   auto lock = std::lock_guard{coord_data_lock_};
 
-  // Find replica we already registered
-  auto registered_replica = std::find_if(
-      registered_instances_.begin(), registered_instances_.end(),
-      [instance_name](const CoordinatorInstance &instance) { return instance.InstanceName() == instance_name; });
+  auto const is_new_main = [&instance_name](CoordinatorInstance const &instance) {
+    return instance.InstanceName() == instance_name;
+  };
+  auto new_main = std::ranges::find_if(registered_instances_, is_new_main);
 
-  // if replica not found...
-  if (registered_replica == registered_instances_.end()) {
-    spdlog::error("You didn't register instance with given name {}", instance_name);
+  if (new_main == registered_instances_.end()) {
+    spdlog::error("Instance {} not registered. Please register it using REGISTER INSTANCE {}", instance_name,
+                  instance_name);
     return SetInstanceToMainCoordinatorStatus::NO_INSTANCE_WITH_NAME;
   }
 
-  registered_replica->client_.PauseFrequentCheck();
+  new_main->PauseFrequentCheck();
+  utils::OnScopeExit scope_exit{[&new_main] { new_main->ResumeFrequentCheck(); }};
 
-  std::vector<CoordinatorClientConfig::ReplicationClientInfo> repl_clients_info;
+  ReplicationClientsInfo repl_clients_info;
   repl_clients_info.reserve(registered_instances_.size() - 1);
-  std::ranges::for_each(registered_instances_,
-                        [registered_replica, &repl_clients_info](const CoordinatorInstance &replica) {
-                          if (replica != *registered_replica) {
-                            repl_clients_info.emplace_back(replica.client_.ReplicationClientInfo());
-                          }
-                        });
 
-  // PROMOTE REPLICA TO MAIN
-  // THIS SHOULD FAIL HERE IF IT IS DOWN
-  if (auto result = registered_replica->client_.SendPromoteReplicaToMainRpc(std::move(repl_clients_info)); !result) {
-    registered_replica->client_.ResumeFrequentCheck();
+  auto const is_not_new_main = [&instance_name](CoordinatorInstance const &instance) {
+    return instance.InstanceName() != instance_name;
+  };
+  std::ranges::transform(registered_instances_ | ranges::views::filter(is_not_new_main),
+                         std::back_inserter(repl_clients_info),
+                         [](const CoordinatorInstance &instance) { return instance.ReplicationClientInfo(); });
+
+  if (!new_main->PromoteToMain(std::move(repl_clients_info), main_succ_cb_, main_fail_cb_)) {
     return SetInstanceToMainCoordinatorStatus::COULD_NOT_PROMOTE_TO_MAIN;
   }
 
-  registered_replica->client_.SetSuccCallback(main_succ_cb_);
-  registered_replica->client_.SetFailCallback(main_fail_cb_);
-  registered_replica->replication_role_ = replication_coordination_glue::ReplicationRole::MAIN;
-  registered_replica->client_.ResumeFrequentCheck();
-
+  spdlog::info("Instance {} promoted to main", instance_name);
   return SetInstanceToMainCoordinatorStatus::SUCCESS;
 }
 
 auto CoordinatorData::RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus {
   auto lock = std::lock_guard{coord_data_lock_};
-  if (std::ranges::any_of(registered_instances_, [&config](const CoordinatorInstance &instance) {
+  if (std::ranges::any_of(registered_instances_, [&config](CoordinatorInstance const &instance) {
         return instance.InstanceName() == config.instance_name;
       })) {
     return RegisterInstanceCoordinatorStatus::NAME_EXISTS;
   }
 
-  if (std::ranges::any_of(registered_instances_, [&config](const CoordinatorInstance &instance) {
-        spdlog::trace("Comparing {} with {}", instance.SocketAddress(), config.SocketAddress());
+  if (std::ranges::any_of(registered_instances_, [&config](CoordinatorInstance const &instance) {
         return instance.SocketAddress() == config.SocketAddress();
       })) {
-    return RegisterInstanceCoordinatorStatus::END_POINT_EXISTS;
+    return RegisterInstanceCoordinatorStatus::ENDPOINT_EXISTS;
   }
 
-  CoordinatorClientConfig::ReplicationClientInfo replication_client_info_copy = config.replication_client_info;
+  try {
+    registered_instances_.emplace_back(this, std::move(config), replica_succ_cb_, replica_fail_cb_);
+    return RegisterInstanceCoordinatorStatus::SUCCESS;
 
-  // TODO (antoniofilipovic) create and then push back
-  auto *instance = &registered_instances_.emplace_back(this, std::move(config), replica_succ_cb_, replica_fail_cb_,
-                                                       replication_coordination_glue::ReplicationRole::REPLICA);
-  if (auto res = instance->client_.SendSetToReplicaRpc(replication_client_info_copy); !res) {
+  } catch (CoordinatorRegisterInstanceException const &) {
     return RegisterInstanceCoordinatorStatus::RPC_FAILED;
   }
-
-  instance->client_.StartFrequentCheck();
-
-  return RegisterInstanceCoordinatorStatus::SUCCESS;
 }
 
 }  // namespace memgraph::coordination

src/coordination/coordinator_handlers.cpp

@@ -0,0 +1,123 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#ifdef MG_ENTERPRISE
+#include "coordination/coordinator_handlers.hpp"
+
+#include <range/v3/view.hpp>
+
+#include "coordination/coordinator_rpc.hpp"
+#include "coordination/include/coordination/coordinator_server.hpp"
+
+namespace memgraph::dbms {
+
+void CoordinatorHandlers::Register(memgraph::coordination::CoordinatorServer &server,
+                                   replication::ReplicationHandler &replication_handler) {
+  server.Register<coordination::PromoteReplicaToMainRpc>(
+      [&](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
+        spdlog::info("Received PromoteReplicaToMainRpc");
+        CoordinatorHandlers::PromoteReplicaToMainHandler(replication_handler, req_reader, res_builder);
+      });
+
+  server.Register<coordination::DemoteMainToReplicaRpc>(
+      [&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
+        spdlog::info("Received DemoteMainToReplicaRpc from coordinator server");
+        CoordinatorHandlers::DemoteMainToReplicaHandler(replication_handler, req_reader, res_builder);
+      });
+}
+
+void CoordinatorHandlers::DemoteMainToReplicaHandler(replication::ReplicationHandler &replication_handler,
+                                                     slk::Reader *req_reader, slk::Builder *res_builder) {
+  spdlog::info("Executing DemoteMainToReplicaHandler");
+
+  if (!replication_handler.IsMain()) {
+    spdlog::error("Setting to replica must be performed on main.");
+    slk::Save(coordination::DemoteMainToReplicaRes{false}, res_builder);
+    return;
+  }
+
+  coordination::DemoteMainToReplicaReq req;
+  slk::Load(&req, req_reader);
+
+  const replication::ReplicationServerConfig clients_config{
+      .ip_address = req.replication_client_info.replication_ip_address,
+      .port = req.replication_client_info.replication_port};
+
+  if (!replication_handler.SetReplicationRoleReplica(clients_config)) {
+    spdlog::error("Demoting main to replica failed!");
+    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+    return;
+  }
+
+  slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
+}
+
+void CoordinatorHandlers::PromoteReplicaToMainHandler(replication::ReplicationHandler &replication_handler,
+                                                      slk::Reader *req_reader, slk::Builder *res_builder) {
+  if (!replication_handler.IsReplica()) {
+    spdlog::error("Failover must be performed on replica!");
+    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+    return;
+  }
+
+  // This can fail because of disk. If it does, the cluster state could get inconsistent.
+  // We don't handle disk issues.
+  if (!replication_handler.DoReplicaToMainPromotion()) {
+    spdlog::error("Promoting replica to main failed!");
+    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+    return;
+  }
+
+  coordination::PromoteReplicaToMainReq req;
+  slk::Load(&req, req_reader);
+
+  auto const converter = [](const auto &repl_info_config) {
+    return replication::ReplicationClientConfig{
+        .name = repl_info_config.instance_name,
+        .mode = repl_info_config.replication_mode,
+        .ip_address = repl_info_config.replication_ip_address,
+        .port = repl_info_config.replication_port,
+    };
+  };
+
+  // registering replicas
+  for (auto const &config : req.replication_clients_info | ranges::views::transform(converter)) {
+    auto instance_client = replication_handler.RegisterReplica(config);
+    if (instance_client.HasError()) {
+      using enum memgraph::replication::RegisterReplicaError;
+      switch (instance_client.GetError()) {
+        // Can't happen, checked on the coordinator side
+        case memgraph::query::RegisterReplicaError::NAME_EXISTS:
+          spdlog::error("Replica with the same name already exists!");
+          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+          return;
+        // Can't happen, checked on the coordinator side
+        case memgraph::query::RegisterReplicaError::ENDPOINT_EXISTS:
+          spdlog::error("Replica with the same endpoint already exists!");
+          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+          return;
+        // We don't handle disk issues
+        case memgraph::query::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
+          spdlog::error("Registered replica could not be persisted!");
+          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+          return;
+        case memgraph::query::RegisterReplicaError::CONNECTION_FAILED:
+          // Connection failure is not a fatal error
+          break;
+      }
+    }
+  }
+
+  slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
+}
+
+}  // namespace memgraph::dbms
+#endif

src/coordination/coordinator_instance.cpp

@@ -0,0 +1,84 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#ifdef MG_ENTERPRISE
+
+#include "coordination/coordinator_instance.hpp"
+
+namespace memgraph::coordination {
+
+CoordinatorInstance::CoordinatorInstance(CoordinatorData *data, CoordinatorClientConfig config,
+                                         HealthCheckCallback succ_cb, HealthCheckCallback fail_cb)
+    : client_(data, std::move(config), std::move(succ_cb), std::move(fail_cb)),
+      replication_role_(replication_coordination_glue::ReplicationRole::REPLICA),
+      is_alive_(true) {
+  if (!client_.DemoteToReplica()) {
+    throw CoordinatorRegisterInstanceException("Failed to demote instance {} to replica", client_.InstanceName());
+  }
+  client_.StartFrequentCheck();
+}
+
+auto CoordinatorInstance::OnSuccessPing() -> void {
+  last_response_time_ = std::chrono::system_clock::now();
+  is_alive_ = true;
+}
+
+auto CoordinatorInstance::OnFailPing() -> bool {
+  is_alive_ =
+      std::chrono::duration_cast<std::chrono::seconds>(std::chrono::system_clock::now() - last_response_time_).count() <
+      CoordinatorClusterConfig::alive_response_time_difference_sec_;
+  return is_alive_;
+}
+
+auto CoordinatorInstance::InstanceName() const -> std::string { return client_.InstanceName(); }
+auto CoordinatorInstance::SocketAddress() const -> std::string { return client_.SocketAddress(); }
+auto CoordinatorInstance::IsAlive() const -> bool { return is_alive_; }
+
+auto CoordinatorInstance::IsReplica() const -> bool {
+  return replication_role_ == replication_coordination_glue::ReplicationRole::REPLICA;
+}
+auto CoordinatorInstance::IsMain() const -> bool {
+  return replication_role_ == replication_coordination_glue::ReplicationRole::MAIN;
+}
+
+auto CoordinatorInstance::PromoteToMain(ReplicationClientsInfo repl_clients_info, HealthCheckCallback main_succ_cb,
+                                        HealthCheckCallback main_fail_cb) -> bool {
+  if (!client_.SendPromoteReplicaToMainRpc(std::move(repl_clients_info))) {
+    return false;
+  }
+
+  replication_role_ = replication_coordination_glue::ReplicationRole::MAIN;
+  client_.SetCallbacks(std::move(main_succ_cb), std::move(main_fail_cb));
+
+  return true;
+}
+
+auto CoordinatorInstance::DemoteToReplica(HealthCheckCallback replica_succ_cb, HealthCheckCallback replica_fail_cb)
+    -> bool {
+  if (!client_.DemoteToReplica()) {
+    return false;
+  }
+
+  replication_role_ = replication_coordination_glue::ReplicationRole::REPLICA;
+  client_.SetCallbacks(std::move(replica_succ_cb), std::move(replica_fail_cb));
+
+  return true;
+}
+
+auto CoordinatorInstance::PauseFrequentCheck() -> void { client_.PauseFrequentCheck(); }
+auto CoordinatorInstance::ResumeFrequentCheck() -> void { client_.ResumeFrequentCheck(); }
+
+auto CoordinatorInstance::ReplicationClientInfo() const -> CoordinatorClientConfig::ReplicationClientInfo {
+  return client_.ReplicationClientInfo();
+}
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/coordinator_rpc.cpp

@@ -36,19 +36,19 @@ void PromoteReplicaToMainRes::Load(PromoteReplicaToMainRes *self, memgraph::slk:
   memgraph::slk::Load(self, reader);
 }
 
-void SetMainToReplicaReq::Save(const SetMainToReplicaReq &self, memgraph::slk::Builder *builder) {
+void DemoteMainToReplicaReq::Save(const DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder) {
   memgraph::slk::Save(self, builder);
 }
 
-void SetMainToReplicaReq::Load(SetMainToReplicaReq *self, memgraph::slk::Reader *reader) {
+void DemoteMainToReplicaReq::Load(DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader) {
   memgraph::slk::Load(self, reader);
 }
 
-void SetMainToReplicaRes::Save(const SetMainToReplicaRes &self, memgraph::slk::Builder *builder) {
+void DemoteMainToReplicaRes::Save(const DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder) {
   memgraph::slk::Save(self, builder);
 }
 
-void SetMainToReplicaRes::Load(SetMainToReplicaRes *self, memgraph::slk::Reader *reader) {
+void DemoteMainToReplicaRes::Load(DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader) {
   memgraph::slk::Load(self, reader);
 }
 
@@ -60,11 +60,11 @@ constexpr utils::TypeInfo coordination::PromoteReplicaToMainReq::kType{utils::Ty
 constexpr utils::TypeInfo coordination::PromoteReplicaToMainRes::kType{utils::TypeId::COORD_FAILOVER_RES,
                                                                        "CoordPromoteReplicaToMainRes", nullptr};
 
-constexpr utils::TypeInfo coordination::SetMainToReplicaReq::kType{utils::TypeId::COORD_SET_REPL_MAIN_REQ,
-                                                                   "CoordSetReplMainReq", nullptr};
+constexpr utils::TypeInfo coordination::DemoteMainToReplicaReq::kType{utils::TypeId::COORD_SET_REPL_MAIN_REQ,
+                                                                      "CoordDemoteToReplicaReq", nullptr};
 
-constexpr utils::TypeInfo coordination::SetMainToReplicaRes::kType{utils::TypeId::COORD_SET_REPL_MAIN_RES,
-                                                                   "CoordSetReplMainRes", nullptr};
+constexpr utils::TypeInfo coordination::DemoteMainToReplicaRes::kType{utils::TypeId::COORD_SET_REPL_MAIN_RES,
+                                                                      "CoordDemoteToReplicaRes", nullptr};
 
 namespace slk {
 
@@ -84,19 +84,19 @@ void Load(memgraph::coordination::PromoteReplicaToMainReq *self, memgraph::slk::
   memgraph::slk::Load(&self->replication_clients_info, reader);
 }
 
-void Save(const memgraph::coordination::SetMainToReplicaReq &self, memgraph::slk::Builder *builder) {
+void Save(const memgraph::coordination::DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder) {
   memgraph::slk::Save(self.replication_client_info, builder);
 }
 
-void Load(memgraph::coordination::SetMainToReplicaReq *self, memgraph::slk::Reader *reader) {
+void Load(memgraph::coordination::DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader) {
   memgraph::slk::Load(&self->replication_client_info, reader);
 }
 
-void Save(const memgraph::coordination::SetMainToReplicaRes &self, memgraph::slk::Builder *builder) {
+void Save(const memgraph::coordination::DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder) {
   memgraph::slk::Save(self.success, builder);
 }
 
-void Load(memgraph::coordination::SetMainToReplicaRes *self, memgraph::slk::Reader *reader) {
+void Load(memgraph::coordination::DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader) {
   memgraph::slk::Load(&self->success, reader);
 }
 

src/coordination/coordinator_state.cpp

@@ -74,12 +74,6 @@ auto CoordinatorState::ShowInstances() const -> std::vector<CoordinatorInstanceS
   return std::get<CoordinatorData>(data_).ShowInstances();
 }
 
-[[nodiscard]] auto CoordinatorState::DoFailover() -> DoFailoverStatus {
-  MG_ASSERT(std::holds_alternative<CoordinatorData>(data_), "Cannot do failover since variant holds wrong alternative");
-  auto &coord_state = std::get<CoordinatorData>(data_);
-  return coord_state.DoFailover();
-}
-
 auto CoordinatorState::GetCoordinatorServer() const -> CoordinatorServer & {
   MG_ASSERT(std::holds_alternative<CoordinatorMainReplicaData>(data_),
             "Cannot get coordinator server since variant holds wrong alternative");

src/coordination/include/coordination/coordinator_client.hpp

@@ -21,12 +21,10 @@ namespace memgraph::coordination {
 
 class CoordinatorData;
 using HealthCheckCallback = std::function<void(CoordinatorData *, std::string_view)>;
+using ReplicationClientsInfo = std::vector<ReplClientInfo>;
 
 class CoordinatorClient {
  public:
-  using ReplClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
-  using ReplicationClientsInfo = std::vector<ReplClientInfo>;
-
   explicit CoordinatorClient(CoordinatorData *coord_data_, CoordinatorClientConfig config, HealthCheckCallback succ_cb,
                              HealthCheckCallback fail_cb);
 
@@ -46,15 +44,12 @@ class CoordinatorClient {
   auto InstanceName() const -> std::string;
   auto SocketAddress() const -> std::string;
 
-  auto SendPromoteReplicaToMainRpc(ReplicationClientsInfo replication_clients_info) const -> bool;
+  [[nodiscard]] auto SendPromoteReplicaToMainRpc(ReplicationClientsInfo replication_clients_info) const -> bool;
+  [[nodiscard]] auto DemoteToReplica() const -> bool;
 
-  auto ReplicationClientInfo() const -> const ReplClientInfo &;
-  auto ResetReplicationClientInfo() -> void;
+  auto ReplicationClientInfo() const -> ReplClientInfo;
 
-  auto SendSetToReplicaRpc(ReplClientInfo replication_client_info) const -> bool;
-
-  auto SetSuccCallback(HealthCheckCallback succ_cb) -> void;
-  auto SetFailCallback(HealthCheckCallback fail_cb) -> void;
+  auto SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void;
 
   friend bool operator==(CoordinatorClient const &first, CoordinatorClient const &second) {
     return first.config_ == second.config_;

src/coordination/include/coordination/coordinator_config.hpp

@@ -32,9 +32,7 @@ struct CoordinatorClientConfig {
 
   auto SocketAddress() const -> std::string { return ip_address + ":" + std::to_string(port); }
 
-  // Info which coordinator will send to new main when performing failover
   struct ReplicationClientInfo {
-    // Must be the same as CoordinatorClientConfig's instance_name
     std::string instance_name;
     replication_coordination_glue::ReplicationMode replication_mode{};
     std::string replication_ip_address;
@@ -43,7 +41,6 @@ struct CoordinatorClientConfig {
     friend bool operator==(ReplicationClientInfo const &, ReplicationClientInfo const &) = default;
   };
 
-  // Each instance has replication config in case it fails
   ReplicationClientInfo replication_client_info;
 
   struct SSL {
@@ -58,6 +55,8 @@ struct CoordinatorClientConfig {
   friend bool operator==(CoordinatorClientConfig const &, CoordinatorClientConfig const &) = default;
 };
 
+using ReplClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
+
 struct CoordinatorServerConfig {
   std::string ip_address;
   uint16_t port{};

src/coordination/include/coordination/coordinator_data.hpp

@@ -16,9 +16,9 @@
 #include "coordination/coordinator_instance.hpp"
 #include "coordination/coordinator_instance_status.hpp"
 #include "coordination/coordinator_server.hpp"
-#include "coordination/failover_status.hpp"
 #include "coordination/register_main_replica_coordinator_status.hpp"
 #include "utils/rw_lock.hpp"
+#include "utils/thread_pool.hpp"
 
 #include <list>
 
@@ -27,17 +27,20 @@ class CoordinatorData {
  public:
   CoordinatorData();
 
-  [[nodiscard]] auto DoFailover() -> DoFailoverStatus;
-
   [[nodiscard]] auto RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
+
   [[nodiscard]] auto SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
 
+  auto TryFailover() -> void;
+
   auto ShowInstances() const -> std::vector<CoordinatorInstanceStatus>;
 
  private:
+  auto ClusterHasAliveMain_() const -> bool;
+
   mutable utils::RWLock coord_data_lock_{utils::RWLock::Priority::READ};
   HealthCheckCallback main_succ_cb_, main_fail_cb_, replica_succ_cb_, replica_fail_cb_;
-  // Must be std::list because we rely on pointer stability
+  // NOTE: Must be std::list because we rely on pointer stability
   std::list<CoordinatorInstance> registered_instances_;
 };
 

src/coordination/include/coordination/coordinator_exceptions.hpp

@@ -16,16 +16,16 @@
 #include "utils/exceptions.hpp"
 
 namespace memgraph::coordination {
-class CoordinatorFailoverException final : public utils::BasicException {
+class CoordinatorRegisterInstanceException final : public utils::BasicException {
  public:
-  explicit CoordinatorFailoverException(const std::string_view what) noexcept
-      : BasicException("Failover didn't complete successfully: " + std::string(what)) {}
+  explicit CoordinatorRegisterInstanceException(const std::string_view what) noexcept
+      : BasicException("Failed to create instance: " + std::string(what)) {}
 
   template <class... Args>
-  explicit CoordinatorFailoverException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
-      : CoordinatorFailoverException(fmt::format(fmt, std::forward<Args>(args)...)) {}
+  explicit CoordinatorRegisterInstanceException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
+      : CoordinatorRegisterInstanceException(fmt::format(fmt, std::forward<Args>(args)...)) {}
 
-  SPECIALIZE_GET_EXCEPTION_NAME(CoordinatorFailoverException)
+  SPECIALIZE_GET_EXCEPTION_NAME(CoordinatorRegisterInstanceException)
 };
 
 }  // namespace memgraph::coordination

src/coordination/include/coordination/coordinator_handlers.hpp

@@ -13,7 +13,9 @@
 
 #ifdef MG_ENTERPRISE
 
-#include "slk/serialization.hpp"
+#include "coordination/coordinator_server.hpp"
+#include "replication_handler/replication_handler.hpp"
+#include "slk/streams.hpp"
 
 namespace memgraph::dbms {
 
@@ -21,12 +23,14 @@ class DbmsHandler;
 
 class CoordinatorHandlers {
  public:
-  static void Register(DbmsHandler &dbms_handler);
+  static void Register(memgraph::coordination::CoordinatorServer &server,
+                       replication::ReplicationHandler &replication_handler);
 
  private:
-  static void PromoteReplicaToMainHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader,
+  static void PromoteReplicaToMainHandler(replication::ReplicationHandler &replication_handler, slk::Reader *req_reader,
                                           slk::Builder *res_builder);
-  static void SetMainToReplicaHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void DemoteMainToReplicaHandler(replication::ReplicationHandler &replication_handler, slk::Reader *req_reader,
+                                         slk::Builder *res_builder);
 };
 
 }  // namespace memgraph::dbms

src/coordination/include/coordination/coordinator_instance.hpp

@@ -15,6 +15,7 @@
 
 #include "coordination/coordinator_client.hpp"
 #include "coordination/coordinator_cluster_config.hpp"
+#include "coordination/coordinator_exceptions.hpp"
 #include "replication_coordination_glue/role.hpp"
 
 namespace memgraph::coordination {
@@ -24,10 +25,7 @@ class CoordinatorData;
 class CoordinatorInstance {
  public:
   CoordinatorInstance(CoordinatorData *data, CoordinatorClientConfig config, HealthCheckCallback succ_cb,
-                      HealthCheckCallback fail_cb, replication_coordination_glue::ReplicationRole replication_role)
-      : client_(data, std::move(config), std::move(succ_cb), std::move(fail_cb)),
-        replication_role_(replication_role),
-        is_alive_(true) {}
+                      HealthCheckCallback fail_cb);
 
   CoordinatorInstance(CoordinatorInstance const &other) = delete;
   CoordinatorInstance &operator=(CoordinatorInstance const &other) = delete;
@@ -35,34 +33,27 @@ class CoordinatorInstance {
   CoordinatorInstance &operator=(CoordinatorInstance &&other) noexcept = delete;
   ~CoordinatorInstance() = default;
 
-  auto UpdateInstanceStatus() -> bool {
-    is_alive_ = std::chrono::duration_cast<std::chrono::seconds>(std::chrono::system_clock::now() - last_response_time_)
-                    .count() < CoordinatorClusterConfig::alive_response_time_difference_sec_;
-    return is_alive_;
-  }
-  auto UpdateLastResponseTime() -> void { last_response_time_ = std::chrono::system_clock::now(); }
+  auto OnSuccessPing() -> void;
+  auto OnFailPing() -> bool;
 
-  auto InstanceName() const -> std::string { return client_.InstanceName(); }
-  auto SocketAddress() const -> std::string { return client_.SocketAddress(); }
-  auto IsAlive() const -> bool { return is_alive_; }
+  auto IsAlive() const -> bool;
 
-  auto IsReplica() const -> bool {
-    return replication_role_ == replication_coordination_glue::ReplicationRole::REPLICA;
-  }
-  auto IsMain() const -> bool { return replication_role_ == replication_coordination_glue::ReplicationRole::MAIN; }
+  auto InstanceName() const -> std::string;
+  auto SocketAddress() const -> std::string;
 
-  auto PrepareForFailover() -> void { client_.PauseFrequentCheck(); }
-  auto RestoreAfterFailedFailover() -> void { client_.ResumeFrequentCheck(); }
+  auto IsReplica() const -> bool;
+  auto IsMain() const -> bool;
 
-  auto PostFailover(HealthCheckCallback main_succ_cb, HealthCheckCallback main_fail_cb) -> void {
-    replication_role_ = replication_coordination_glue::ReplicationRole::MAIN;
-    client_.SetSuccCallback(std::move(main_succ_cb));
-    client_.SetFailCallback(std::move(main_fail_cb));
-    // Comment with Andi but we shouldn't delete this, what if this MAIN FAILS AGAIN
-    // client_.ResetReplicationClientInfo();
-    client_.ResumeFrequentCheck();
-  }
+  auto PromoteToMain(ReplicationClientsInfo repl_clients_info, HealthCheckCallback main_succ_cb,
+                     HealthCheckCallback main_fail_cb) -> bool;
+  auto DemoteToReplica(HealthCheckCallback replica_succ_cb, HealthCheckCallback replica_fail_cb) -> bool;
 
+  auto PauseFrequentCheck() -> void;
+  auto ResumeFrequentCheck() -> void;
+
+  auto ReplicationClientInfo() const -> ReplClientInfo;
+
+ private:
   CoordinatorClient client_;
   replication_coordination_glue::ReplicationRole replication_role_;
   std::chrono::system_clock::time_point last_response_time_{};

src/coordination/include/coordination/coordinator_rpc.hpp

@@ -48,35 +48,35 @@ struct PromoteReplicaToMainRes {
 
 using PromoteReplicaToMainRpc = rpc::RequestResponse<PromoteReplicaToMainReq, PromoteReplicaToMainRes>;
 
-struct SetMainToReplicaReq {
+struct DemoteMainToReplicaReq {
   static const utils::TypeInfo kType;
   static const utils::TypeInfo &GetTypeInfo() { return kType; }
 
-  static void Load(SetMainToReplicaReq *self, memgraph::slk::Reader *reader);
-  static void Save(const SetMainToReplicaReq &self, memgraph::slk::Builder *builder);
+  static void Load(DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader);
+  static void Save(const DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder);
 
-  explicit SetMainToReplicaReq(CoordinatorClientConfig::ReplicationClientInfo replication_client_info)
+  explicit DemoteMainToReplicaReq(CoordinatorClientConfig::ReplicationClientInfo replication_client_info)
       : replication_client_info(std::move(replication_client_info)) {}
 
-  SetMainToReplicaReq() = default;
+  DemoteMainToReplicaReq() = default;
 
   CoordinatorClientConfig::ReplicationClientInfo replication_client_info;
 };
 
-struct SetMainToReplicaRes {
+struct DemoteMainToReplicaRes {
   static const utils::TypeInfo kType;
   static const utils::TypeInfo &GetTypeInfo() { return kType; }
 
-  static void Load(SetMainToReplicaRes *self, memgraph::slk::Reader *reader);
-  static void Save(const SetMainToReplicaRes &self, memgraph::slk::Builder *builder);
+  static void Load(DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader);
+  static void Save(const DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder);
 
-  explicit SetMainToReplicaRes(bool success) : success(success) {}
-  SetMainToReplicaRes() = default;
+  explicit DemoteMainToReplicaRes(bool success) : success(success) {}
+  DemoteMainToReplicaRes() = default;
 
   bool success;
 };
 
-using SetMainToReplicaRpc = rpc::RequestResponse<SetMainToReplicaReq, SetMainToReplicaRes>;
+using DemoteMainToReplicaRpc = rpc::RequestResponse<DemoteMainToReplicaReq, DemoteMainToReplicaRes>;
 
 }  // namespace memgraph::coordination
 
@@ -91,13 +91,13 @@ void Save(const memgraph::coordination::PromoteReplicaToMainReq &self, memgraph:
 
 void Load(memgraph::coordination::PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader);
 
-void Save(const memgraph::coordination::SetMainToReplicaRes &self, memgraph::slk::Builder *builder);
+void Save(const memgraph::coordination::DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder);
 
-void Load(memgraph::coordination::SetMainToReplicaRes *self, memgraph::slk::Reader *reader);
+void Load(memgraph::coordination::DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader);
 
-void Save(const memgraph::coordination::SetMainToReplicaReq &self, memgraph::slk::Builder *builder);
+void Save(const memgraph::coordination::DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder);
 
-void Load(memgraph::coordination::SetMainToReplicaReq *self, memgraph::slk::Reader *reader);
+void Load(memgraph::coordination::DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader);
 
 }  // namespace memgraph::slk
 

src/coordination/include/coordination/coordinator_state.hpp

@@ -16,7 +16,6 @@
 #include "coordination/coordinator_data.hpp"
 #include "coordination/coordinator_instance_status.hpp"
 #include "coordination/coordinator_server.hpp"
-#include "coordination/failover_status.hpp"
 #include "coordination/register_main_replica_coordinator_status.hpp"
 
 #include <variant>
@@ -28,8 +27,8 @@ class CoordinatorState {
   CoordinatorState();
   ~CoordinatorState() = default;
 
-  CoordinatorState(const CoordinatorState &) = delete;
-  CoordinatorState &operator=(const CoordinatorState &) = delete;
+  CoordinatorState(CoordinatorState const &) = delete;
+  CoordinatorState &operator=(CoordinatorState const &) = delete;
 
   CoordinatorState(CoordinatorState &&) noexcept = delete;
   CoordinatorState &operator=(CoordinatorState &&) noexcept = delete;
@@ -43,8 +42,6 @@ class CoordinatorState {
   // The client code must check that the server exists before calling this method.
   auto GetCoordinatorServer() const -> CoordinatorServer &;
 
-  [[nodiscard]] auto DoFailover() -> DoFailoverStatus;
-
  private:
   std::variant<CoordinatorData, CoordinatorMainReplicaData> data_;
 };

src/coordination/include/coordination/register_main_replica_coordinator_status.hpp

@@ -19,8 +19,7 @@ namespace memgraph::coordination {
 
 enum class RegisterInstanceCoordinatorStatus : uint8_t {
   NAME_EXISTS,
-  END_POINT_EXISTS,
-  COULD_NOT_BE_PERSISTED,
+  ENDPOINT_EXISTS,
   NOT_COORDINATOR,
   RPC_FAILED,
   SUCCESS

src/dbms/CMakeLists.txt

@@ -1,2 +1,10 @@
-add_library(mg-dbms STATIC dbms_handler.cpp database.cpp replication_handler.cpp coordinator_handler.cpp replication_client.cpp inmemory/replication_handlers.cpp coordinator_handlers.cpp)
-target_link_libraries(mg-dbms mg-utils mg-storage-v2 mg-query mg-replication mg-coordination)
+add_library(mg-dbms STATIC
+        dbms_handler.cpp
+        database.cpp
+        coordinator_handler.cpp
+        inmemory/replication_handlers.cpp
+        replication_handlers.cpp
+        rpc.cpp
+
+)
+target_link_libraries(mg-dbms mg-utils mg-storage-v2 mg-query mg-auth mg-replication mg-coordination)

src/dbms/coordinator_handler.cpp

@@ -18,20 +18,21 @@
 
 namespace memgraph::dbms {
 
-CoordinatorHandler::CoordinatorHandler(DbmsHandler &dbms_handler) : dbms_handler_(dbms_handler) {}
+CoordinatorHandler::CoordinatorHandler(coordination::CoordinatorState &coordinator_state)
+    : coordinator_state_(coordinator_state) {}
 
 auto CoordinatorHandler::RegisterInstance(memgraph::coordination::CoordinatorClientConfig config)
     -> coordination::RegisterInstanceCoordinatorStatus {
-  return dbms_handler_.CoordinatorState().RegisterInstance(config);
+  return coordinator_state_.RegisterInstance(config);
 }
 
 auto CoordinatorHandler::SetInstanceToMain(std::string instance_name)
     -> coordination::SetInstanceToMainCoordinatorStatus {
-  return dbms_handler_.CoordinatorState().SetInstanceToMain(std::move(instance_name));
+  return coordinator_state_.SetInstanceToMain(std::move(instance_name));
 }
 
 auto CoordinatorHandler::ShowInstances() const -> std::vector<coordination::CoordinatorInstanceStatus> {
-  return dbms_handler_.CoordinatorState().ShowInstances();
+  return coordinator_state_.ShowInstances();
 }
 }  // namespace memgraph::dbms
 

src/dbms/coordinator_handler.hpp

@@ -13,15 +13,11 @@
 
 #ifdef MG_ENTERPRISE
 
-#include "utils/result.hpp"
-
 #include "coordination/coordinator_config.hpp"
 #include "coordination/coordinator_instance_status.hpp"
-#include "coordination/failover_status.hpp"
+#include "coordination/coordinator_state.hpp"
 #include "coordination/register_main_replica_coordinator_status.hpp"
 
-#include <cstdint>
-#include <optional>
 #include <vector>
 
 namespace memgraph::dbms {
@@ -30,7 +26,7 @@ class DbmsHandler;
 
 class CoordinatorHandler {
  public:
-  explicit CoordinatorHandler(DbmsHandler &dbms_handler);
+  explicit CoordinatorHandler(coordination::CoordinatorState &coordinator_state);
 
   auto RegisterInstance(coordination::CoordinatorClientConfig config)
       -> coordination::RegisterInstanceCoordinatorStatus;
@@ -40,7 +36,7 @@ class CoordinatorHandler {
   auto ShowInstances() const -> std::vector<coordination::CoordinatorInstanceStatus>;
 
  private:
-  DbmsHandler &dbms_handler_;
+  coordination::CoordinatorState &coordinator_state_;
 };
 
 }  // namespace memgraph::dbms

src/dbms/coordinator_handlers.cpp

@@ -1,153 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-#ifdef MG_ENTERPRISE
-
-#include "dbms/coordinator_handlers.hpp"
-#include "dbms/utils.hpp"
-
-#include "coordination/coordinator_exceptions.hpp"
-#include "coordination/coordinator_rpc.hpp"
-#include "dbms/dbms_handler.hpp"
-#include "dbms/replication_client.hpp"
-
-#include "range/v3/view.hpp"
-
-namespace memgraph::dbms {
-
-void CoordinatorHandlers::Register(DbmsHandler &dbms_handler) {
-  auto &server = dbms_handler.CoordinatorState().GetCoordinatorServer();
-
-  server.Register<coordination::PromoteReplicaToMainRpc>(
-      [&dbms_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
-        spdlog::info("Received PromoteReplicaToMainRpc");
-        CoordinatorHandlers::PromoteReplicaToMainHandler(dbms_handler, req_reader, res_builder);
-      });
-
-  server.Register<coordination::SetMainToReplicaRpc>(
-      [&dbms_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
-        spdlog::info("Received SetMainToReplicaRpc from coordinator server");
-        CoordinatorHandlers::SetMainToReplicaHandler(dbms_handler, req_reader, res_builder);
-      });
-}
-
-void CoordinatorHandlers::SetMainToReplicaHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader,
-                                                  slk::Builder *res_builder) {
-  auto &repl_state = dbms_handler.ReplicationState();
-
-  if (!repl_state.IsMain()) {
-    spdlog::error("Setting to replica must be performed on main.");
-    slk::Save(coordination::SetMainToReplicaRes{false}, res_builder);
-    return;
-  }
-
-  coordination::SetMainToReplicaReq req;
-  slk::Load(&req, req_reader);
-
-  replication::ReplicationServerConfig clients_config{.ip_address = req.replication_client_info.replication_ip_address,
-                                                      .port = req.replication_client_info.replication_port};
-
-  if (bool success = memgraph::dbms::SetReplicationRoleReplica(dbms_handler, clients_config); !success) {
-    spdlog::error("Setting main to replica failed!");
-    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-    return;
-  }
-
-  slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
-}
-
-void CoordinatorHandlers::PromoteReplicaToMainHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader,
-                                                      slk::Builder *res_builder) {
-  auto &repl_state = dbms_handler.ReplicationState();
-
-  if (!repl_state.IsReplica()) {
-    spdlog::error("Failover must be performed on replica!");
-    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-    return;
-  }
-
-  auto repl_server_config = std::get<replication::RoleReplicaData>(repl_state.ReplicationData()).config;
-
-  // This can fail because of disk. If it does, the cluster state could get inconsistent.
-  // We don't handle disk issues.
-  if (bool success = memgraph::dbms::DoReplicaToMainPromotion(dbms_handler); !success) {
-    spdlog::error("Promoting replica to main failed!");
-    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-    return;
-  }
-
-  coordination::PromoteReplicaToMainReq req;
-  slk::Load(&req, req_reader);
-
-  auto const converter = [](const auto &repl_info_config) {
-    return replication::ReplicationClientConfig{
-        .name = repl_info_config.instance_name,
-        .mode = repl_info_config.replication_mode,
-        .ip_address = repl_info_config.replication_ip_address,
-        .port = repl_info_config.replication_port,
-    };
-  };
-
-  MG_ASSERT(
-      std::get<replication::RoleMainData>(repl_state.ReplicationData()).registered_replicas_.empty(),
-      "No replicas should be registered after promoting replica to main and before registering replication clients!");
-
-  // registering replicas
-  for (auto const &config : req.replication_clients_info | ranges::views::transform(converter)) {
-    auto instance_client = repl_state.RegisterReplica(config);
-    if (instance_client.HasError()) {
-      switch (instance_client.GetError()) {
-        // Can't happen, we are already replica
-        case memgraph::replication::RegisterReplicaError::NOT_MAIN:
-          spdlog::error("Failover must be performed to main!");
-          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-          return;
-        // Can't happen, checked on the coordinator side
-        case memgraph::replication::RegisterReplicaError::NAME_EXISTS:
-          spdlog::error("Replica with the same name already exists!");
-          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-          return;
-        // Can't happen, checked on the coordinator side
-        case memgraph::replication::RegisterReplicaError::ENDPOINT_EXISTS:
-          spdlog::error("Replica with the same endpoint already exists!");
-          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-          return;
-        // We don't handle disk issues
-        case memgraph::replication::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
-          spdlog::error("Registered replica could not be persisted!");
-          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
-          return;
-        case memgraph::replication::RegisterReplicaError::SUCCESS:
-          break;
-      }
-    }
-    if (!allow_mt_repl && dbms_handler.All().size() > 1) {
-      spdlog::warn("Multi-tenant replication is currently not supported!");
-    }
-
-    auto &instance_client_ref = *instance_client.GetValue();
-
-    // Update system before enabling individual storage <-> replica clients
-    dbms_handler.SystemRestore(instance_client_ref);
-
-    // TODO: (andi) Policy for register all databases
-    // Will be resolved after deciding about choosing new replica
-    const bool all_clients_good = memgraph::dbms::RegisterAllDatabasesClients(dbms_handler, instance_client_ref);
-    MG_ASSERT(all_clients_good, "Failed to register one or more databases to the REPLICA \"{}\".", config.name);
-
-    StartReplicaClient(dbms_handler, instance_client_ref);
-  }
-
-  slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
-}
-
-}  // namespace memgraph::dbms
-#endif

src/dbms/database.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -11,10 +11,7 @@
 
 #include "dbms/database.hpp"
 #include "dbms/inmemory/storage_helper.hpp"
-#include "dbms/replication_handler.hpp"
-#include "flags/storage_mode.hpp"
 #include "storage/v2/disk/storage.hpp"
-#include "storage/v2/inmemory/storage.hpp"
 #include "storage/v2/storage_mode.hpp"
 
 template struct memgraph::utils::Gatekeeper<memgraph::dbms::Database>;

src/dbms/dbms_handler.cpp

@@ -11,29 +11,73 @@
 
 #include "dbms/dbms_handler.hpp"
 
-#include "dbms/coordinator_handlers.hpp"
-#include "flags/replication.hpp"
-
 #include <cstdint>
 #include <filesystem>
 
 #include "dbms/constants.hpp"
 #include "dbms/global.hpp"
-#include "dbms/replication_client.hpp"
 #include "spdlog/spdlog.h"
+#include "system/include/system/system.hpp"
 #include "utils/exceptions.hpp"
 #include "utils/logging.hpp"
 #include "utils/uuid.hpp"
 
 namespace memgraph::dbms {
 
-#ifdef MG_ENTERPRISE
-
 namespace {
-constexpr std::string_view kDBPrefix = "database:";                               // Key prefix for database durability
-constexpr std::string_view kLastCommitedSystemTsKey = "last_commited_system_ts";  // Key for timestamp durability
+constexpr std::string_view kDBPrefix = "database:";  // Key prefix for database durability
+
+std::string RegisterReplicaErrorToString(query::RegisterReplicaError error) {
+  switch (error) {
+    using enum query::RegisterReplicaError;
+    case NAME_EXISTS:
+      return "NAME_EXISTS";
+    case ENDPOINT_EXISTS:
+      return "ENDPOINT_EXISTS";
+    case CONNECTION_FAILED:
+      return "CONNECTION_FAILED";
+    case COULD_NOT_BE_PERSISTED:
+      return "COULD_NOT_BE_PERSISTED";
+  }
+}
+
+// Per storage
+// NOTE Storage will connect to all replicas. Future work might change this
+void RestoreReplication(replication::RoleMainData &mainData, DatabaseAccess db_acc) {
+  spdlog::info("Restoring replication role.");
+
+  // Each individual client has already been restored and started. Here we just go through each database and start its
+  // client
+  for (auto &instance_client : mainData.registered_replicas_) {
+    spdlog::info("Replica {} restoration started for {}.", instance_client.name_, db_acc->name());
+    const auto &ret = db_acc->storage()->repl_storage_state_.replication_clients_.WithLock(
+        [&, db_acc](auto &storage_clients) mutable -> utils::BasicResult<query::RegisterReplicaError> {
+          auto client = std::make_unique<storage::ReplicationStorageClient>(instance_client);
+          auto *storage = db_acc->storage();
+          client->Start(storage, std::move(db_acc));
+          // After start the storage <-> replica state should be READY or RECOVERING (if correctly started)
+          // MAYBE_BEHIND isn't a statement of the current state, this is the default value
+          // Failed to start due to branching of MAIN and REPLICA
+          if (client->State() == storage::replication::ReplicaState::MAYBE_BEHIND) {
+            spdlog::warn("Connection failed when registering replica {}. Replica will still be registered.",
+                         instance_client.name_);
+          }
+          storage_clients.push_back(std::move(client));
+          return {};
+        });
+
+    if (ret.HasError()) {
+      MG_ASSERT(query::RegisterReplicaError::CONNECTION_FAILED != ret.GetError());
+      LOG_FATAL("Failure when restoring replica {}: {}.", instance_client.name_,
+                RegisterReplicaErrorToString(ret.GetError()));
+    }
+    spdlog::info("Replica {} restored for {}.", instance_client.name_, db_acc->name());
+  }
+  spdlog::info("Replication role restored to MAIN.");
+}
 }  // namespace
 
+#ifdef MG_ENTERPRISE
 struct Durability {
   enum class DurabilityVersion : uint8_t {
     V0 = 0,
@@ -112,11 +156,9 @@ struct Durability {
   }
 };
 
-DbmsHandler::DbmsHandler(
-    storage::Config config,
-    memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
-    bool recovery_on_startup)
-    : default_config_{std::move(config)}, repl_state_{ReplicationStateRootPath(default_config_)} {
+DbmsHandler::DbmsHandler(storage::Config config, memgraph::system::System &system,
+                         replication::ReplicationState &repl_state, auth::SynchedAuth &auth, bool recovery_on_startup)
+    : default_config_{std::move(config)}, auth_{auth}, repl_state_{repl_state}, system_{&system} {
   // TODO: Decouple storage config from dbms config
   // TODO: Save individual db configs inside the kvstore and restore from there
 
@@ -150,19 +192,13 @@ DbmsHandler::DbmsHandler(
       const auto uuid = json.at("uuid").get<utils::UUID>();
       const auto rel_dir = json.at("rel_dir").get<std::filesystem::path>();
       spdlog::info("Restoring database {} at {}.", name, rel_dir);
-      auto new_db = New_(name, uuid, rel_dir);
+      auto new_db = New_(name, uuid, nullptr, rel_dir);
       MG_ASSERT(!new_db.HasError(), "Failed while creating database {}.", name);
       directories.emplace(rel_dir.filename());
       spdlog::info("Database {} restored.", name);
     }
-    // Read the last timestamp
-    auto lcst = durability_->Get(kLastCommitedSystemTsKey);
-    if (lcst) {
-      last_commited_system_timestamp_ = std::stoul(*lcst);
-      system_timestamp_ = last_commited_system_timestamp_;
-    }
   } else {  // Clear databases from the durability list and auth
-    auto locked_auth = auth->Lock();
+    auto locked_auth = auth_.Lock();
     auto it = durability_->begin(std::string{kDBPrefix});
     auto end = durability_->end(std::string{kDBPrefix});
     for (; it != end; ++it) {
@@ -172,8 +208,6 @@ DbmsHandler::DbmsHandler(
       locked_auth->DeleteDatabase(name);
       durability_->Delete(key);
     }
-    // Delete the last timestamp
-    durability_->Delete(kLastCommitedSystemTsKey);
   }
 
   /*
@@ -198,45 +232,29 @@ DbmsHandler::DbmsHandler(
    */
   // Setup the default DB
   SetupDefault_();
-
-  /*
-   * REPLICATION RECOVERY AND STARTUP
-   */
-  // Startup replication state (if recovered at startup)
-  auto replica = [this](replication::RoleReplicaData const &data) { return StartRpcServer(*this, data); };
-  // Replication recovery and frequent check start
-  auto main = [this](replication::RoleMainData &data) {
-    for (auto &client : data.registered_replicas_) {
-      SystemRestore(client);
-    }
-    ForEach([this](DatabaseAccess db) { RecoverReplication(db); });
-    for (auto &client : data.registered_replicas_) {
-      StartReplicaClient(*this, client);
-    }
-    return true;
-  };
-  // Startup proccess for main/replica
-  MG_ASSERT(std::visit(memgraph::utils::Overloaded{replica, main}, repl_state_.ReplicationData()),
-            "Replica recovery failure!");
-
-  // Warning
-  if (default_config_.durability.snapshot_wal_mode == storage::Config::Durability::SnapshotWalMode::DISABLED &&
-      repl_state_.IsMain()) {
-    spdlog::warn(
-        "The instance has the MAIN replication role, but durability logs and snapshots are disabled. Please "
-        "consider "
-        "enabling durability by using --storage-snapshot-interval-sec and --storage-wal-enabled flags because "
-        "without write-ahead logs this instance is not replicating any data.");
-  }
-
-  // MAIN or REPLICA instance
-  if (FLAGS_coordinator_server_port) {
-    CoordinatorHandlers::Register(*this);
-    MG_ASSERT(coordinator_state_.GetCoordinatorServer().Start(), "Failed to start coordinator server!");
-  }
 }
 
-DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name) {
+struct DropDatabase : memgraph::system::ISystemAction {
+  explicit DropDatabase(utils::UUID uuid) : uuid_{uuid} {}
+  void DoDurability() override { /* Done during DBMS execution */
+  }
+
+  bool DoReplication(replication::ReplicationClient &client, replication::ReplicationEpoch const &epoch,
+                     memgraph::system::Transaction const &txn) const override {
+    auto check_response = [](const storage::replication::DropDatabaseRes &response) {
+      return response.result != storage::replication::DropDatabaseRes::Result::FAILURE;
+    };
+
+    return client.SteamAndFinalizeDelta<storage::replication::DropDatabaseRpc>(
+        check_response, epoch.id(), txn.last_committed_system_timestamp(), txn.timestamp(), uuid_);
+  }
+  void PostReplication(replication::RoleMainData &mainData) const override {}
+
+ private:
+  utils::UUID uuid_;
+};
+
+DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name, system::Transaction *transaction) {
   std::lock_guard<LockT> wr(lock_);
   if (db_name == kDefaultDB) {
     // MSG cannot delete the default db
@@ -273,9 +291,10 @@ DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name) {
 
   // Success
   // Save delta
-  if (system_transaction_) {
-    system_transaction_->delta.emplace(SystemTransaction::Delta::drop_database, uuid);
+  if (transaction) {
+    transaction->AddAction<DropDatabase>(uuid);
   }
+
   return {};
 }
 
@@ -296,18 +315,48 @@ DbmsHandler::DeleteResult DbmsHandler::Delete(utils::UUID uuid) {
   return Delete_(db_name);
 }
 
-DbmsHandler::NewResultT DbmsHandler::New_(storage::Config storage_config) {
+struct CreateDatabase : memgraph::system::ISystemAction {
+  explicit CreateDatabase(storage::SalientConfig config, DatabaseAccess db_acc)
+      : config_{std::move(config)}, db_acc(db_acc) {}
+
+  void DoDurability() override {
+    // Done during dbms execution
+  }
+
+  bool DoReplication(replication::ReplicationClient &client, replication::ReplicationEpoch const &epoch,
+                     memgraph::system::Transaction const &txn) const override {
+    auto check_response = [](const storage::replication::CreateDatabaseRes &response) {
+      return response.result != storage::replication::CreateDatabaseRes::Result::FAILURE;
+    };
+
+    return client.SteamAndFinalizeDelta<storage::replication::CreateDatabaseRpc>(
+        check_response, epoch.id(), txn.last_committed_system_timestamp(), txn.timestamp(), config_);
+  }
+
+  void PostReplication(replication::RoleMainData &mainData) const override {
+    // Sync database with REPLICAs
+    // NOTE: The function bellow is used to create ReplicationStorageClient, so it must be called on a new storage
+    // We don't need to have it here, since the function won't fail even if the replication client fails to
+    // connect We will just have everything ready, for recovery at some point.
+    dbms::DbmsHandler::RecoverStorageReplication(db_acc, mainData);
+  }
+
+ private:
+  storage::SalientConfig config_;
+  DatabaseAccess db_acc;
+};
+
+DbmsHandler::NewResultT DbmsHandler::New_(storage::Config storage_config, system::Transaction *txn) {
   auto new_db = db_handler_.New(storage_config, repl_state_);
 
   if (new_db.HasValue()) {  // Success
                             // Save delta
-    if (system_transaction_) {
-      system_transaction_->delta.emplace(SystemTransaction::Delta::create_database, storage_config.salient);
-    }
     UpdateDurability(storage_config);
-    return new_db.GetValue();
+    if (txn) {
+      txn->AddAction<CreateDatabase>(storage_config.salient, new_db.GetValue());
+    }
   }
-  return new_db.GetError();
+  return new_db;
 }
 
 DbmsHandler::DeleteResult DbmsHandler::Delete_(std::string_view db_name) {
@@ -361,89 +410,16 @@ void DbmsHandler::UpdateDurability(const storage::Config &config, std::optional<
   durability_->Put(key, val);
 }
 
-AllSyncReplicaStatus DbmsHandler::Commit() {
-  if (system_transaction_ == std::nullopt || system_transaction_->delta == std::nullopt)
-    return AllSyncReplicaStatus::AllCommitsConfirmed;  // Nothing to commit
-  const auto &delta = *system_transaction_->delta;
-
-  auto sync_status = AllSyncReplicaStatus::AllCommitsConfirmed;
-  // TODO Create a system client that can handle all of this automatically
-  switch (delta.action) {
-    using enum SystemTransaction::Delta::Action;
-    case CREATE_DATABASE: {
-      // Replication
-      auto main_handler = [&](memgraph::replication::RoleMainData &main_data) {
-        // TODO: data race issue? registered_replicas_ access not protected
-        // This is sync in any case, as this is the startup
-        for (auto &client : main_data.registered_replicas_) {
-          bool completed = SteamAndFinalizeDelta<storage::replication::CreateDatabaseRpc>(
-              client,
-              [](const storage::replication::CreateDatabaseRes &response) {
-                return response.result != storage::replication::CreateDatabaseRes::Result::FAILURE;
-              },
-              std::string(main_data.epoch_.id()), last_commited_system_timestamp_,
-              system_transaction_->system_timestamp, delta.config);
-          // TODO: reduce duplicate code
-          if (!completed && client.mode_ == replication_coordination_glue::ReplicationMode::SYNC) {
-            sync_status = AllSyncReplicaStatus::SomeCommitsUnconfirmed;
-          }
-        }
-        // Sync database with REPLICAs
-        RecoverReplication(Get_(delta.config.name));
-      };
-      auto replica_handler = [](memgraph::replication::RoleReplicaData &) { /* Nothing to do */ };
-      std::visit(utils::Overloaded{main_handler, replica_handler}, repl_state_.ReplicationData());
-    } break;
-    case DROP_DATABASE: {
-      // Replication
-      auto main_handler = [&](memgraph::replication::RoleMainData &main_data) {
-        // TODO: data race issue? registered_replicas_ access not protected
-        // This is sync in any case, as this is the startup
-        for (auto &client : main_data.registered_replicas_) {
-          bool completed = SteamAndFinalizeDelta<storage::replication::DropDatabaseRpc>(
-              client,
-              [](const storage::replication::DropDatabaseRes &response) {
-                return response.result != storage::replication::DropDatabaseRes::Result::FAILURE;
-              },
-              std::string(main_data.epoch_.id()), last_commited_system_timestamp_,
-              system_transaction_->system_timestamp, delta.uuid);
-          // TODO: reduce duplicate code
-          if (!completed && client.mode_ == replication_coordination_glue::ReplicationMode::SYNC) {
-            sync_status = AllSyncReplicaStatus::SomeCommitsUnconfirmed;
-          }
-        }
-      };
-      auto replica_handler = [](memgraph::replication::RoleReplicaData &) { /* Nothing to do */ };
-      std::visit(utils::Overloaded{main_handler, replica_handler}, repl_state_.ReplicationData());
-    } break;
-  }
-
-  durability_->Put(kLastCommitedSystemTsKey, std::to_string(system_transaction_->system_timestamp));
-  last_commited_system_timestamp_ = system_transaction_->system_timestamp;
-  ResetSystemTransaction();
-  return sync_status;
-}
-
-#else  // not MG_ENTERPRISE
-
-AllSyncReplicaStatus DbmsHandler::Commit() {
-  if (system_transaction_ == std::nullopt || system_transaction_->delta == std::nullopt) {
-    return AllSyncReplicaStatus::AllCommitsConfirmed;  // Nothing to commit
-  }
-  const auto &delta = *system_transaction_->delta;
-
-  switch (delta.action) {
-    using enum SystemTransaction::Delta::Action;
-    case CREATE_DATABASE:
-    case DROP_DATABASE:
-      /* Community edition doesn't support multi-tenant replication */
-      break;
-  }
-
-  last_commited_system_timestamp_ = system_transaction_->system_timestamp;
-  ResetSystemTransaction();
-  return AllSyncReplicaStatus::AllCommitsConfirmed;
-}
-
 #endif
+
+void DbmsHandler::RecoverStorageReplication(DatabaseAccess db_acc, replication::RoleMainData &role_main_data) {
+  if (allow_mt_repl || db_acc->name() == dbms::kDefaultDB) {
+    // Handle global replication state
+    spdlog::info("Replication configuration will be stored and will be automatically restored in case of a crash.");
+    // RECOVER REPLICA CONNECTIONS
+    memgraph::dbms::RestoreReplication(role_main_data, db_acc);
+  } else if (!role_main_data.registered_replicas_.empty()) {
+    spdlog::warn("Multi-tenant replication is currently not supported!");
+  }
+}
 }  // namespace memgraph::dbms

src/dbms/dbms_handler.hpp

@@ -25,24 +25,24 @@
 #include "constants.hpp"
 #include "dbms/database.hpp"
 #include "dbms/inmemory/replication_handlers.hpp"
-#include "dbms/replication_handler.hpp"
+#include "dbms/rpc.hpp"
 #include "kvstore/kvstore.hpp"
+#include "license/license.hpp"
 #include "replication/replication_client.hpp"
 #include "storage/v2/config.hpp"
-#include "storage/v2/replication/enums.hpp"
-#include "storage/v2/replication/rpc.hpp"
 #include "storage/v2/transaction.hpp"
+#include "system/system.hpp"
 #include "utils/thread_pool.hpp"
 #ifdef MG_ENTERPRISE
 #include "coordination/coordinator_state.hpp"
 #include "dbms/database_handler.hpp"
 #endif
-#include "dbms/transaction.hpp"
 #include "global.hpp"
 #include "query/config.hpp"
 #include "query/interpreter_context.hpp"
 #include "spdlog/spdlog.h"
 #include "storage/v2/isolation_level.hpp"
+#include "system/system.hpp"
 #include "utils/logging.hpp"
 #include "utils/result.hpp"
 #include "utils/rw_lock.hpp"
@@ -51,11 +51,6 @@
 
 namespace memgraph::dbms {
 
-enum class AllSyncReplicaStatus {
-  AllCommitsConfirmed,
-  SomeCommitsUnconfirmed,
-};
-
 struct Statistics {
   uint64_t num_vertex;           //!< Sum of vertexes in every database
   uint64_t num_edges;            //!< Sum of edges in every database
@@ -111,8 +106,8 @@ class DbmsHandler {
    * @param auth pointer to the global authenticator
    * @param recovery_on_startup restore databases (and its content) and authentication data
    */
-  DbmsHandler(storage::Config config,
-              memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
+  DbmsHandler(storage::Config config, memgraph::system::System &system, replication::ReplicationState &repl_state,
+              auth::SynchedAuth &auth,
               bool recovery_on_startup);  // TODO If more arguments are added use a config struct
 #else
   /**
@@ -120,15 +115,14 @@ class DbmsHandler {
    *
    * @param configs storage configuration
    */
-  DbmsHandler(storage::Config config)
-      : repl_state_{ReplicationStateRootPath(config)},
+  DbmsHandler(storage::Config config, memgraph::system::System &system, replication::ReplicationState &repl_state)
+      : repl_state_{repl_state},
+        system_{&system},
         db_gatekeeper_{[&] {
                          config.salient.name = kDefaultDB;
                          return std::move(config);
                        }(),
-                       repl_state_} {
-    RecoverReplication(Get());
-  }
+                       repl_state_} {}
 #endif
 
 #ifdef MG_ENTERPRISE
@@ -138,10 +132,10 @@ class DbmsHandler {
    * @param name name of the database
    * @return NewResultT context on success, error on failure
    */
-  NewResultT New(const std::string &name) {
+  NewResultT New(const std::string &name, system::Transaction *txn = nullptr) {
     std::lock_guard<LockT> wr(lock_);
     const auto uuid = utils::UUID{};
-    return New_(name, uuid);
+    return New_(name, uuid, txn);
   }
 
   /**
@@ -234,7 +228,7 @@ class DbmsHandler {
    * @param db_name database name
    * @return DeleteResult error on failure
    */
-  DeleteResult TryDelete(std::string_view db_name);
+  DeleteResult TryDelete(std::string_view db_name, system::Transaction *transaction = nullptr);
 
   /**
    * @brief Delete or defer deletion of database.
@@ -267,23 +261,12 @@ class DbmsHandler {
 #endif
   }
 
-  replication::ReplicationState &ReplicationState() { return repl_state_; }
-  replication::ReplicationState const &ReplicationState() const { return repl_state_; }
-
-  bool IsMain() const { return repl_state_.IsMain(); }
-  bool IsReplica() const { return repl_state_.IsReplica(); }
-
-#ifdef MG_ENTERPRISE
-  coordination::CoordinatorState &CoordinatorState() { return coordinator_state_; }
-#endif
-
   /**
    * @brief Return the statistics all databases.
    *
    * @return Statistics
    */
-  Statistics Stats() {
-    auto const replication_role = repl_state_.GetRole();
+  Statistics Stats(memgraph::replication_coordination_glue::ReplicationRole replication_role) {
     Statistics stats{};
     // TODO: Handle overflow?
 #ifdef MG_ENTERPRISE
@@ -319,8 +302,7 @@ class DbmsHandler {
    *
    * @return std::vector<DatabaseInfo>
    */
-  std::vector<DatabaseInfo> Info() {
-    auto const replication_role = repl_state_.GetRole();
+  std::vector<DatabaseInfo> Info(memgraph::replication_coordination_glue::ReplicationRole replication_role) {
     std::vector<DatabaseInfo> res;
 #ifdef MG_ENTERPRISE
     std::shared_lock<LockT> rd(lock_);
@@ -407,98 +389,17 @@ class DbmsHandler {
     }
   }
 
-  void NewSystemTransaction() {
-    DMG_ASSERT(!system_transaction_, "Already running a system transaction");
-    system_transaction_.emplace(++system_timestamp_);
-  }
-
-  void ResetSystemTransaction() { system_transaction_.reset(); }
-
-  //! \tparam RPC An rpc::RequestResponse
-  //! \tparam Args the args type
-  //! \param client the client to use for rpc communication
-  //! \param check predicate to check response is ok
-  //! \param args arguments to forward to the rpc request
-  //! \return If replica stream is completed or enqueued
-  template <typename RPC, typename... Args>
-  bool SteamAndFinalizeDelta(auto &client, auto &&check, Args &&...args) {
-    try {
-      auto stream = client.rpc_client_.template Stream<RPC>(std::forward<Args>(args)...);
-      auto task = [&client, check = std::forward<decltype(check)>(check), stream = std::move(stream)]() mutable {
-        if (stream.IsDefunct()) {
-          client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
-          return false;
-        }
-        try {
-          if (check(stream.AwaitResponse())) {
-            return true;
-          }
-        } catch (memgraph::rpc::GenericRpcFailedException const &e) {
-          // swallow error, fallthrough to error handling
-        }
-        // This replica needs SYSTEM recovery
-        client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
-        return false;
-      };
-
-      if (client.mode_ == memgraph::replication_coordination_glue::ReplicationMode::ASYNC) {
-        client.thread_pool_.AddTask([task = utils::CopyMovableFunctionWrapper{std::move(task)}]() mutable { task(); });
-        return true;
-      }
-
-      return task();
-    } catch (memgraph::rpc::GenericRpcFailedException const &e) {
-      // This replica needs SYSTEM recovery
-      client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
-      return false;
-    }
-  };
-
-  AllSyncReplicaStatus Commit();
-
-  auto LastCommitedTS() const -> uint64_t { return last_commited_system_timestamp_; }
-  void SetLastCommitedTS(uint64_t new_ts) { last_commited_system_timestamp_.store(new_ts); }
+  static void RecoverStorageReplication(DatabaseAccess db_acc, replication::RoleMainData &role_main_data);
 
+  auto default_config() const -> storage::Config const & {
 #ifdef MG_ENTERPRISE
-  // When being called by intepreter no need to gain lock, it should already be under a system transaction
-  // But concurrently the FrequentCheck is running and will need to lock before reading last_commited_system_timestamp_
-  template <bool REQUIRE_LOCK = false>
-  void SystemRestore(replication::ReplicationClient &client) {
-    // Check if system is up to date
-    if (client.state_.WithLock(
-            [](auto &state) { return state == memgraph::replication::ReplicationClient::State::READY; }))
-      return;
-
-    // Try to recover...
-    {
-      auto [database_configs, last_commited_system_timestamp] = std::invoke([&] {
-        auto sys_guard =
-            std::unique_lock{system_lock_, std::defer_lock};  // ensure no other system transaction in progress
-        if constexpr (REQUIRE_LOCK) {
-          sys_guard.lock();
-        }
-        auto configs = std::vector<storage::SalientConfig>{};
-        ForEach([&configs](DatabaseAccess acc) { configs.emplace_back(acc->config().salient); });
-        return std::pair{configs, last_commited_system_timestamp_.load()};
-      });
-      try {
-        auto stream = client.rpc_client_.Stream<storage::replication::SystemRecoveryRpc>(last_commited_system_timestamp,
-                                                                                         std::move(database_configs));
-        const auto response = stream.AwaitResponse();
-        if (response.result == storage::replication::SystemRecoveryRes::Result::FAILURE) {
-          client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
-          return;
-        }
-      } catch (memgraph::rpc::GenericRpcFailedException const &e) {
-        client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
-        return;
-      }
-    }
-
-    // Successfully recovered
-    client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::READY; });
-  }
+    return default_config_;
+#else
+    const auto acc = db_gatekeeper_.access();
+    MG_ASSERT(acc, "Failed to get default database!");
+    return acc->get()->config();
 #endif
+  }
 
  private:
 #ifdef MG_ENTERPRISE
@@ -524,7 +425,8 @@ class DbmsHandler {
    * @param uuid undelying RocksDB directory
    * @return NewResultT context on success, error on failure
    */
-  NewResultT New_(std::string_view name, utils::UUID uuid, std::optional<std::filesystem::path> rel_dir = {}) {
+  NewResultT New_(std::string_view name, utils::UUID uuid, system::Transaction *txn = nullptr,
+                  std::optional<std::filesystem::path> rel_dir = {}) {
     auto config_copy = default_config_;
     config_copy.salient.name = name;
     config_copy.salient.uuid = uuid;
@@ -535,7 +437,7 @@ class DbmsHandler {
       storage::UpdatePaths(config_copy,
                            default_config_.durability.storage_directory / kMultiTenantDir / std::string{uuid});
     }
-    return New_(std::move(config_copy));
+    return New_(std::move(config_copy), txn);
   }
 
   /**
@@ -544,11 +446,11 @@ class DbmsHandler {
    * @param config configuration to be used
    * @return NewResultT context on success, error on failure
    */
-  NewResultT New_(const storage::SalientConfig &config) {
+  NewResultT New_(const storage::SalientConfig &config, system::Transaction *txn = nullptr) {
     auto config_copy = default_config_;
     config_copy.salient = config;  // name, uuid, mode, etc
     UpdatePaths(config_copy, config_copy.durability.storage_directory / kMultiTenantDir / std::string{config.uuid});
-    return New_(std::move(config_copy));
+    return New_(std::move(config_copy), txn);
   }
 
   /**
@@ -557,7 +459,7 @@ class DbmsHandler {
    * @param storage_config storage configuration
    * @return NewResultT context on success, error on failure
    */
-  NewResultT New_(storage::Config storage_config);
+  DbmsHandler::NewResultT New_(storage::Config storage_config, system::Transaction *txn = nullptr);
 
   // TODO: new overload of Delete_ with DatabaseAccess
   DeleteResult Delete_(std::string_view db_name);
@@ -572,7 +474,8 @@ class DbmsHandler {
       Get(kDefaultDB);
     } catch (const UnknownDatabaseException &) {
       // No default DB restored, create it
-      MG_ASSERT(New_(kDefaultDB, {/* random UUID */}, ".").HasValue(), "Failed while creating the default database");
+      MG_ASSERT(New_(kDefaultDB, {/* random UUID */}, nullptr, ".").HasValue(),
+                "Failed while creating the default database");
     }
 
     // For back-compatibility...
@@ -659,35 +562,24 @@ class DbmsHandler {
   }
 #endif
 
-  void RecoverReplication(DatabaseAccess db_acc) {
-    if (allow_mt_repl || db_acc->name() == dbms::kDefaultDB) {
-      // Handle global replication state
-      spdlog::info("Replication configuration will be stored and will be automatically restored in case of a crash.");
-      // RECOVER REPLICA CONNECTIONS
-      memgraph::dbms::RestoreReplication(repl_state_, std::move(db_acc));
-    } else if (const ::memgraph::replication::RoleMainData *data =
-                   std::get_if<::memgraph::replication::RoleMainData>(&repl_state_.ReplicationData());
-               data && !data->registered_replicas_.empty()) {
-      spdlog::warn("Multi-tenant replication is currently not supported!");
-    }
-  }
-
 #ifdef MG_ENTERPRISE
   mutable LockT lock_{utils::RWLock::Priority::READ};  //!< protective lock
   storage::Config default_config_;                     //!< Storage configuration used when creating new databases
   DatabaseHandler db_handler_;                         //!< multi-tenancy storage handler
-  std::unique_ptr<kvstore::KVStore> durability_;       //!< list of active dbs (pointer so we can postpone its creation)
-  coordination::CoordinatorState coordinator_state_;  //!< Replication coordinator
+  // TODO: move to be common
+  std::unique_ptr<kvstore::KVStore> durability_;  //!< list of active dbs (pointer so we can postpone its creation)
+  auth::SynchedAuth &auth_;                       //!< Synchronized auth::Auth
 #endif
-  // TODO: Make an api
- public:
-  utils::ResourceLock system_lock_{};  //!> Ensure exclusive access for system queries
  private:
-  std::optional<SystemTransaction> system_transaction_;      //!< Current system transaction (only one at a time)
-  uint64_t system_timestamp_{storage::kTimestampInitialId};  //!< System timestamp
-  std::atomic_uint64_t last_commited_system_timestamp_{
-      storage::kTimestampInitialId};          //!< Last commited system timestamp
-  replication::ReplicationState repl_state_;  //!< Global replication state
+  // NOTE: atm the only reason this exists here, is because we pass it into the construction of New Database's
+  //       Database only uses it as a convience to make the correct Access without out needing to be told the
+  //       current replication role. TODO: make Database Access explicit about the role and remove this from
+  //       dbms stuff
+  replication::ReplicationState &repl_state_;  //!< Ref to global replication state
+ public:
+  // TODO fix to be non public/remove from dbms....maybe
+  system::System *system_;
+
 #ifndef MG_ENTERPRISE
   mutable utils::Gatekeeper<Database> db_gatekeeper_;  //!< Single databases gatekeeper
 #endif

src/dbms/inmemory/storage_helper.hpp

@@ -11,10 +11,6 @@
 
 #pragma once
 
-#include <variant>
-
-#include "dbms/constants.hpp"
-#include "dbms/replication_handler.hpp"
 #include "replication/state.hpp"
 #include "storage/v2/config.hpp"
 #include "storage/v2/inmemory/storage.hpp"

src/dbms/replication_client.cpp

@@ -1,43 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-#include "dbms/replication_client.hpp"
-#include "replication/replication_client.hpp"
-
-namespace memgraph::dbms {
-
-void StartReplicaClient(DbmsHandler &dbms_handler, replication::ReplicationClient &client) {
-  // No client error, start instance level client
-  auto const &endpoint = client.rpc_client_.Endpoint();
-  spdlog::trace("Replication client started at: {}:{}", endpoint.address, endpoint.port);
-  client.StartFrequentCheck([&dbms_handler](bool reconnect, replication::ReplicationClient &client) {
-    // Working connection
-    // Check if system needs restoration
-    if (reconnect) {
-      client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
-    }
-#ifdef MG_ENTERPRISE
-    dbms_handler.SystemRestore<true>(client);
-#endif
-    // Check if any database has been left behind
-    dbms_handler.ForEach([&name = client.name_, reconnect](dbms::DatabaseAccess db_acc) {
-      // Specific database <-> replica client
-      db_acc->storage()->repl_storage_state_.WithClient(name, [&](storage::ReplicationStorageClient *client) {
-        if (reconnect || client->State() == storage::replication::ReplicaState::MAYBE_BEHIND) {
-          // Database <-> replica might be behind, check and recover
-          client->TryCheckReplicaStateAsync(db_acc->storage(), db_acc);
-        }
-      });
-    });
-  });
-}  // namespace memgraph::dbms
-
-}  // namespace memgraph::dbms

src/dbms/replication_handler.cpp

@@ -1,400 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-#include "dbms/replication_handler.hpp"
-
-#include <algorithm>
-
-#include "dbms/constants.hpp"
-#include "dbms/dbms_handler.hpp"
-#include "dbms/global.hpp"
-#include "dbms/inmemory/replication_handlers.hpp"
-#include "dbms/replication_client.hpp"
-#include "dbms/utils.hpp"
-#include "replication/messages.hpp"
-#include "replication/state.hpp"
-#include "spdlog/spdlog.h"
-#include "storage/v2/config.hpp"
-#include "storage/v2/replication/rpc.hpp"
-#include "utils/on_scope_exit.hpp"
-
-using memgraph::replication::RoleMainData;
-using memgraph::replication::RoleReplicaData;
-
-namespace memgraph::dbms {
-
-namespace {
-
-std::string RegisterReplicaErrorToString(RegisterReplicaError error) {
-  switch (error) {
-    using enum RegisterReplicaError;
-    case NAME_EXISTS:
-      return "NAME_EXISTS";
-    case ENDPOINT_EXISTS:
-      return "ENDPOINT_EXISTS";
-    case CONNECTION_FAILED:
-      return "CONNECTION_FAILED";
-    case COULD_NOT_BE_PERSISTED:
-      return "COULD_NOT_BE_PERSISTED";
-  }
-}
-}  // namespace
-
-ReplicationHandler::ReplicationHandler(DbmsHandler &dbms_handler) : dbms_handler_(dbms_handler) {}
-
-bool ReplicationHandler::SetReplicationRoleMain() {
-  auto const main_handler = [](RoleMainData &) {
-    // If we are already MAIN, we don't want to change anything
-    return false;
-  };
-
-  auto const replica_handler = [this](RoleReplicaData const &) {
-    return memgraph::dbms::DoReplicaToMainPromotion(dbms_handler_);
-  };
-
-  // TODO: under lock
-  return std::visit(utils::Overloaded{main_handler, replica_handler},
-                    dbms_handler_.ReplicationState().ReplicationData());
-}
-
-bool ReplicationHandler::SetReplicationRoleReplica(const memgraph::replication::ReplicationServerConfig &config) {
-  // We don't want to restart the server if we're already a REPLICA
-  if (dbms_handler_.ReplicationState().IsReplica()) {
-    return false;
-  }
-
-  // TODO StorageState needs to be synched. Could have a dangling reference if someone adds a database as we are
-  //      deleting the replica.
-  // Remove database specific clients
-  dbms_handler_.ForEach([&](DatabaseAccess db_acc) {
-    auto *storage = db_acc->storage();
-    storage->repl_storage_state_.replication_clients_.WithLock([](auto &clients) { clients.clear(); });
-  });
-  // Remove instance level clients
-  std::get<RoleMainData>(dbms_handler_.ReplicationState().ReplicationData()).registered_replicas_.clear();
-
-  // Creates the server
-  dbms_handler_.ReplicationState().SetReplicationRoleReplica(config);
-
-  // Start
-  const auto success =
-      std::visit(utils::Overloaded{[](RoleMainData const &) {
-                                     // ASSERT
-                                     return false;
-                                   },
-                                   [this](RoleReplicaData const &data) { return StartRpcServer(dbms_handler_, data); }},
-                 dbms_handler_.ReplicationState().ReplicationData());
-  // TODO Handle error (restore to main?)
-  return success;
-}
-
-auto ReplicationHandler::RegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
-    -> memgraph::utils::BasicResult<RegisterReplicaError> {
-  MG_ASSERT(dbms_handler_.ReplicationState().IsMain(), "Only main instance can register a replica!");
-
-  auto maybe_client = dbms_handler_.ReplicationState().RegisterReplica(config);
-  if (maybe_client.HasError()) {
-    switch (maybe_client.GetError()) {
-      case memgraph::replication::RegisterReplicaError::NOT_MAIN:
-        MG_ASSERT(false, "Only main instance can register a replica!");
-        return {};
-      case memgraph::replication::RegisterReplicaError::NAME_EXISTS:
-        return memgraph::dbms::RegisterReplicaError::NAME_EXISTS;
-      case memgraph::replication::RegisterReplicaError::ENDPOINT_EXISTS:
-        return memgraph::dbms::RegisterReplicaError::ENDPOINT_EXISTS;
-      case memgraph::replication::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
-        return memgraph::dbms::RegisterReplicaError::COULD_NOT_BE_PERSISTED;
-      case memgraph::replication::RegisterReplicaError::SUCCESS:
-        break;
-    }
-  }
-
-  if (!allow_mt_repl && dbms_handler_.All().size() > 1) {
-    spdlog::warn("Multi-tenant replication is currently not supported!");
-  }
-
-#ifdef MG_ENTERPRISE
-  // Update system before enabling individual storage <-> replica clients
-  dbms_handler_.SystemRestore(*maybe_client.GetValue());
-#endif
-
-  const auto dbms_error = memgraph::dbms::HandleRegisterReplicaStatus(maybe_client);
-  if (dbms_error.has_value()) {
-    return *dbms_error;
-  }
-  auto &instance_client_ptr = maybe_client.GetValue();
-  const bool all_clients_good = memgraph::dbms::RegisterAllDatabasesClients(dbms_handler_, *instance_client_ptr);
-
-  // NOTE Currently if any databases fails, we revert back
-  if (!all_clients_good) {
-    spdlog::error("Failed to register all databases on the REPLICA \"{}\"", config.name);
-    UnregisterReplica(config.name);
-    return RegisterReplicaError::CONNECTION_FAILED;
-  }
-
-  // No client error, start instance level client
-  StartReplicaClient(dbms_handler_, *instance_client_ptr);
-  return {};
-}
-
-auto ReplicationHandler::UnregisterReplica(std::string_view name) -> UnregisterReplicaResult {
-  auto const replica_handler = [](RoleReplicaData const &) -> UnregisterReplicaResult {
-    return UnregisterReplicaResult::NOT_MAIN;
-  };
-  auto const main_handler = [this, name](RoleMainData &mainData) -> UnregisterReplicaResult {
-    if (!dbms_handler_.ReplicationState().TryPersistUnregisterReplica(name)) {
-      return UnregisterReplicaResult::COULD_NOT_BE_PERSISTED;
-    }
-    // Remove database specific clients
-    dbms_handler_.ForEach([name](DatabaseAccess db_acc) {
-      db_acc->storage()->repl_storage_state_.replication_clients_.WithLock([&name](auto &clients) {
-        std::erase_if(clients, [name](const auto &client) { return client->Name() == name; });
-      });
-    });
-    // Remove instance level clients
-    auto const n_unregistered =
-        std::erase_if(mainData.registered_replicas_, [name](auto const &client) { return client.name_ == name; });
-    return n_unregistered != 0 ? UnregisterReplicaResult::SUCCESS : UnregisterReplicaResult::CAN_NOT_UNREGISTER;
-  };
-
-  return std::visit(utils::Overloaded{main_handler, replica_handler},
-                    dbms_handler_.ReplicationState().ReplicationData());
-}
-
-auto ReplicationHandler::GetRole() const -> memgraph::replication_coordination_glue::ReplicationRole {
-  return dbms_handler_.ReplicationState().GetRole();
-}
-
-bool ReplicationHandler::IsMain() const { return dbms_handler_.ReplicationState().IsMain(); }
-
-bool ReplicationHandler::IsReplica() const { return dbms_handler_.ReplicationState().IsReplica(); }
-
-// Per storage
-// NOTE Storage will connect to all replicas. Future work might change this
-void RestoreReplication(replication::ReplicationState &repl_state, DatabaseAccess db_acc) {
-  spdlog::info("Restoring replication role.");
-
-  /// MAIN
-  auto const recover_main = [db_acc = std::move(db_acc)](RoleMainData &mainData) mutable {  // NOLINT
-    // Each individual client has already been restored and started. Here we just go through each database and start its
-    // client
-    for (auto &instance_client : mainData.registered_replicas_) {
-      spdlog::info("Replica {} restoration started for {}.", instance_client.name_, db_acc->name());
-      const auto &ret = db_acc->storage()->repl_storage_state_.replication_clients_.WithLock(
-          [&, db_acc](auto &storage_clients) mutable -> utils::BasicResult<RegisterReplicaError> {
-            auto client = std::make_unique<storage::ReplicationStorageClient>(instance_client);
-            auto *storage = db_acc->storage();
-            client->Start(storage, std::move(db_acc));
-            // After start the storage <-> replica state should be READY or RECOVERING (if correctly started)
-            // MAYBE_BEHIND isn't a statement of the current state, this is the default value
-            // Failed to start due to branching of MAIN and REPLICA
-            if (client->State() == storage::replication::ReplicaState::MAYBE_BEHIND) {
-              spdlog::warn("Connection failed when registering replica {}. Replica will still be registered.",
-                           instance_client.name_);
-            }
-            storage_clients.push_back(std::move(client));
-            return {};
-          });
-
-      if (ret.HasError()) {
-        MG_ASSERT(RegisterReplicaError::CONNECTION_FAILED != ret.GetError());
-        LOG_FATAL("Failure when restoring replica {}: {}.", instance_client.name_,
-                  RegisterReplicaErrorToString(ret.GetError()));
-      }
-      spdlog::info("Replica {} restored for {}.", instance_client.name_, db_acc->name());
-    }
-    spdlog::info("Replication role restored to MAIN.");
-  };
-
-  /// REPLICA
-  auto const recover_replica = [](RoleReplicaData const &data) { /*nothing to do*/ };
-
-  std::visit(
-      utils::Overloaded{
-          recover_main,
-          recover_replica,
-      },
-      repl_state.ReplicationData());
-}
-
-namespace system_replication {
-#ifdef MG_ENTERPRISE
-void SystemHeartbeatHandler(const uint64_t ts, slk::Reader *req_reader, slk::Builder *res_builder) {
-  replication::SystemHeartbeatReq req;
-  replication::SystemHeartbeatReq::Load(&req, req_reader);
-
-  replication::SystemHeartbeatRes res(ts);
-  memgraph::slk::Save(res, res_builder);
-}
-
-void CreateDatabaseHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder) {
-  memgraph::storage::replication::CreateDatabaseReq req;
-  memgraph::slk::Load(&req, req_reader);
-
-  using memgraph::storage::replication::CreateDatabaseRes;
-  CreateDatabaseRes res(CreateDatabaseRes::Result::FAILURE);
-
-  // Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
-  //       of the set of databases. Hence no history exists to maintain regarding epoch change.
-  //       If MAIN has changed we need to check this new group_timestamp is consistent with
-  //       what we have so far.
-
-  if (req.expected_group_timestamp != dbms_handler.LastCommitedTS()) {
-    spdlog::debug("CreateDatabaseHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
-                  dbms_handler.LastCommitedTS());
-    memgraph::slk::Save(res, res_builder);
-    return;
-  }
-
-  try {
-    // Create new
-    auto new_db = dbms_handler.Update(req.config);
-    if (new_db.HasValue()) {
-      // Successfully create db
-      dbms_handler.SetLastCommitedTS(req.new_group_timestamp);
-      res = CreateDatabaseRes(CreateDatabaseRes::Result::SUCCESS);
-      spdlog::debug("CreateDatabaseHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
-    }
-  } catch (...) {
-    // Failure
-  }
-
-  memgraph::slk::Save(res, res_builder);
-}
-
-void DropDatabaseHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder) {
-  memgraph::storage::replication::DropDatabaseReq req;
-  memgraph::slk::Load(&req, req_reader);
-
-  using memgraph::storage::replication::DropDatabaseRes;
-  DropDatabaseRes res(DropDatabaseRes::Result::FAILURE);
-
-  // Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
-  //       of the set of databases. Hence no history exists to maintain regarding epoch change.
-  //       If MAIN has changed we need to check this new group_timestamp is consistent with
-  //       what we have so far.
-
-  if (req.expected_group_timestamp != dbms_handler.LastCommitedTS()) {
-    spdlog::debug("DropDatabaseHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
-                  dbms_handler.LastCommitedTS());
-    memgraph::slk::Save(res, res_builder);
-    return;
-  }
-
-  try {
-    // NOTE: Single communication channel can exist at a time, no other database can be deleted/created at the moment.
-    auto new_db = dbms_handler.Delete(req.uuid);
-    if (new_db.HasError()) {
-      if (new_db.GetError() == DeleteError::NON_EXISTENT) {
-        // Nothing to drop
-        dbms_handler.SetLastCommitedTS(req.new_group_timestamp);
-        res = DropDatabaseRes(DropDatabaseRes::Result::NO_NEED);
-      }
-    } else {
-      // Successfully drop db
-      dbms_handler.SetLastCommitedTS(req.new_group_timestamp);
-      res = DropDatabaseRes(DropDatabaseRes::Result::SUCCESS);
-      spdlog::debug("DropDatabaseHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
-    }
-  } catch (...) {
-    // Failure
-  }
-
-  memgraph::slk::Save(res, res_builder);
-}
-
-void SystemRecoveryHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder) {
-  // TODO Speed up
-  memgraph::storage::replication::SystemRecoveryReq req;
-  memgraph::slk::Load(&req, req_reader);
-
-  using memgraph::storage::replication::SystemRecoveryRes;
-  SystemRecoveryRes res(SystemRecoveryRes::Result::FAILURE);
-
-  utils::OnScopeExit send_on_exit([&]() { memgraph::slk::Save(res, res_builder); });
-
-  // Get all current dbs
-  auto old = dbms_handler.All();
-
-  // Check/create the incoming dbs
-  for (const auto &config : req.database_configs) {
-    // Missing db
-    try {
-      if (dbms_handler.Update(config).HasError()) {
-        spdlog::debug("SystemRecoveryHandler: Failed to update database \"{}\".", config.name);
-        return;  // Send failure on exit
-      }
-    } catch (const UnknownDatabaseException &) {
-      spdlog::debug("SystemRecoveryHandler: UnknownDatabaseException");
-      return;  // Send failure on exit
-    }
-    const auto it = std::find(old.begin(), old.end(), config.name);
-    if (it != old.end()) old.erase(it);
-  }
-
-  // Delete all the leftover old dbs
-  for (const auto &remove_db : old) {
-    const auto del = dbms_handler.Delete(remove_db);
-    if (del.HasError()) {
-      // Some errors are not terminal
-      if (del.GetError() == DeleteError::DEFAULT_DB || del.GetError() == DeleteError::NON_EXISTENT) {
-        spdlog::debug("SystemRecoveryHandler: Dropped database \"{}\".", remove_db);
-        continue;
-      }
-      spdlog::debug("SystemRecoveryHandler: Failed to drop database \"{}\".", remove_db);
-      return;  // Send failure on exit
-    }
-  }
-  // Successfully recovered
-  dbms_handler.SetLastCommitedTS(req.forced_group_timestamp);
-  spdlog::debug("SystemRecoveryHandler: SUCCESS updated LCTS to {}", req.forced_group_timestamp);
-  res = SystemRecoveryRes(SystemRecoveryRes::Result::SUCCESS);
-}
-#endif
-
-void Register(replication::RoleReplicaData const &data, dbms::DbmsHandler &dbms_handler) {
-#ifdef MG_ENTERPRISE
-  data.server->rpc_server_.Register<replication::SystemHeartbeatRpc>(
-      [&dbms_handler](auto *req_reader, auto *res_builder) {
-        spdlog::debug("Received SystemHeartbeatRpc");
-        SystemHeartbeatHandler(dbms_handler.LastCommitedTS(), req_reader, res_builder);
-      });
-  data.server->rpc_server_.Register<storage::replication::CreateDatabaseRpc>(
-      [&dbms_handler](auto *req_reader, auto *res_builder) {
-        spdlog::debug("Received CreateDatabaseRpc");
-        CreateDatabaseHandler(dbms_handler, req_reader, res_builder);
-      });
-  data.server->rpc_server_.Register<storage::replication::DropDatabaseRpc>(
-      [&dbms_handler](auto *req_reader, auto *res_builder) {
-        spdlog::debug("Received DropDatabaseRpc");
-        DropDatabaseHandler(dbms_handler, req_reader, res_builder);
-      });
-  data.server->rpc_server_.Register<storage::replication::SystemRecoveryRpc>(
-      [&dbms_handler](auto *req_reader, auto *res_builder) {
-        spdlog::debug("Received SystemRecoveryRpc");
-        SystemRecoveryHandler(dbms_handler, req_reader, res_builder);
-      });
-#endif
-}
-}  // namespace system_replication
-
-bool StartRpcServer(DbmsHandler &dbms_handler, const replication::RoleReplicaData &data) {
-  // Register handlers
-  InMemoryReplicationHandlers::Register(&dbms_handler, *data.server);
-  system_replication::Register(data, dbms_handler);
-  // Start server
-  if (!data.server->Start()) {
-    spdlog::error("Unable to start the replication server.");
-    return false;
-  }
-  return true;
-}
-}  // namespace memgraph::dbms

src/dbms/replication_handler.hpp

@@ -1,82 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-#pragma once
-
-#include "replication_coordination_glue/role.hpp"
-#include "dbms/database.hpp"
-#include "utils/result.hpp"
-
-namespace memgraph::replication {
-struct ReplicationState;
-struct ReplicationServerConfig;
-struct ReplicationClientConfig;
-}  // namespace memgraph::replication
-
-namespace memgraph::dbms {
-
-class DbmsHandler;
-
-enum class RegisterReplicaError : uint8_t { NAME_EXISTS, ENDPOINT_EXISTS, CONNECTION_FAILED, COULD_NOT_BE_PERSISTED };
-
-enum class UnregisterReplicaResult : uint8_t {
-  NOT_MAIN,
-  COULD_NOT_BE_PERSISTED,
-  CAN_NOT_UNREGISTER,
-  SUCCESS,
-};
-
-/// A handler type that keep in sync current ReplicationState and the MAIN/REPLICA-ness of Storage
-/// TODO: extend to do multiple storages
-struct ReplicationHandler {
-  explicit ReplicationHandler(DbmsHandler &dbms_handler);
-
-  // as REPLICA, become MAIN
-  bool SetReplicationRoleMain();
-
-  // as MAIN, become REPLICA
-  bool SetReplicationRoleReplica(const memgraph::replication::ReplicationServerConfig &config);
-
-  // as MAIN, define and connect to REPLICAs
-  auto RegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
-      -> utils::BasicResult<RegisterReplicaError>;
-
-  // as MAIN, remove a REPLICA connection
-  auto UnregisterReplica(std::string_view name) -> UnregisterReplicaResult;
-
-  // Helper pass-through (TODO: remove)
-  auto GetRole() const -> memgraph::replication_coordination_glue::ReplicationRole;
-  bool IsMain() const;
-  bool IsReplica() const;
-
- private:
-  DbmsHandler &dbms_handler_;
-};
-
-/// A handler type that keep in sync current ReplicationState and the MAIN/REPLICA-ness of Storage
-/// TODO: extend to do multiple storages
-void RestoreReplication(replication::ReplicationState &repl_state, DatabaseAccess db_acc);
-
-namespace system_replication {
-// System handlers
-#ifdef MG_ENTERPRISE
-void CreateDatabaseHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
-void SystemHeartbeatHandler(uint64_t ts, slk::Reader *req_reader, slk::Builder *res_builder);
-void SystemRecoveryHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
-#endif
-
-/// Register all DBMS level RPC handlers
-void Register(replication::RoleReplicaData const &data, DbmsHandler &dbms_handler);
-}  // namespace system_replication
-
-bool StartRpcServer(DbmsHandler &dbms_handler, const replication::RoleReplicaData &data);
-
-}  // namespace memgraph::dbms

src/dbms/replication_handlers.cpp

@@ -0,0 +1,191 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "dbms/replication_handlers.hpp"
+
+#include "dbms/database.hpp"
+#include "dbms/dbms_handler.hpp"
+#include "storage/v2/storage.hpp"
+#include "system/state.hpp"
+
+namespace memgraph::dbms {
+
+#ifdef MG_ENTERPRISE
+
+void CreateDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                           DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder) {
+  using memgraph::storage::replication::CreateDatabaseRes;
+  CreateDatabaseRes res(CreateDatabaseRes::Result::FAILURE);
+
+  // Ignore if no license
+  if (!license::global_license_checker.IsEnterpriseValidFast()) {
+    spdlog::error("Handling CreateDatabase, an enterprise RPC message, without license.");
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  memgraph::storage::replication::CreateDatabaseReq req;
+  memgraph::slk::Load(&req, req_reader);
+
+  // Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
+  //       of the set of databases. Hence no history exists to maintain regarding epoch change.
+  //       If MAIN has changed we need to check this new group_timestamp is consistent with
+  //       what we have so far.
+
+  if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
+    spdlog::debug("CreateDatabaseHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
+                  system_state_access.LastCommitedTS());
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  try {
+    // Create new
+    auto new_db = dbms_handler.Update(req.config);
+    if (new_db.HasValue()) {
+      // Successfully create db
+      system_state_access.SetLastCommitedTS(req.new_group_timestamp);
+      res = CreateDatabaseRes(CreateDatabaseRes::Result::SUCCESS);
+      spdlog::debug("CreateDatabaseHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
+    }
+  } catch (...) {
+    // Failure
+  }
+
+  memgraph::slk::Save(res, res_builder);
+}
+
+void DropDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access, DbmsHandler &dbms_handler,
+                         slk::Reader *req_reader, slk::Builder *res_builder) {
+  using memgraph::storage::replication::DropDatabaseRes;
+  DropDatabaseRes res(DropDatabaseRes::Result::FAILURE);
+
+  // Ignore if no license
+  if (!license::global_license_checker.IsEnterpriseValidFast()) {
+    spdlog::error("Handling DropDatabase, an enterprise RPC message, without license.");
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  memgraph::storage::replication::DropDatabaseReq req;
+  memgraph::slk::Load(&req, req_reader);
+
+  // Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
+  //       of the set of databases. Hence no history exists to maintain regarding epoch change.
+  //       If MAIN has changed we need to check this new group_timestamp is consistent with
+  //       what we have so far.
+
+  if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
+    spdlog::debug("DropDatabaseHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
+                  system_state_access.LastCommitedTS());
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  try {
+    // NOTE: Single communication channel can exist at a time, no other database can be deleted/created at the moment.
+    auto new_db = dbms_handler.Delete(req.uuid);
+    if (new_db.HasError()) {
+      if (new_db.GetError() == DeleteError::NON_EXISTENT) {
+        // Nothing to drop
+        system_state_access.SetLastCommitedTS(req.new_group_timestamp);
+        res = DropDatabaseRes(DropDatabaseRes::Result::NO_NEED);
+      }
+    } else {
+      // Successfully drop db
+      system_state_access.SetLastCommitedTS(req.new_group_timestamp);
+      res = DropDatabaseRes(DropDatabaseRes::Result::SUCCESS);
+      spdlog::debug("DropDatabaseHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
+    }
+  } catch (...) {
+    // Failure
+  }
+
+  memgraph::slk::Save(res, res_builder);
+}
+
+bool SystemRecoveryHandler(DbmsHandler &dbms_handler, const std::vector<storage::SalientConfig> &database_configs) {
+  /*
+   * NO LICENSE
+   */
+  if (!license::global_license_checker.IsEnterpriseValidFast()) {
+    spdlog::error("Handling SystemRecovery, an enterprise RPC message, without license.");
+    for (const auto &config : database_configs) {
+      // Only handle default DB
+      if (config.name != kDefaultDB) continue;
+      try {
+        if (dbms_handler.Update(config).HasError()) {
+          return false;
+        }
+      } catch (const UnknownDatabaseException &) {
+        return false;
+      }
+    }
+    return true;
+  }
+
+  /*
+   * MULTI-TENANCY
+   */
+  // Get all current dbs
+  auto old = dbms_handler.All();
+  // Check/create the incoming dbs
+  for (const auto &config : database_configs) {
+    // Missing db
+    try {
+      if (dbms_handler.Update(config).HasError()) {
+        spdlog::debug("SystemRecoveryHandler: Failed to update database \"{}\".", config.name);
+        return false;
+      }
+    } catch (const UnknownDatabaseException &) {
+      spdlog::debug("SystemRecoveryHandler: UnknownDatabaseException");
+      return false;
+    }
+    const auto it = std::find(old.begin(), old.end(), config.name);
+    if (it != old.end()) old.erase(it);
+  }
+
+  // Delete all the leftover old dbs
+  for (const auto &remove_db : old) {
+    const auto del = dbms_handler.Delete(remove_db);
+    if (del.HasError()) {
+      // Some errors are not terminal
+      if (del.GetError() == DeleteError::DEFAULT_DB || del.GetError() == DeleteError::NON_EXISTENT) {
+        spdlog::debug("SystemRecoveryHandler: Dropped database \"{}\".", remove_db);
+        continue;
+      }
+      spdlog::debug("SystemRecoveryHandler: Failed to drop database \"{}\".", remove_db);
+      return false;
+    }
+  }
+
+  /*
+   * SUCCESS
+   */
+  return true;
+}
+
+void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
+              dbms::DbmsHandler &dbms_handler) {
+  // NOTE: Register even without license as the user could add a license at run-time
+  data.server->rpc_server_.Register<storage::replication::CreateDatabaseRpc>(
+      [system_state_access, &dbms_handler](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received CreateDatabaseRpc");
+        CreateDatabaseHandler(system_state_access, dbms_handler, req_reader, res_builder);
+      });
+  data.server->rpc_server_.Register<storage::replication::DropDatabaseRpc>(
+      [system_state_access, &dbms_handler](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received DropDatabaseRpc");
+        DropDatabaseHandler(system_state_access, dbms_handler, req_reader, res_builder);
+      });
+}
+#endif
+}  // namespace memgraph::dbms

src/dbms/replication_handlers.hpp

@@ -0,0 +1,32 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include "dbms/dbms_handler.hpp"
+#include "replication/state.hpp"
+#include "slk/streams.hpp"
+#include "system/state.hpp"
+
+namespace memgraph::dbms {
+#ifdef MG_ENTERPRISE
+// RPC handlers
+void CreateDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                           DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+void DropDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access, DbmsHandler &dbms_handler,
+                         slk::Reader *req_reader, slk::Builder *res_builder);
+bool SystemRecoveryHandler(DbmsHandler &dbms_handler, const std::vector<storage::SalientConfig> &database_configs);
+
+// RPC registration
+void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
+              dbms::DbmsHandler &dbms_handler);
+#endif
+}  // namespace memgraph::dbms

src/dbms/rpc.cpp

@@ -0,0 +1,118 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "dbms/rpc.hpp"
+
+#include "slk/streams.hpp"
+#include "storage/v2/replication/rpc.hpp"
+#include "utils/enum.hpp"
+#include "utils/typeinfo.hpp"
+
+namespace memgraph {
+
+namespace storage::replication {
+
+void CreateDatabaseReq::Save(const CreateDatabaseReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void CreateDatabaseReq::Load(CreateDatabaseReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+void CreateDatabaseRes::Save(const CreateDatabaseRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void CreateDatabaseRes::Load(CreateDatabaseRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+void DropDatabaseReq::Save(const DropDatabaseReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void DropDatabaseReq::Load(DropDatabaseReq *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
+void DropDatabaseRes::Save(const DropDatabaseRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void DropDatabaseRes::Load(DropDatabaseRes *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
+
+const utils::TypeInfo CreateDatabaseReq::kType{utils::TypeId::REP_CREATE_DATABASE_REQ, "CreateDatabaseReq", nullptr};
+
+const utils::TypeInfo CreateDatabaseRes::kType{utils::TypeId::REP_CREATE_DATABASE_RES, "CreateDatabaseRes", nullptr};
+
+const utils::TypeInfo DropDatabaseReq::kType{utils::TypeId::REP_DROP_DATABASE_REQ, "DropDatabaseReq", nullptr};
+
+const utils::TypeInfo DropDatabaseRes::kType{utils::TypeId::REP_DROP_DATABASE_RES, "DropDatabaseRes", nullptr};
+
+}  // namespace storage::replication
+
+// Autogenerated SLK serialization code
+namespace slk {
+
+// Serialize code for CreateDatabaseReq
+
+void Save(const memgraph::storage::replication::CreateDatabaseReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.epoch_id, builder);
+  memgraph::slk::Save(self.expected_group_timestamp, builder);
+  memgraph::slk::Save(self.new_group_timestamp, builder);
+  memgraph::slk::Save(self.config, builder);
+}
+
+void Load(memgraph::storage::replication::CreateDatabaseReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->epoch_id, reader);
+  memgraph::slk::Load(&self->expected_group_timestamp, reader);
+  memgraph::slk::Load(&self->new_group_timestamp, reader);
+  memgraph::slk::Load(&self->config, reader);
+}
+
+// Serialize code for CreateDatabaseRes
+
+void Save(const memgraph::storage::replication::CreateDatabaseRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(utils::EnumToNum<uint8_t>(self.result), builder);
+}
+
+void Load(memgraph::storage::replication::CreateDatabaseRes *self, memgraph::slk::Reader *reader) {
+  uint8_t res = 0;
+  memgraph::slk::Load(&res, reader);
+  if (!utils::NumToEnum(res, self->result)) {
+    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
+  }
+}
+
+// Serialize code for DropDatabaseReq
+
+void Save(const memgraph::storage::replication::DropDatabaseReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.epoch_id, builder);
+  memgraph::slk::Save(self.expected_group_timestamp, builder);
+  memgraph::slk::Save(self.new_group_timestamp, builder);
+  memgraph::slk::Save(self.uuid, builder);
+}
+
+void Load(memgraph::storage::replication::DropDatabaseReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->epoch_id, reader);
+  memgraph::slk::Load(&self->expected_group_timestamp, reader);
+  memgraph::slk::Load(&self->new_group_timestamp, reader);
+  memgraph::slk::Load(&self->uuid, reader);
+}
+
+// Serialize code for DropDatabaseRes
+
+void Save(const memgraph::storage::replication::DropDatabaseRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(utils::EnumToNum<uint8_t>(self.result), builder);
+}
+
+void Load(memgraph::storage::replication::DropDatabaseRes *self, memgraph::slk::Reader *reader) {
+  uint8_t res = 0;
+  memgraph::slk::Load(&res, reader);
+  if (!utils::NumToEnum(res, self->result)) {
+    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
+  }
+}
+
+}  // namespace slk
+}  // namespace memgraph

src/dbms/rpc.hpp

@@ -0,0 +1,118 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include <cstdint>
+#include <string>
+#include <utility>
+
+#include "rpc/messages.hpp"
+#include "slk/streams.hpp"
+#include "storage/v2/config.hpp"
+#include "utils/uuid.hpp"
+
+namespace memgraph::storage::replication {
+
+struct CreateDatabaseReq {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(CreateDatabaseReq *self, memgraph::slk::Reader *reader);
+  static void Save(const CreateDatabaseReq &self, memgraph::slk::Builder *builder);
+  CreateDatabaseReq() = default;
+  CreateDatabaseReq(std::string_view epoch_id, uint64_t expected_group_timestamp, uint64_t new_group_timestamp,
+                    storage::SalientConfig config)
+      : epoch_id(std::string(epoch_id)),
+        expected_group_timestamp{expected_group_timestamp},
+        new_group_timestamp(new_group_timestamp),
+        config(std::move(config)) {}
+
+  std::string epoch_id;
+  uint64_t expected_group_timestamp;
+  uint64_t new_group_timestamp;
+  storage::SalientConfig config;
+};
+
+struct CreateDatabaseRes {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
+
+  static void Load(CreateDatabaseRes *self, memgraph::slk::Reader *reader);
+  static void Save(const CreateDatabaseRes &self, memgraph::slk::Builder *builder);
+  CreateDatabaseRes() = default;
+  explicit CreateDatabaseRes(Result res) : result(res) {}
+
+  Result result;
+};
+
+using CreateDatabaseRpc = rpc::RequestResponse<CreateDatabaseReq, CreateDatabaseRes>;
+
+struct DropDatabaseReq {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(DropDatabaseReq *self, memgraph::slk::Reader *reader);
+  static void Save(const DropDatabaseReq &self, memgraph::slk::Builder *builder);
+  DropDatabaseReq() = default;
+  DropDatabaseReq(std::string_view epoch_id, uint64_t expected_group_timestamp, uint64_t new_group_timestamp,
+                  const utils::UUID &uuid)
+      : epoch_id(std::string(epoch_id)),
+        expected_group_timestamp{expected_group_timestamp},
+        new_group_timestamp(new_group_timestamp),
+        uuid(uuid) {}
+
+  std::string epoch_id;
+  uint64_t expected_group_timestamp;
+  uint64_t new_group_timestamp;
+  utils::UUID uuid;
+};
+
+struct DropDatabaseRes {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
+
+  static void Load(DropDatabaseRes *self, memgraph::slk::Reader *reader);
+  static void Save(const DropDatabaseRes &self, memgraph::slk::Builder *builder);
+  DropDatabaseRes() = default;
+  explicit DropDatabaseRes(Result res) : result(res) {}
+
+  Result result;
+};
+
+using DropDatabaseRpc = rpc::RequestResponse<DropDatabaseReq, DropDatabaseRes>;
+
+}  // namespace memgraph::storage::replication
+
+// SLK serialization declarations
+namespace memgraph::slk {
+
+void Save(const memgraph::storage::replication::CreateDatabaseReq &self, memgraph::slk::Builder *builder);
+
+void Load(memgraph::storage::replication::CreateDatabaseReq *self, memgraph::slk::Reader *reader);
+
+void Save(const memgraph::storage::replication::CreateDatabaseRes &self, memgraph::slk::Builder *builder);
+
+void Load(memgraph::storage::replication::CreateDatabaseRes *self, memgraph::slk::Reader *reader);
+
+void Save(const memgraph::storage::replication::DropDatabaseReq &self, memgraph::slk::Builder *builder);
+
+void Load(memgraph::storage::replication::DropDatabaseReq *self, memgraph::slk::Reader *reader);
+
+void Save(const memgraph::storage::replication::DropDatabaseRes &self, memgraph::slk::Builder *builder);
+
+void Load(memgraph::storage::replication::DropDatabaseRes *self, memgraph::slk::Reader *reader);
+
+}  // namespace memgraph::slk

src/dbms/transaction.hpp

@@ -11,7 +11,10 @@
 
 #pragma once
 
+#include <list>
 #include <memory>
+#include <optional>
+#include "auth/models.hpp"
 #include "storage/v2/config.hpp"
 
 namespace memgraph::dbms {
@@ -20,17 +23,70 @@ struct SystemTransaction {
     enum class Action {
       CREATE_DATABASE,
       DROP_DATABASE,
+      UPDATE_AUTH_DATA,
+      DROP_AUTH_DATA,
+      /**
+       *
+       * CREATE USER user_name [IDENTIFIED BY 'password'];
+       * SET PASSWORD FOR user_name TO 'new_password';
+       * ^ SaveUser
+       *
+       * DROP USER user_name;
+       * ^ Directly on KVStore
+       *
+       * CREATE ROLE role_name;
+       * ^ SaveRole
+       *
+       * DROP ROLE
+       * ^ RemoveRole
+       *
+       * SET ROLE FOR user_name TO role_name;
+       * CLEAR ROLE FOR user_name;
+       * ^ Do stuff then do SaveUser
+       *
+       * GRANT privilege_list TO user_or_role;
+       * DENY AUTH, INDEX TO moderator:
+       * REVOKE AUTH, INDEX TO moderator:
+       * GRANT permission_level ON (LABELS | EDGE_TYPES) label_list TO user_or_role;
+       * REVOKE (LABELS | EDGE_TYPES) label_or_edge_type_list FROM user_or_role
+       * DENY (LABELS | EDGE_TYPES) label_or_edge_type_list TO user_or_role
+       * ^ all of these are EditPermissions <-> SaveUser/Role
+       *
+       * Multi-tenant TODO Doc;
+       * ^ Should all call SaveUser
+       *
+       */
     };
 
     static constexpr struct CreateDatabase {
     } create_database;
     static constexpr struct DropDatabase {
     } drop_database;
+    static constexpr struct UpdateAuthData {
+    } update_auth_data;
+    static constexpr struct DropAuthData {
+    } drop_auth_data;
 
+    enum class AuthData { USER, ROLE };
+
+    // Multi-tenancy
     Delta(CreateDatabase /*tag*/, storage::SalientConfig config)
         : action(Action::CREATE_DATABASE), config(std::move(config)) {}
     Delta(DropDatabase /*tag*/, const utils::UUID &uuid) : action(Action::DROP_DATABASE), uuid(uuid) {}
 
+    // Auth
+    Delta(UpdateAuthData /*tag*/, std::optional<auth::User> user)
+        : action(Action::UPDATE_AUTH_DATA), auth_data{std::move(user), std::nullopt} {}
+    Delta(UpdateAuthData /*tag*/, std::optional<auth::Role> role)
+        : action(Action::UPDATE_AUTH_DATA), auth_data{std::nullopt, std::move(role)} {}
+    Delta(DropAuthData /*tag*/, AuthData type, std::string_view name)
+        : action(Action::DROP_AUTH_DATA),
+          auth_data_key{
+              .type = type,
+              .name = std::string{name},
+          } {}
+
+    // Generic
     Delta(const Delta &) = delete;
     Delta(Delta &&) = delete;
     Delta &operator=(const Delta &) = delete;
@@ -42,8 +98,14 @@ struct SystemTransaction {
           std::destroy_at(&config);
           break;
         case Action::DROP_DATABASE:
+          std::destroy_at(&uuid);
+          break;
+        case Action::UPDATE_AUTH_DATA:
+          std::destroy_at(&auth_data);
+          break;
+        case Action::DROP_AUTH_DATA:
+          std::destroy_at(&auth_data_key);
           break;
-          // Some deltas might have special destructor handling
       }
     }
 
@@ -51,13 +113,20 @@ struct SystemTransaction {
     union {
       storage::SalientConfig config;
       utils::UUID uuid;
+      struct {
+        std::optional<auth::User> user;
+        std::optional<auth::Role> role;
+      } auth_data;
+      struct {
+        AuthData type;
+        std::string name;
+      } auth_data_key;
     };
   };
 
   explicit SystemTransaction(uint64_t timestamp) : system_timestamp(timestamp) {}
 
-  // Currently system transitions support a single delta
-  std::optional<Delta> delta{};
+  std::list<Delta> deltas{};
   uint64_t system_timestamp;
 };
 

src/dbms/utils.hpp

@@ -1,133 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-#pragma once
-
-#include "dbms/dbms_handler.hpp"
-#include "dbms/replication_handler.hpp"
-#include "replication/include/replication/state.hpp"
-#include "utils/result.hpp"
-
-namespace memgraph::dbms {
-
-inline bool DoReplicaToMainPromotion(dbms::DbmsHandler &dbms_handler) {
-  auto &repl_state = dbms_handler.ReplicationState();
-  // STEP 1) bring down all REPLICA servers
-  dbms_handler.ForEach([](DatabaseAccess db_acc) {
-    auto *storage = db_acc->storage();
-    // Remember old epoch + storage timestamp association
-    storage->PrepareForNewEpoch();
-  });
-
-  // STEP 2) Change to MAIN
-  // TODO: restore replication servers if false?
-  if (!repl_state.SetReplicationRoleMain()) {
-    // TODO: Handle recovery on failure???
-    return false;
-  }
-
-  // STEP 3) We are now MAIN, update storage local epoch
-  const auto &epoch =
-      std::get<replication::RoleMainData>(std::as_const(dbms_handler.ReplicationState()).ReplicationData()).epoch_;
-  dbms_handler.ForEach([&](DatabaseAccess db_acc) {
-    auto *storage = db_acc->storage();
-    storage->repl_storage_state_.epoch_ = epoch;
-  });
-
-  return true;
-};
-
-inline bool SetReplicationRoleReplica(dbms::DbmsHandler &dbms_handler,
-                                      const memgraph::replication::ReplicationServerConfig &config) {
-  if (dbms_handler.ReplicationState().IsReplica()) {
-    return false;
-  }
-
-  // TODO StorageState needs to be synched. Could have a dangling reference if someone adds a database as we are
-  //      deleting the replica.
-  // Remove database specific clients
-  dbms_handler.ForEach([&](DatabaseAccess db_acc) {
-    auto *storage = db_acc->storage();
-    storage->repl_storage_state_.replication_clients_.WithLock([](auto &clients) { clients.clear(); });
-  });
-  // Remove instance level clients
-  std::get<replication::RoleMainData>(dbms_handler.ReplicationState().ReplicationData()).registered_replicas_.clear();
-
-  // Creates the server
-  dbms_handler.ReplicationState().SetReplicationRoleReplica(config);
-
-  // Start
-  const auto success = std::visit(utils::Overloaded{[](replication::RoleMainData const &) {
-                                                      // ASSERT
-                                                      return false;
-                                                    },
-                                                    [&dbms_handler](replication::RoleReplicaData const &data) {
-                                                      return StartRpcServer(dbms_handler, data);
-                                                    }},
-                                  dbms_handler.ReplicationState().ReplicationData());
-  // TODO Handle error (restore to main?)
-  return success;
-}
-
-inline bool RegisterAllDatabasesClients(dbms::DbmsHandler &dbms_handler,
-                                        replication::ReplicationClient &instance_client) {
-  if (!allow_mt_repl && dbms_handler.All().size() > 1) {
-    spdlog::warn("Multi-tenant replication is currently not supported!");
-  }
-
-  bool all_clients_good = true;
-
-  // Add database specific clients (NOTE Currently all databases are connected to each replica)
-  dbms_handler.ForEach([&](DatabaseAccess db_acc) {
-    auto *storage = db_acc->storage();
-    if (!allow_mt_repl && storage->name() != kDefaultDB) {
-      return;
-    }
-    // TODO: ATM only IN_MEMORY_TRANSACTIONAL, fix other modes
-    if (storage->storage_mode_ != storage::StorageMode::IN_MEMORY_TRANSACTIONAL) return;
-
-    all_clients_good &= storage->repl_storage_state_.replication_clients_.WithLock(
-        [storage, &instance_client, db_acc = std::move(db_acc)](auto &storage_clients) mutable {  // NOLINT
-          auto client = std::make_unique<storage::ReplicationStorageClient>(instance_client);
-          // All good, start replica client
-          client->Start(storage, std::move(db_acc));
-          // After start the storage <-> replica state should be READY or RECOVERING (if correctly started)
-          // MAYBE_BEHIND isn't a statement of the current state, this is the default value
-          // Failed to start due an error like branching of MAIN and REPLICA
-          if (client->State() == storage::replication::ReplicaState::MAYBE_BEHIND) {
-            return false;  // TODO: sometimes we need to still add to storage_clients
-          }
-          storage_clients.push_back(std::move(client));
-          return true;
-        });
-  });
-
-  return all_clients_good;
-}
-
-inline std::optional<RegisterReplicaError> HandleRegisterReplicaStatus(
-    utils::BasicResult<replication::RegisterReplicaError, replication::ReplicationClient *> &instance_client) {
-  if (instance_client.HasError()) switch (instance_client.GetError()) {
-      case replication::RegisterReplicaError::NOT_MAIN:
-        MG_ASSERT(false, "Only main instance can register a replica!");
-        return {};
-      case replication::RegisterReplicaError::NAME_EXISTS:
-        return dbms::RegisterReplicaError::NAME_EXISTS;
-      case replication::RegisterReplicaError::ENDPOINT_EXISTS:
-        return dbms::RegisterReplicaError::ENDPOINT_EXISTS;
-      case replication::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
-        return dbms::RegisterReplicaError::COULD_NOT_BE_PERSISTED;
-      case replication::RegisterReplicaError::SUCCESS:
-        break;
-    }
-  return {};
-}
-
-}  // namespace memgraph::dbms

src/glue/ServerT.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -33,7 +33,7 @@ class WritePrioritizedRWLock;
 
 struct Context {
   memgraph::query::InterpreterContext *ic;
-  memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth;
+  memgraph::auth::SynchedAuth *auth;
 #if MG_ENTERPRISE
   memgraph::audit::Log *audit_log;
 #endif

src/glue/SessionHL.cpp

@@ -319,8 +319,7 @@ void SessionHL::Configure(const std::map<std::string, memgraph::communication::b
 SessionHL::SessionHL(memgraph::query::InterpreterContext *interpreter_context,
                      memgraph::communication::v2::ServerEndpoint endpoint,
                      memgraph::communication::v2::InputStream *input_stream,
-                     memgraph::communication::v2::OutputStream *output_stream,
-                     memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth
+                     memgraph::communication::v2::OutputStream *output_stream, memgraph::auth::SynchedAuth *auth
 #ifdef MG_ENTERPRISE
                      ,
                      memgraph::audit::Log *audit_log

src/glue/SessionHL.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -25,8 +25,7 @@ class SessionHL final : public memgraph::communication::bolt::Session<memgraph::
   SessionHL(memgraph::query::InterpreterContext *interpreter_context,
             memgraph::communication::v2::ServerEndpoint endpoint,
             memgraph::communication::v2::InputStream *input_stream,
-            memgraph::communication::v2::OutputStream *output_stream,
-            memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth
+            memgraph::communication::v2::OutputStream *output_stream, memgraph::auth::SynchedAuth *auth
 #ifdef MG_ENTERPRISE
             ,
             memgraph::audit::Log *audit_log
@@ -88,7 +87,7 @@ class SessionHL final : public memgraph::communication::bolt::Session<memgraph::
   memgraph::audit::Log *audit_log_;
   bool in_explicit_db_{false};  //!< If true, the user has defined the database to use via metadata
 #endif
-  memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth_;
+  memgraph::auth::SynchedAuth *auth_;
   memgraph::communication::v2::ServerEndpoint endpoint_;
   std::optional<std::string> implicit_db_;
 };

src/glue/auth_checker.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -66,9 +66,7 @@ bool IsUserAuthorizedEdgeType(const memgraph::auth::User &user, const memgraph::
 #endif
 namespace memgraph::glue {
 
-AuthChecker::AuthChecker(
-    memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth)
-    : auth_(auth) {}
+AuthChecker::AuthChecker(memgraph::auth::SynchedAuth *auth) : auth_(auth) {}
 
 bool AuthChecker::IsUserAuthorized(const std::optional<std::string> &username,
                                    const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,

src/glue/auth_checker.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -22,8 +22,7 @@ namespace memgraph::glue {
 
 class AuthChecker : public query::AuthChecker {
  public:
-  explicit AuthChecker(
-      memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth);
+  explicit AuthChecker(memgraph::auth::SynchedAuth *auth);
 
   bool IsUserAuthorized(const std::optional<std::string> &username,
                         const std::vector<query::AuthQuery::Privilege> &privileges,
@@ -41,7 +40,7 @@ class AuthChecker : public query::AuthChecker {
                                              const std::string &db_name = "");
 
  private:
-  memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth_;
+  memgraph::auth::SynchedAuth *auth_;
   mutable memgraph::utils::Synchronized<auth::User, memgraph::utils::SpinLock> user_;  // cached user
 };
 #ifdef MG_ENTERPRISE

src/glue/auth_handler.cpp

@@ -210,16 +210,25 @@ std::vector<std::vector<memgraph::query::TypedValue>> ShowFineGrainedUserPrivile
   if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
     return {};
   }
-  const auto &label_permissions = user->GetFineGrainedAccessLabelPermissions();
-  const auto &edge_type_permissions = user->GetFineGrainedAccessEdgeTypePermissions();
 
-  auto all_fine_grained_permissions =
-      GetFineGrainedPermissionForPrivilegeForUserOrRole(label_permissions, "LABEL", "USER");
-  auto edge_type_fine_grained_permissions =
-      GetFineGrainedPermissionForPrivilegeForUserOrRole(edge_type_permissions, "EDGE_TYPE", "USER");
+  auto all_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
+      user->GetUserFineGrainedAccessLabelPermissions(), "LABEL", "USER");
+  auto all_role_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
+      user->GetRoleFineGrainedAccessLabelPermissions(), "LABEL", "ROLE");
+  all_fine_grained_permissions.insert(all_fine_grained_permissions.end(),
+                                      std::make_move_iterator(all_role_fine_grained_permissions.begin()),
+                                      std::make_move_iterator(all_role_fine_grained_permissions.end()));
 
-  all_fine_grained_permissions.insert(all_fine_grained_permissions.end(), edge_type_fine_grained_permissions.begin(),
-                                      edge_type_fine_grained_permissions.end());
+  auto edge_type_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
+      user->GetUserFineGrainedAccessEdgeTypePermissions(), "EDGE_TYPE", "USER");
+  auto role_edge_type_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
+      user->GetRoleFineGrainedAccessEdgeTypePermissions(), "EDGE_TYPE", "ROLE");
+  all_fine_grained_permissions.insert(all_fine_grained_permissions.end(),
+                                      std::make_move_iterator(edge_type_fine_grained_permissions.begin()),
+                                      std::make_move_iterator(edge_type_fine_grained_permissions.end()));
+  all_fine_grained_permissions.insert(all_fine_grained_permissions.end(),
+                                      std::make_move_iterator(role_edge_type_fine_grained_permissions.begin()),
+                                      std::make_move_iterator(role_edge_type_fine_grained_permissions.end()));
 
   return ConstructFineGrainedPrivilegesResult(all_fine_grained_permissions);
 }
@@ -233,9 +242,9 @@ std::vector<std::vector<memgraph::query::TypedValue>> ShowFineGrainedRolePrivile
   const auto &edge_type_permissions = role->GetFineGrainedAccessEdgeTypePermissions();
 
   auto all_fine_grained_permissions =
-      GetFineGrainedPermissionForPrivilegeForUserOrRole(label_permissions, "LABEL", "USER");
+      GetFineGrainedPermissionForPrivilegeForUserOrRole(label_permissions, "LABEL", "ROLE");
   auto edge_type_fine_grained_permissions =
-      GetFineGrainedPermissionForPrivilegeForUserOrRole(edge_type_permissions, "EDGE_TYPE", "USER");
+      GetFineGrainedPermissionForPrivilegeForUserOrRole(edge_type_permissions, "EDGE_TYPE", "ROLE");
 
   all_fine_grained_permissions.insert(all_fine_grained_permissions.end(), edge_type_fine_grained_permissions.begin(),
                                       edge_type_fine_grained_permissions.end());
@@ -248,16 +257,15 @@ std::vector<std::vector<memgraph::query::TypedValue>> ShowFineGrainedRolePrivile
 
 namespace memgraph::glue {
 
-AuthQueryHandler::AuthQueryHandler(
-    memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth)
-    : auth_(auth) {}
+AuthQueryHandler::AuthQueryHandler(memgraph::auth::SynchedAuth *auth) : auth_(auth) {}
 
-bool AuthQueryHandler::CreateUser(const std::string &username, const std::optional<std::string> &password) {
+bool AuthQueryHandler::CreateUser(const std::string &username, const std::optional<std::string> &password,
+                                  system::Transaction *system_tx) {
   try {
     const auto [first_user, user_added] = std::invoke([&, this] {
       auto locked_auth = auth_->Lock();
       const auto first_user = !locked_auth->HasUsers();
-      const auto user_added = locked_auth->AddUser(username, password).has_value();
+      const auto user_added = locked_auth->AddUser(username, password, system_tx).has_value();
       return std::make_pair(first_user, user_added);
     });
 
@@ -276,10 +284,11 @@ bool AuthQueryHandler::CreateUser(const std::string &username, const std::option
             }
           }
 #endif
-      );
+          ,
+          system_tx);
 #ifdef MG_ENTERPRISE
-      GrantDatabaseToUser(auth::kAllDatabases, username);
-      SetMainDatabase(dbms::kDefaultDB, username);
+      GrantDatabaseToUser(auth::kAllDatabases, username, system_tx);
+      SetMainDatabase(dbms::kDefaultDB, username, system_tx);
 #endif
     }
 
@@ -289,18 +298,19 @@ bool AuthQueryHandler::CreateUser(const std::string &username, const std::option
   }
 }
 
-bool AuthQueryHandler::DropUser(const std::string &username) {
+bool AuthQueryHandler::DropUser(const std::string &username, system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
     if (!user) return false;
-    return locked_auth->RemoveUser(username);
+    return locked_auth->RemoveUser(username, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 
-void AuthQueryHandler::SetPassword(const std::string &username, const std::optional<std::string> &password) {
+void AuthQueryHandler::SetPassword(const std::string &username, const std::optional<std::string> &password,
+                                   system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
@@ -308,39 +318,41 @@ void AuthQueryHandler::SetPassword(const std::string &username, const std::optio
       throw memgraph::query::QueryRuntimeException("User '{}' doesn't exist.", username);
     }
     locked_auth->UpdatePassword(*user, password);
-    locked_auth->SaveUser(*user);
+    locked_auth->SaveUser(*user, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 
-bool AuthQueryHandler::CreateRole(const std::string &rolename) {
+bool AuthQueryHandler::CreateRole(const std::string &rolename, system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
-    return locked_auth->AddRole(rolename).has_value();
+    return locked_auth->AddRole(rolename, system_tx).has_value();
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 
 #ifdef MG_ENTERPRISE
-bool AuthQueryHandler::RevokeDatabaseFromUser(const std::string &db, const std::string &username) {
+bool AuthQueryHandler::RevokeDatabaseFromUser(const std::string &db_name, const std::string &username,
+                                              system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
     if (!user) return false;
-    return locked_auth->RevokeDatabaseFromUser(db, username);
+    return locked_auth->RevokeDatabaseFromUser(db_name, username, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 
-bool AuthQueryHandler::GrantDatabaseToUser(const std::string &db, const std::string &username) {
+bool AuthQueryHandler::GrantDatabaseToUser(const std::string &db_name, const std::string &username,
+                                           system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
     if (!user) return false;
-    return locked_auth->GrantDatabaseToUser(db, username);
+    return locked_auth->GrantDatabaseToUser(db_name, username, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
@@ -360,27 +372,28 @@ std::vector<std::vector<memgraph::query::TypedValue>> AuthQueryHandler::GetDatab
   }
 }
 
-bool AuthQueryHandler::SetMainDatabase(std::string_view db, const std::string &username) {
+bool AuthQueryHandler::SetMainDatabase(std::string_view db_name, const std::string &username,
+                                       system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
     if (!user) return false;
-    return locked_auth->SetMainDatabase(db, username);
+    return locked_auth->SetMainDatabase(db_name, username, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 
-void AuthQueryHandler::DeleteDatabase(std::string_view db) {
+void AuthQueryHandler::DeleteDatabase(std::string_view db_name, system::Transaction *system_tx) {
   try {
-    auth_->Lock()->DeleteDatabase(std::string(db));
+    auth_->Lock()->DeleteDatabase(std::string(db_name), system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 #endif
 
-bool AuthQueryHandler::DropRole(const std::string &rolename) {
+bool AuthQueryHandler::DropRole(const std::string &rolename, system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto role = locked_auth->GetRole(rolename);
@@ -389,7 +402,7 @@ bool AuthQueryHandler::DropRole(const std::string &rolename) {
       return false;
     };
 
-    return locked_auth->RemoveRole(rolename);
+    return locked_auth->RemoveRole(rolename, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
@@ -461,7 +474,8 @@ std::vector<memgraph::query::TypedValue> AuthQueryHandler::GetUsernamesForRole(c
   }
 }
 
-void AuthQueryHandler::SetRole(const std::string &username, const std::string &rolename) {
+void AuthQueryHandler::SetRole(const std::string &username, const std::string &rolename,
+                               system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
@@ -477,13 +491,13 @@ void AuthQueryHandler::SetRole(const std::string &username, const std::string &r
                                                    current_role->rolename());
     }
     user->SetRole(*role);
-    locked_auth->SaveUser(*user);
+    locked_auth->SaveUser(*user, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
 }
 
-void AuthQueryHandler::ClearRole(const std::string &username) {
+void AuthQueryHandler::ClearRole(const std::string &username, system::Transaction *system_tx) {
   try {
     auto locked_auth = auth_->Lock();
     auto user = locked_auth->GetUser(username);
@@ -491,7 +505,7 @@ void AuthQueryHandler::ClearRole(const std::string &username) {
       throw memgraph::query::QueryRuntimeException("User '{}' doesn't exist .", username);
     }
     user->ClearRole();
-    locked_auth->SaveUser(*user);
+    locked_auth->SaveUser(*user, system_tx);
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());
   }
@@ -545,7 +559,8 @@ void AuthQueryHandler::GrantPrivilege(
     const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
         &edge_type_privileges
 #endif
-) {
+    ,
+    system::Transaction *system_tx) {
   EditPermissions(
       user_or_role, privileges,
 #ifdef MG_ENTERPRISE
@@ -568,11 +583,13 @@ void AuthQueryHandler::GrantPrivilege(
         }
       }
 #endif
-  );
+      ,
+      system_tx);
 }  // namespace memgraph::glue
 
 void AuthQueryHandler::DenyPrivilege(const std::string &user_or_role,
-                                     const std::vector<memgraph::query::AuthQuery::Privilege> &privileges) {
+                                     const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,
+                                     system::Transaction *system_tx) {
   EditPermissions(
       user_or_role, privileges,
 #ifdef MG_ENTERPRISE
@@ -588,7 +605,8 @@ void AuthQueryHandler::DenyPrivilege(const std::string &user_or_role,
       ,
       [](auto &fine_grained_permissions, const auto &privilege_collection) {}
 #endif
-  );
+      ,
+      system_tx);
 }
 
 void AuthQueryHandler::RevokePrivilege(
@@ -600,7 +618,8 @@ void AuthQueryHandler::RevokePrivilege(
     const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
         &edge_type_privileges
 #endif
-) {
+    ,
+    system::Transaction *system_tx) {
   EditPermissions(
       user_or_role, privileges,
 #ifdef MG_ENTERPRISE
@@ -622,7 +641,8 @@ void AuthQueryHandler::RevokePrivilege(
         }
       }
 #endif
-  );
+      ,
+      system_tx);
 }  // namespace memgraph::glue
 
 template <class TEditPermissionsFun
@@ -646,7 +666,8 @@ void AuthQueryHandler::EditPermissions(
     ,
     const TEditFineGrainedPermissionsFun &edit_fine_grained_permissions_fun
 #endif
-) {
+    ,
+    system::Transaction *system_tx) {
   try {
     std::vector<memgraph::auth::Permission> permissions;
     permissions.reserve(privileges.size());
@@ -675,7 +696,7 @@ void AuthQueryHandler::EditPermissions(
         }
       }
 #endif
-      locked_auth->SaveUser(*user);
+      locked_auth->SaveUser(*user, system_tx);
     } else {
       for (const auto &permission : permissions) {
         edit_permissions_fun(role->permissions(), permission);
@@ -691,7 +712,7 @@ void AuthQueryHandler::EditPermissions(
         }
       }
 #endif
-      locked_auth->SaveRole(*role);
+      locked_auth->SaveRole(*role, system_tx);
     }
   } catch (const memgraph::auth::AuthException &e) {
     throw memgraph::query::QueryRuntimeException(e.what());

src/glue/auth_handler.hpp

@@ -23,32 +23,36 @@
 namespace memgraph::glue {
 
 class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
-  memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth_;
+  memgraph::auth::SynchedAuth *auth_;
 
  public:
-  AuthQueryHandler(memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth);
+  explicit AuthQueryHandler(memgraph::auth::SynchedAuth *auth);
 
-  bool CreateUser(const std::string &username, const std::optional<std::string> &password) override;
+  bool CreateUser(const std::string &username, const std::optional<std::string> &password,
+                  system::Transaction *system_tx) override;
 
-  bool DropUser(const std::string &username) override;
+  bool DropUser(const std::string &username, system::Transaction *system_tx) override;
 
-  void SetPassword(const std::string &username, const std::optional<std::string> &password) override;
+  void SetPassword(const std::string &username, const std::optional<std::string> &password,
+                   system::Transaction *system_tx) override;
 
 #ifdef MG_ENTERPRISE
-  bool RevokeDatabaseFromUser(const std::string &db, const std::string &username) override;
+  bool RevokeDatabaseFromUser(const std::string &db_name, const std::string &username,
+                              system::Transaction *system_tx) override;
 
-  bool GrantDatabaseToUser(const std::string &db, const std::string &username) override;
+  bool GrantDatabaseToUser(const std::string &db_name, const std::string &username,
+                           system::Transaction *system_tx) override;
 
   std::vector<std::vector<memgraph::query::TypedValue>> GetDatabasePrivileges(const std::string &username) override;
 
-  bool SetMainDatabase(std::string_view db, const std::string &username) override;
+  bool SetMainDatabase(std::string_view db_name, const std::string &username, system::Transaction *system_tx) override;
 
-  void DeleteDatabase(std::string_view db) override;
+  void DeleteDatabase(std::string_view db_name, system::Transaction *system_tx) override;
 #endif
 
-  bool CreateRole(const std::string &rolename) override;
+  bool CreateRole(const std::string &rolename, system::Transaction *system_tx) override;
 
-  bool DropRole(const std::string &rolename) override;
+  bool DropRole(const std::string &rolename, system::Transaction *system_tx) override;
 
   std::vector<memgraph::query::TypedValue> GetUsernames() override;
 
@@ -58,9 +62,9 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
 
   std::vector<memgraph::query::TypedValue> GetUsernamesForRole(const std::string &rolename) override;
 
-  void SetRole(const std::string &username, const std::string &rolename) override;
+  void SetRole(const std::string &username, const std::string &rolename, system::Transaction *system_tx) override;
 
-  void ClearRole(const std::string &username) override;
+  void ClearRole(const std::string &username, system::Transaction *system_tx) override;
 
   std::vector<std::vector<memgraph::query::TypedValue>> GetPrivileges(const std::string &user_or_role) override;
 
@@ -74,10 +78,12 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
       const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
           &edge_type_privileges
 #endif
-      ) override;
+      ,
+      system::Transaction *system_tx) override;
 
   void DenyPrivilege(const std::string &user_or_role,
-                     const std::vector<memgraph::query::AuthQuery::Privilege> &privileges) override;
+                     const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,
+                     system::Transaction *system_tx) override;
 
   void RevokePrivilege(
       const std::string &user_or_role, const std::vector<memgraph::query::AuthQuery::Privilege> &privileges
@@ -88,7 +94,8 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
       const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
           &edge_type_privileges
 #endif
-      ) override;
+      ,
+      system::Transaction *system_tx) override;
 
  private:
   template <class TEditPermissionsFun
@@ -112,6 +119,7 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
       ,
       const TEditFineGrainedPermissionsFun &edit_fine_grained_permissions_fun
 #endif
-  );
+      ,
+      system::Transaction *system_tx);
 };
 }  // namespace memgraph::glue

src/kvstore/kvstore.cpp

@@ -37,6 +37,7 @@ KVStore::KVStore(std::filesystem::path storage) : pimpl_(std::make_unique<impl>(
 }
 
 KVStore::~KVStore() {
+  if (pimpl_ == nullptr) return;
   spdlog::debug("Destroying KVStore at {}", pimpl_->storage.string());
   const auto sync = pimpl_->db->SyncWAL();
   if (!sync.ok()) spdlog::error("KVStore sync failed!");

src/memgraph.cpp

@@ -11,9 +11,12 @@
 
 #include <cstdint>
 #include "audit/log.hpp"
+#include "auth/auth.hpp"
 #include "communication/websocket/auth.hpp"
 #include "communication/websocket/server.hpp"
+#include "coordination/coordinator_handlers.hpp"
 #include "dbms/constants.hpp"
+#include "dbms/dbms_handler.hpp"
 #include "dbms/inmemory/replication_handlers.hpp"
 #include "flags/all.hpp"
 #include "glue/MonitoringServerT.hpp"
@@ -24,14 +27,19 @@
 #include "helpers.hpp"
 #include "license/license_sender.hpp"
 #include "memory/global_memory_control.hpp"
+#include "query/auth_query_handler.hpp"
 #include "query/config.hpp"
 #include "query/discard_value_stream.hpp"
 #include "query/interpreter.hpp"
+#include "query/interpreter_context.hpp"
 #include "query/procedure/callable_alias_mapper.hpp"
 #include "query/procedure/module.hpp"
 #include "query/procedure/py_module.hpp"
+#include "replication_handler/replication_handler.hpp"
+#include "replication_handler/system_replication.hpp"
 #include "requests/requests.hpp"
 #include "storage/v2/durability/durability.hpp"
+#include "system/system.hpp"
 #include "telemetry/telemetry.hpp"
 #include "utils/signals.hpp"
 #include "utils/sysinfo/memory.hpp"
@@ -39,10 +47,6 @@
 #include "utils/terminate_handler.hpp"
 #include "version.hpp"
 
-#include "dbms/dbms_handler.hpp"
-#include "query/auth_query_handler.hpp"
-#include "query/interpreter_context.hpp"
-
 namespace {
 constexpr const char *kMgUser = "MEMGRAPH_USER";
 constexpr const char *kMgPassword = "MEMGRAPH_PASSWORD";
@@ -356,44 +360,75 @@ int main(int argc, char **argv) {
       .stream_transaction_conflict_retries = FLAGS_stream_transaction_conflict_retries,
       .stream_transaction_retry_interval = std::chrono::milliseconds(FLAGS_stream_transaction_retry_interval)};
 
-  auto auth_glue =
-      [](memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
-         std::unique_ptr<memgraph::query::AuthQueryHandler> &ah, std::unique_ptr<memgraph::query::AuthChecker> &ac) {
-        // Glue high level auth implementations to the query side
-        ah = std::make_unique<memgraph::glue::AuthQueryHandler>(auth);
-        ac = std::make_unique<memgraph::glue::AuthChecker>(auth);
-        // Handle users passed via arguments
-        auto *maybe_username = std::getenv(kMgUser);
-        auto *maybe_password = std::getenv(kMgPassword);
-        auto *maybe_pass_file = std::getenv(kMgPassfile);
-        if (maybe_username && maybe_password) {
-          ah->CreateUser(maybe_username, maybe_password);
-        } else if (maybe_pass_file) {
-          const auto [username, password] = LoadUsernameAndPassword(maybe_pass_file);
-          if (!username.empty() && !password.empty()) {
-            ah->CreateUser(username, password);
-          }
-        }
-      };
+  auto auth_glue = [](memgraph::auth::SynchedAuth *auth, std::unique_ptr<memgraph::query::AuthQueryHandler> &ah,
+                      std::unique_ptr<memgraph::query::AuthChecker> &ac) {
+    // Glue high level auth implementations to the query side
+    ah = std::make_unique<memgraph::glue::AuthQueryHandler>(auth);
+    ac = std::make_unique<memgraph::glue::AuthChecker>(auth);
+    // Handle users passed via arguments
+    auto *maybe_username = std::getenv(kMgUser);
+    auto *maybe_password = std::getenv(kMgPassword);
+    auto *maybe_pass_file = std::getenv(kMgPassfile);
+    if (maybe_username && maybe_password) {
+      ah->CreateUser(maybe_username, maybe_password, nullptr);
+    } else if (maybe_pass_file) {
+      const auto [username, password] = LoadUsernameAndPassword(maybe_pass_file);
+      if (!username.empty() && !password.empty()) {
+        ah->CreateUser(username, password, nullptr);
+      }
+    }
+  };
 
   memgraph::auth::Auth::Config auth_config{FLAGS_auth_user_or_role_name_regex, FLAGS_auth_password_strength_regex,
                                            FLAGS_auth_password_permit_null};
-  memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> auth_{
-      data_directory / "auth", auth_config};
+  memgraph::auth::SynchedAuth auth_{data_directory / "auth", auth_config};
   std::unique_ptr<memgraph::query::AuthQueryHandler> auth_handler;
   std::unique_ptr<memgraph::query::AuthChecker> auth_checker;
   auth_glue(&auth_, auth_handler, auth_checker);
 
-  memgraph::dbms::DbmsHandler dbms_handler(db_config
+  auto system = memgraph::system::System{db_config.durability.storage_directory, FLAGS_data_recovery_on_startup};
+
+  // singleton replication state
+  memgraph::replication::ReplicationState repl_state{ReplicationStateRootPath(db_config)};
+
+  // singleton coordinator state
+#ifdef MG_ENTERPRISE
+  memgraph::coordination::CoordinatorState coordinator_state;
+#endif
+
+  memgraph::dbms::DbmsHandler dbms_handler(db_config, system, repl_state
 #ifdef MG_ENTERPRISE
                                            ,
-                                           &auth_, FLAGS_data_recovery_on_startup
+                                           auth_, FLAGS_data_recovery_on_startup
 #endif
   );
+
+  // Note: Now that all system's subsystems are initialised (dbms & auth)
+  //       We can now initialise the recovery of replication (which will include those subsystems)
+  //       ReplicationHandler will handle the recovery
+  auto replication_handler = memgraph::replication::ReplicationHandler{repl_state, dbms_handler
+#ifdef MG_ENTERPRISE
+                                                                       ,
+                                                                       &system, auth_
+#endif
+  };
+
+#ifdef MG_ENTERPRISE
+  // MAIN or REPLICA instance
+  if (FLAGS_coordinator_server_port) {
+    memgraph::dbms::CoordinatorHandlers::Register(coordinator_state.GetCoordinatorServer(), replication_handler);
+    MG_ASSERT(coordinator_state.GetCoordinatorServer().Start(), "Failed to start coordinator server!");
+  }
+#endif
+
   auto db_acc = dbms_handler.Get();
 
-  memgraph::query::InterpreterContext interpreter_context_(
-      interp_config, &dbms_handler, &dbms_handler.ReplicationState(), auth_handler.get(), auth_checker.get());
+  memgraph::query::InterpreterContext interpreter_context_(interp_config, &dbms_handler, &repl_state, system,
+#ifdef MG_ENTERPRISE
+                                                           &coordinator_state,
+#endif
+                                                           auth_handler.get(), auth_checker.get(),
+                                                           &replication_handler);
   MG_ASSERT(db_acc, "Failed to access the main database");
 
   memgraph::query::procedure::gModuleRegistry.SetModulesDirectory(memgraph::flags::ParseQueryModulesDirectory(),
@@ -460,9 +495,9 @@ int main(int argc, char **argv) {
   if (FLAGS_telemetry_enabled) {
     telemetry.emplace(telemetry_server, data_directory / "telemetry", memgraph::glue::run_id_, machine_id,
                       service_name == "BoltS", FLAGS_data_directory, std::chrono::minutes(10));
-    telemetry->AddStorageCollector(dbms_handler, auth_);
+    telemetry->AddStorageCollector(dbms_handler, auth_, repl_state);
 #ifdef MG_ENTERPRISE
-    telemetry->AddDatabaseCollector(dbms_handler);
+    telemetry->AddDatabaseCollector(dbms_handler, repl_state);
 #else
     telemetry->AddDatabaseCollector();
 #endif

src/query/CMakeLists.txt

@@ -56,6 +56,7 @@ target_link_libraries(mg-query PUBLIC dl
     mg-kvstore
     mg-memory
     mg::csv
+    mg::system
     mg-flags
     mg-dbms
     mg-events)

src/query/auth_query_handler.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -18,6 +18,7 @@
 
 #include "query/frontend/ast/ast.hpp"  // overkill
 #include "query/typed_value.hpp"
+#include "system/system.hpp"
 
 namespace memgraph::query {
 
@@ -33,23 +34,27 @@ class AuthQueryHandler {
 
   /// Return false if the user already exists.
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool CreateUser(const std::string &username, const std::optional<std::string> &password) = 0;
+  virtual bool CreateUser(const std::string &username, const std::optional<std::string> &password,
+                          system::Transaction *system_tx) = 0;
 
   /// Return false if the user does not exist.
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool DropUser(const std::string &username) = 0;
+  virtual bool DropUser(const std::string &username, system::Transaction *system_tx) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual void SetPassword(const std::string &username, const std::optional<std::string> &password) = 0;
+  virtual void SetPassword(const std::string &username, const std::optional<std::string> &password,
+                           system::Transaction *system_tx) = 0;
 
 #ifdef MG_ENTERPRISE
   /// Return true if access revoked successfully
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool RevokeDatabaseFromUser(const std::string &db, const std::string &username) = 0;
+  virtual bool RevokeDatabaseFromUser(const std::string &db, const std::string &username,
+                                      system::Transaction *system_tx) = 0;
 
   /// Return true if access granted successfully
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool GrantDatabaseToUser(const std::string &db, const std::string &username) = 0;
+  virtual bool GrantDatabaseToUser(const std::string &db, const std::string &username,
+                                   system::Transaction *system_tx) = 0;
 
   /// Returns database access rights for the user
   /// @throw QueryRuntimeException if an error ocurred.
@@ -57,20 +62,20 @@ class AuthQueryHandler {
 
   /// Return true if main database set successfully
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool SetMainDatabase(std::string_view db, const std::string &username) = 0;
+  virtual bool SetMainDatabase(std::string_view db, const std::string &username, system::Transaction *system_tx) = 0;
 
   /// Delete database from all users
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual void DeleteDatabase(std::string_view db) = 0;
+  virtual void DeleteDatabase(std::string_view db, system::Transaction *system_tx) = 0;
 #endif
 
   /// Return false if the role already exists.
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool CreateRole(const std::string &rolename) = 0;
+  virtual bool CreateRole(const std::string &rolename, system::Transaction *system_tx) = 0;
 
   /// Return false if the role does not exist.
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual bool DropRole(const std::string &rolename) = 0;
+  virtual bool DropRole(const std::string &rolename, system::Transaction *system_tx) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
   virtual std::vector<memgraph::query::TypedValue> GetUsernames() = 0;
@@ -85,10 +90,10 @@ class AuthQueryHandler {
   virtual std::vector<memgraph::query::TypedValue> GetUsernamesForRole(const std::string &rolename) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual void SetRole(const std::string &username, const std::string &rolename) = 0;
+  virtual void SetRole(const std::string &username, const std::string &rolename, system::Transaction *system_tx) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual void ClearRole(const std::string &username) = 0;
+  virtual void ClearRole(const std::string &username, system::Transaction *system_tx) = 0;
 
   virtual std::vector<std::vector<memgraph::query::TypedValue>> GetPrivileges(const std::string &user_or_role) = 0;
 
@@ -103,11 +108,13 @@ class AuthQueryHandler {
       const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
           &edge_type_privileges
 #endif
-      ) = 0;
+      ,
+      system::Transaction *system_tx) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
   virtual void DenyPrivilege(const std::string &user_or_role,
-                             const std::vector<memgraph::query::AuthQuery::Privilege> &privileges) = 0;
+                             const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,
+                             system::Transaction *system_tx) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
   virtual void RevokePrivilege(
@@ -120,7 +127,8 @@ class AuthQueryHandler {
       const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
           &edge_type_privileges
 #endif
-      ) = 0;
+      ,
+      system::Transaction *system_tx) = 0;
 };
 
 }  // namespace memgraph::query

src/query/frontend/opencypher/grammar/MemgraphCypher.g4

@@ -386,7 +386,7 @@ replicationSocketAddress : literal ;
 registerReplica : REGISTER REPLICA instanceName ( SYNC | ASYNC )
                 TO socketAddress ;
 
-registerInstanceOnCoordinator : REGISTER INSTANCE  instanceName ON coordinatorSocketAddress ( AS ASYNC ) ?  WITH  replicationSocketAddress ;
+registerInstanceOnCoordinator : REGISTER INSTANCE instanceName ON coordinatorSocketAddress ( AS ASYNC ) ? WITH replicationSocketAddress ;
 
 setInstanceToMain : SET INSTANCE instanceName TO MAIN ;
 

src/query/interpreter.cpp

@@ -34,7 +34,9 @@
 #include "auth/auth.hpp"
 #include "auth/models.hpp"
 #include "csv/parsing.hpp"
+#include "dbms/coordinator_handler.hpp"
 #include "dbms/database.hpp"
+#include "dbms/dbms_handler.hpp"
 #include "dbms/global.hpp"
 #include "dbms/inmemory/storage_helper.hpp"
 #include "flags/replication.hpp"
@@ -43,6 +45,7 @@
 #include "license/license.hpp"
 #include "memory/global_memory_control.hpp"
 #include "memory/query_memory_control.hpp"
+#include "query/auth_query_handler.hpp"
 #include "query/config.hpp"
 #include "query/constants.hpp"
 #include "query/context.hpp"
@@ -58,12 +61,14 @@
 #include "query/frontend/semantic/symbol_generator.hpp"
 #include "query/interpret/eval.hpp"
 #include "query/interpret/frame.hpp"
+#include "query/interpreter_context.hpp"
 #include "query/metadata.hpp"
 #include "query/plan/hint_provider.hpp"
 #include "query/plan/planner.hpp"
 #include "query/plan/profile.hpp"
 #include "query/plan/vertex_count_cache.hpp"
 #include "query/procedure/module.hpp"
+#include "query/replication_query_handler.hpp"
 #include "query/stream.hpp"
 #include "query/stream/common.hpp"
 #include "query/stream/sources.hpp"
@@ -71,6 +76,7 @@
 #include "query/trigger.hpp"
 #include "query/typed_value.hpp"
 #include "replication/config.hpp"
+#include "replication/state.hpp"
 #include "spdlog/spdlog.h"
 #include "storage/v2/disk/storage.hpp"
 #include "storage/v2/edge.hpp"
@@ -101,13 +107,6 @@
 #include "utils/typeinfo.hpp"
 #include "utils/variant_helpers.hpp"
 
-#include "dbms/coordinator_handler.hpp"
-#include "dbms/dbms_handler.hpp"
-#include "dbms/replication_handler.hpp"
-#include "query/auth_query_handler.hpp"
-#include "query/interpreter_context.hpp"
-#include "replication/state.hpp"
-
 #ifdef MG_ENTERPRISE
 #include "coordination/constants.hpp"
 #endif
@@ -306,17 +305,18 @@ class ReplQueryHandler {
     ReplicationQuery::ReplicaState state;
   };
 
-  explicit ReplQueryHandler(dbms::DbmsHandler *dbms_handler) : handler_{*dbms_handler} {}
+  explicit ReplQueryHandler(query::ReplicationQueryHandler &replication_query_handler)
+      : handler_{&replication_query_handler} {}
 
   /// @throw QueryRuntimeException if an error ocurred.
   void SetReplicationRole(ReplicationQuery::ReplicationRole replication_role, std::optional<int64_t> port) {
     auto ValidatePort = [](std::optional<int64_t> port) -> void {
-      if (*port < 0 || *port > std::numeric_limits<uint16_t>::max()) {
+      if (!port || *port < 0 || *port > std::numeric_limits<uint16_t>::max()) {
         throw QueryRuntimeException("Port number invalid!");
       }
     };
     if (replication_role == ReplicationQuery::ReplicationRole::MAIN) {
-      if (!handler_.SetReplicationRoleMain()) {
+      if (!handler_->SetReplicationRoleMain()) {
         throw QueryRuntimeException("Couldn't set replication role to main!");
       }
     } else {
@@ -327,7 +327,7 @@ class ReplQueryHandler {
           .port = static_cast<uint16_t>(*port),
       };
 
-      if (!handler_.SetReplicationRoleReplica(config)) {
+      if (!handler_->SetReplicationRoleReplica(config)) {
         throw QueryRuntimeException("Couldn't set role to replica!");
       }
     }
@@ -335,7 +335,7 @@ class ReplQueryHandler {
 
   /// @throw QueryRuntimeException if an error ocurred.
   ReplicationQuery::ReplicationRole ShowReplicationRole() const {
-    switch (handler_.GetRole()) {
+    switch (handler_->GetRole()) {
       case memgraph::replication_coordination_glue::ReplicationRole::MAIN:
         return ReplicationQuery::ReplicationRole::MAIN;
       case memgraph::replication_coordination_glue::ReplicationRole::REPLICA:
@@ -349,7 +349,7 @@ class ReplQueryHandler {
                        const ReplicationQuery::SyncMode sync_mode, const std::chrono::seconds replica_check_frequency) {
     // Coordinator is main by default so this check is OK although it should actually be nothing (neither main nor
     // replica)
-    if (handler_.IsReplica()) {
+    if (handler_->IsReplica()) {
       // replica can't register another replica
       throw QueryRuntimeException("Replica can't register another replica!");
     }
@@ -368,7 +368,7 @@ class ReplQueryHandler {
                                                .replica_check_frequency = replica_check_frequency,
                                                .ssl = std::nullopt};
 
-      const auto error = handler_.RegisterReplica(replication_config).HasError();
+      const auto error = handler_->TryRegisterReplica(replication_config).HasError();
 
       if (error) {
         throw QueryRuntimeException(fmt::format("Couldn't register replica '{}'!", name));
@@ -381,9 +381,9 @@ class ReplQueryHandler {
 
   /// @throw QueryRuntimeException if an error occurred.
   void DropReplica(std::string_view replica_name) {
-    auto const result = handler_.UnregisterReplica(replica_name);
+    auto const result = handler_->UnregisterReplica(replica_name);
     switch (result) {
-      using enum memgraph::dbms::UnregisterReplicaResult;
+      using enum memgraph::query::UnregisterReplicaResult;
       case NOT_MAIN:
         throw QueryRuntimeException("Replica can't unregister a replica!");
       case COULD_NOT_BE_PERSISTED:
@@ -396,7 +396,7 @@ class ReplQueryHandler {
   }
 
   std::vector<ReplicaInfo> ShowReplicas(const dbms::Database &db) const {
-    if (handler_.IsReplica()) {
+    if (handler_->IsReplica()) {
       // replica can't show registered replicas (it shouldn't have any)
       throw QueryRuntimeException("Replica can't show registered replicas (it shouldn't have any)!");
     }
@@ -447,19 +447,16 @@ class ReplQueryHandler {
   }
 
  private:
-  dbms::ReplicationHandler handler_;
+  query::ReplicationQueryHandler *handler_;
 };
 
+#ifdef MG_ENTERPRISE
 class CoordQueryHandler final : public query::CoordinatorQueryHandler {
  public:
-  explicit CoordQueryHandler(dbms::DbmsHandler *dbms_handler) : handler_ { *dbms_handler }
-#ifdef MG_ENTERPRISE
-  , coordinator_handler_(*dbms_handler)
-#endif
-  {
-  }
+  explicit CoordQueryHandler(coordination::CoordinatorState &coordinator_state)
+
+      : coordinator_handler_(coordinator_state) {}
 
-#ifdef MG_ENTERPRISE
   /// @throw QueryRuntimeException if an error ocurred.
   void RegisterInstance(const std::string &coordinator_socket_address, const std::string &replication_socket_address,
                         const std::chrono::seconds instance_check_frequency, const std::string &instance_name,
@@ -497,17 +494,15 @@ class CoordQueryHandler final : public query::CoordinatorQueryHandler {
       using enum memgraph::coordination::RegisterInstanceCoordinatorStatus;
       case NAME_EXISTS:
         throw QueryRuntimeException("Couldn't register replica instance since instance with such name already exists!");
-      case END_POINT_EXISTS:
+      case ENDPOINT_EXISTS:
         throw QueryRuntimeException(
             "Couldn't register replica instance since instance with such endpoint already exists!");
-      case COULD_NOT_BE_PERSISTED:
-        throw QueryRuntimeException("Couldn't register replica instance since it couldn't be persisted!");
       case NOT_COORDINATOR:
-        throw QueryRuntimeException("Couldn't register replica instance since this instance is not a coordinator!");
+        throw QueryRuntimeException("REGISTER INSTANCE query can only be run on a coordinator!");
       case RPC_FAILED:
         throw QueryRuntimeException(
-            "Couldn't register replica because promotion on replica failed! Check logs on replica to find out more "
-            "info!");
+            "Couldn't register replica instance because setting instance to replica failed! Check logs on replica to "
+            "find out more info!");
       case SUCCESS:
         break;
     }
@@ -520,35 +515,29 @@ class CoordQueryHandler final : public query::CoordinatorQueryHandler {
       case NO_INSTANCE_WITH_NAME:
         throw QueryRuntimeException("No instance with such name!");
       case NOT_COORDINATOR:
-        throw QueryRuntimeException("Couldn't set replica instance to main since this instance is not a coordinator!");
+        throw QueryRuntimeException("SET INSTANCE TO MAIN query can only be run on a coordinator!");
       case COULD_NOT_PROMOTE_TO_MAIN:
         throw QueryRuntimeException(
-            "Couldn't set replica instance to main. Check coordinator and replica for more logs");
+            "Couldn't set replica instance to main!. Check coordinator and replica for more logs");
       case SUCCESS:
         break;
     }
   }
 
-#endif
-
-#ifdef MG_ENTERPRISE
   std::vector<coordination::CoordinatorInstanceStatus> ShowInstances() const override {
     return coordinator_handler_.ShowInstances();
   }
 
-#endif
-
  private:
-  dbms::ReplicationHandler handler_;
-#ifdef MG_ENTERPRISE
   dbms::CoordinatorHandler coordinator_handler_;
-#endif
 };
+#endif
 
 /// returns false if the replication role can't be set
 /// @throw QueryRuntimeException if an error ocurred.
 
-Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_context, const Parameters &parameters) {
+Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_context, const Parameters &parameters,
+                         Interpreter &interpreter) {
   AuthQueryHandler *auth = interpreter_context->auth;
 #ifdef MG_ENTERPRISE
   auto *db_handler = interpreter_context->dbms_handler;
@@ -597,63 +586,111 @@ Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_
         license::LicenseCheckErrorToString(license_check_result.GetError(), "advanced authentication features"));
   }
 
+  const auto forbid_on_replica = [has_license = license_check_result.HasError(),
+                                  is_replica = interpreter_context->repl_state->IsReplica()]() {
+    if (is_replica) {
+#if MG_ENTERPRISE
+      if (has_license) {
+        throw QueryException(
+            "Query forbidden on the replica! Update on MAIN, as it is the only source of truth for authentication "
+            "data. MAIN will then replicate the update to connected REPLICAs");
+      }
+      throw QueryException(
+          "Query forbidden on the replica! Switch role to MAIN and update user data, then switch back to REPLICA.");
+#else
+      throw QueryException(
+          "Query forbidden on the replica! Switch role to MAIN and update user data, then switch back to REPLICA.");
+#endif
+    }
+  };
+
   switch (auth_query->action_) {
     case AuthQuery::Action::CREATE_USER:
-      callback.fn = [auth, username, password, valid_enterprise_license = !license_check_result.HasError()] {
+      forbid_on_replica();
+      callback.fn = [auth, username, password, valid_enterprise_license = !license_check_result.HasError(),
+                     interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         MG_ASSERT(password.IsString() || password.IsNull());
-        if (!auth->CreateUser(username, password.IsString() ? std::make_optional(std::string(password.ValueString()))
-                                                            : std::nullopt)) {
+        if (!auth->CreateUser(
+                username, password.IsString() ? std::make_optional(std::string(password.ValueString())) : std::nullopt,
+                &*interpreter->system_transaction_)) {
           throw UserAlreadyExistsException("User '{}' already exists.", username);
         }
 
         // If the license is not valid we create users with admin access
         if (!valid_enterprise_license) {
           spdlog::warn("Granting all the privileges to {}.", username);
-          auth->GrantPrivilege(username, kPrivilegesAll
+          auth->GrantPrivilege(
+              username, kPrivilegesAll
 #ifdef MG_ENTERPRISE
-                               ,
-                               {{{AuthQuery::FineGrainedPrivilege::CREATE_DELETE, {query::kAsterisk}}}},
-                               {
-                                 {
-                                   {
-                                     AuthQuery::FineGrainedPrivilege::CREATE_DELETE, { query::kAsterisk }
-                                   }
-                                 }
-                               }
+              ,
+              {{{AuthQuery::FineGrainedPrivilege::CREATE_DELETE, {query::kAsterisk}}}},
+              {
+                {
+                  {
+                    AuthQuery::FineGrainedPrivilege::CREATE_DELETE, { query::kAsterisk }
+                  }
+                }
+              }
 #endif
-          );
+              ,
+              &*interpreter->system_transaction_);
         }
 
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::DROP_USER:
-      callback.fn = [auth, username] {
-        if (!auth->DropUser(username)) {
+      forbid_on_replica();
+      callback.fn = [auth, username, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
+        if (!auth->DropUser(username, &*interpreter->system_transaction_)) {
           throw QueryRuntimeException("User '{}' doesn't exist.", username);
         }
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::SET_PASSWORD:
-      callback.fn = [auth, username, password] {
+      forbid_on_replica();
+      callback.fn = [auth, username, password, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         MG_ASSERT(password.IsString() || password.IsNull());
         auth->SetPassword(username,
-                          password.IsString() ? std::make_optional(std::string(password.ValueString())) : std::nullopt);
+                          password.IsString() ? std::make_optional(std::string(password.ValueString())) : std::nullopt,
+                          &*interpreter->system_transaction_);
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::CREATE_ROLE:
-      callback.fn = [auth, rolename] {
-        if (!auth->CreateRole(rolename)) {
+      forbid_on_replica();
+      callback.fn = [auth, rolename, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
+        if (!auth->CreateRole(rolename, &*interpreter->system_transaction_)) {
           throw QueryRuntimeException("Role '{}' already exists.", rolename);
         }
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::DROP_ROLE:
-      callback.fn = [auth, rolename] {
-        if (!auth->DropRole(rolename)) {
+      forbid_on_replica();
+      callback.fn = [auth, rolename, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
+        if (!auth->DropRole(rolename, &*interpreter->system_transaction_)) {
           throw QueryRuntimeException("Role '{}' doesn't exist.", rolename);
         }
         return std::vector<std::vector<TypedValue>>();
@@ -684,52 +721,79 @@ Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_
       };
       return callback;
     case AuthQuery::Action::SET_ROLE:
-      callback.fn = [auth, username, rolename] {
-        auth->SetRole(username, rolename);
+      forbid_on_replica();
+      callback.fn = [auth, username, rolename, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
+        auth->SetRole(username, rolename, &*interpreter->system_transaction_);
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::CLEAR_ROLE:
-      callback.fn = [auth, username] {
-        auth->ClearRole(username);
+      forbid_on_replica();
+      callback.fn = [auth, username, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
+        auth->ClearRole(username, &*interpreter->system_transaction_);
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::GRANT_PRIVILEGE:
-      callback.fn = [auth, user_or_role, privileges
+      forbid_on_replica();
+      callback.fn = [auth, user_or_role, privileges, interpreter = &interpreter
 #ifdef MG_ENTERPRISE
                      ,
                      label_privileges, edge_type_privileges
 #endif
       ] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         auth->GrantPrivilege(user_or_role, privileges
 #ifdef MG_ENTERPRISE
                              ,
                              label_privileges, edge_type_privileges
 #endif
-        );
+                             ,
+                             &*interpreter->system_transaction_);
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::DENY_PRIVILEGE:
-      callback.fn = [auth, user_or_role, privileges] {
-        auth->DenyPrivilege(user_or_role, privileges);
+      forbid_on_replica();
+      callback.fn = [auth, user_or_role, privileges, interpreter = &interpreter] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
+        auth->DenyPrivilege(user_or_role, privileges, &*interpreter->system_transaction_);
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
     case AuthQuery::Action::REVOKE_PRIVILEGE: {
-      callback.fn = [auth, user_or_role, privileges
+      forbid_on_replica();
+      callback.fn = [auth, user_or_role, privileges, interpreter = &interpreter
 #ifdef MG_ENTERPRISE
                      ,
                      label_privileges, edge_type_privileges
 #endif
       ] {
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         auth->RevokePrivilege(user_or_role, privileges
 #ifdef MG_ENTERPRISE
                               ,
                               label_privileges, edge_type_privileges
 #endif
-        );
+                              ,
+                              &*interpreter->system_transaction_);
         return std::vector<std::vector<TypedValue>>();
       };
       return callback;
@@ -759,15 +823,20 @@ Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_
       };
       return callback;
     case AuthQuery::Action::GRANT_DATABASE_TO_USER:
+      forbid_on_replica();
 #ifdef MG_ENTERPRISE
-      callback.fn = [auth, database, username, db_handler] {  // NOLINT
+      callback.fn = [auth, database, username, db_handler, interpreter = &interpreter] {  // NOLINT
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         try {
           std::optional<memgraph::dbms::DatabaseAccess> db =
               std::nullopt;  // Hold pointer to database to protect it until query is done
           if (database != memgraph::auth::kAllDatabases) {
             db = db_handler->Get(database);  // Will throw if databases doesn't exist and protect it during pull
           }
-          if (!auth->GrantDatabaseToUser(database, username)) {
+          if (!auth->GrantDatabaseToUser(database, username, &*interpreter->system_transaction_)) {
             throw QueryRuntimeException("Failed to grant database {} to user {}.", database, username);
           }
         } catch (memgraph::dbms::UnknownDatabaseException &e) {
@@ -780,15 +849,20 @@ Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_
       };
       return callback;
     case AuthQuery::Action::REVOKE_DATABASE_FROM_USER:
+      forbid_on_replica();
 #ifdef MG_ENTERPRISE
-      callback.fn = [auth, database, username, db_handler] {  // NOLINT
+      callback.fn = [auth, database, username, db_handler, interpreter = &interpreter] {  // NOLINT
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         try {
           std::optional<memgraph::dbms::DatabaseAccess> db =
               std::nullopt;  // Hold pointer to database to protect it until query is done
           if (database != memgraph::auth::kAllDatabases) {
             db = db_handler->Get(database);  // Will throw if databases doesn't exist and protect it during pull
           }
-          if (!auth->RevokeDatabaseFromUser(database, username)) {
+          if (!auth->RevokeDatabaseFromUser(database, username, &*interpreter->system_transaction_)) {
             throw QueryRuntimeException("Failed to revoke database {} from user {}.", database, username);
           }
         } catch (memgraph::dbms::UnknownDatabaseException &e) {
@@ -813,12 +887,17 @@ Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_
 #endif
       return callback;
     case AuthQuery::Action::SET_MAIN_DATABASE:
+      forbid_on_replica();
 #ifdef MG_ENTERPRISE
-      callback.fn = [auth, database, username, db_handler] {  // NOLINT
+      callback.fn = [auth, database, username, db_handler, interpreter = &interpreter] {  // NOLINT
+        if (!interpreter->system_transaction_) {
+          throw QueryException("Expected to be in a system transaction");
+        }
+
         try {
           const auto db =
               db_handler->Get(database);  // Will throw if databases doesn't exist and protect it during pull
-          if (!auth->SetMainDatabase(database, username)) {
+          if (!auth->SetMainDatabase(database, username, &*interpreter->system_transaction_)) {
             throw QueryRuntimeException("Failed to set main database {} for user {}.", database, username);
           }
         } catch (memgraph::dbms::UnknownDatabaseException &e) {
@@ -836,7 +915,7 @@ Callback HandleAuthQuery(AuthQuery *auth_query, InterpreterContext *interpreter_
 }  // namespace
 
 Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &parameters,
-                                dbms::DbmsHandler *dbms_handler, CurrentDB &current_db,
+                                ReplicationQueryHandler &replication_query_handler, CurrentDB &current_db,
                                 const query::InterpreterConfig &config, std::vector<Notification> *notifications) {
   // TODO: MemoryResource for EvaluationContext, it should probably be passed as
   // the argument to Callback.
@@ -866,7 +945,8 @@ Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &
         notifications->emplace_back(SeverityLevel::WARNING, NotificationCode::REPLICA_PORT_WARNING,
                                     "Be careful the replication port must be different from the memgraph port!");
       }
-      callback.fn = [handler = ReplQueryHandler{dbms_handler}, role = repl_query->role_, maybe_port]() mutable {
+      callback.fn = [handler = ReplQueryHandler{replication_query_handler}, role = repl_query->role_,
+                     maybe_port]() mutable {
         handler.SetReplicationRole(role, maybe_port);
         return std::vector<std::vector<TypedValue>>();
       };
@@ -884,7 +964,7 @@ Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &
 #endif
 
       callback.header = {"replication role"};
-      callback.fn = [handler = ReplQueryHandler{dbms_handler}] {
+      callback.fn = [handler = ReplQueryHandler{replication_query_handler}] {
         auto mode = handler.ShowReplicationRole();
         switch (mode) {
           case ReplicationQuery::ReplicationRole::MAIN: {
@@ -908,7 +988,7 @@ Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &
       auto socket_address = repl_query->socket_address_->Accept(evaluator);
       const auto replica_check_frequency = config.replication_replica_check_frequency;
 
-      callback.fn = [handler = ReplQueryHandler{dbms_handler}, name, socket_address, sync_mode,
+      callback.fn = [handler = ReplQueryHandler{replication_query_handler}, name, socket_address, sync_mode,
                      replica_check_frequency]() mutable {
         handler.RegisterReplica(name, std::string(socket_address.ValueString()), sync_mode, replica_check_frequency);
         return std::vector<std::vector<TypedValue>>();
@@ -925,7 +1005,7 @@ Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &
       }
 #endif
       const auto &name = repl_query->instance_name_;
-      callback.fn = [handler = ReplQueryHandler{dbms_handler}, name]() mutable {
+      callback.fn = [handler = ReplQueryHandler{replication_query_handler}, name]() mutable {
         handler.DropReplica(name);
         return std::vector<std::vector<TypedValue>>();
       };
@@ -943,7 +1023,7 @@ Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &
       callback.header = {
           "name", "socket_address", "sync_mode", "current_timestamp_of_replica", "number_of_timestamp_behind_master",
           "state"};
-      callback.fn = [handler = ReplQueryHandler{dbms_handler}, replica_nfields = callback.header.size(),
+      callback.fn = [handler = ReplQueryHandler{replication_query_handler}, replica_nfields = callback.header.size(),
                      db_acc = current_db.db_acc_] {
         const auto &replicas = handler.ShowReplicas(*db_acc->get());
         auto typed_replicas = std::vector<std::vector<TypedValue>>{};
@@ -991,16 +1071,17 @@ Callback HandleReplicationQuery(ReplicationQuery *repl_query, const Parameters &
   }
 }
 
+#ifdef MG_ENTERPRISE
 Callback HandleCoordinatorQuery(CoordinatorQuery *coordinator_query, const Parameters &parameters,
-                                dbms::DbmsHandler *dbms_handler, const query::InterpreterConfig &config,
-                                std::vector<Notification> *notifications) {
+                                coordination::CoordinatorState *coordinator_state,
+                                const query::InterpreterConfig &config, std::vector<Notification> *notifications) {
   Callback callback;
   switch (coordinator_query->action_) {
     case CoordinatorQuery::Action::REGISTER_INSTANCE: {
       if (!license::global_license_checker.IsEnterpriseValidFast()) {
         throw QueryException("Trying to use enterprise feature without a valid license.");
       }
-#ifdef MG_ENTERPRISE
+
       if constexpr (!coordination::allow_ha) {
         throw QueryRuntimeException(
             "High availability is experimental feature. Please set MG_EXPERIMENTAL_HIGH_AVAILABILITY compile flag to "
@@ -1016,7 +1097,7 @@ Callback HandleCoordinatorQuery(CoordinatorQuery *coordinator_query, const Param
 
       auto coordinator_socket_address_tv = coordinator_query->coordinator_socket_address_->Accept(evaluator);
       auto replication_socket_address_tv = coordinator_query->replication_socket_address_->Accept(evaluator);
-      callback.fn = [handler = CoordQueryHandler{dbms_handler}, coordinator_socket_address_tv,
+      callback.fn = [handler = CoordQueryHandler{*coordinator_state}, coordinator_socket_address_tv,
                      replication_socket_address_tv, main_check_frequency = config.replication_replica_check_frequency,
                      instance_name = coordinator_query->instance_name_,
                      sync_mode = coordinator_query->sync_mode_]() mutable {
@@ -1031,13 +1112,11 @@ Callback HandleCoordinatorQuery(CoordinatorQuery *coordinator_query, const Param
           fmt::format("Coordinator has registered coordinator server on {} for instance {}.",
                       coordinator_socket_address_tv.ValueString(), coordinator_query->instance_name_));
       return callback;
-#endif
     }
     case CoordinatorQuery::Action::SET_INSTANCE_TO_MAIN: {
       if (!license::global_license_checker.IsEnterpriseValidFast()) {
         throw QueryException("Trying to use enterprise feature without a valid license.");
       }
-#ifdef MG_ENTERPRISE
       if constexpr (!coordination::allow_ha) {
         throw QueryRuntimeException(
             "High availability is experimental feature. Please set MG_EXPERIMENTAL_HIGH_AVAILABILITY compile flag to "
@@ -1051,20 +1130,18 @@ Callback HandleCoordinatorQuery(CoordinatorQuery *coordinator_query, const Param
       EvaluationContext evaluation_context{.timestamp = QueryTimestamp(), .parameters = parameters};
       auto evaluator = PrimitiveLiteralExpressionEvaluator{evaluation_context};
 
-      callback.fn = [handler = CoordQueryHandler{dbms_handler},
+      callback.fn = [handler = CoordQueryHandler{*coordinator_state},
                      instance_name = coordinator_query->instance_name_]() mutable {
         handler.SetInstanceToMain(instance_name);
         return std::vector<std::vector<TypedValue>>();
       };
 
       return callback;
-#endif
     }
     case CoordinatorQuery::Action::SHOW_REPLICATION_CLUSTER: {
       if (!license::global_license_checker.IsEnterpriseValidFast()) {
         throw QueryException("Trying to use enterprise feature without a valid license.");
       }
-#ifdef MG_ENTERPRISE
       if constexpr (!coordination::allow_ha) {
         throw QueryRuntimeException(
             "High availability is experimental feature. Please set MG_EXPERIMENTAL_HIGH_AVAILABILITY compile flag to "
@@ -1075,7 +1152,8 @@ Callback HandleCoordinatorQuery(CoordinatorQuery *coordinator_query, const Param
       }
 
       callback.header = {"name", "socket_address", "alive", "role"};
-      callback.fn = [handler = CoordQueryHandler{dbms_handler}, replica_nfields = callback.header.size()]() mutable {
+      callback.fn = [handler = CoordQueryHandler{*coordinator_state},
+                     replica_nfields = callback.header.size()]() mutable {
         auto const instances = handler.ShowInstances();
         std::vector<std::vector<TypedValue>> result{};
         result.reserve(result.size());
@@ -1089,11 +1167,11 @@ Callback HandleCoordinatorQuery(CoordinatorQuery *coordinator_query, const Param
         return result;
       };
       return callback;
-#endif
     }
       return callback;
   }
 }
+#endif
 
 stream::CommonStreamInfo GetCommonStreamInfo(StreamQuery *stream_query, ExpressionVisitor<TypedValue> &evaluator) {
   return {
@@ -2495,14 +2573,14 @@ PreparedQuery PrepareIndexQuery(ParsedQuery parsed_query, bool in_explicit_trans
 }
 
 PreparedQuery PrepareAuthQuery(ParsedQuery parsed_query, bool in_explicit_transaction,
-                               InterpreterContext *interpreter_context) {
+                               InterpreterContext *interpreter_context, Interpreter &interpreter) {
   if (in_explicit_transaction) {
     throw UserModificationInMulticommandTxException();
   }
 
   auto *auth_query = utils::Downcast<AuthQuery>(parsed_query.query);
 
-  auto callback = HandleAuthQuery(auth_query, interpreter_context, parsed_query.parameters);
+  auto callback = HandleAuthQuery(auth_query, interpreter_context, parsed_query.parameters, interpreter);
 
   return PreparedQuery{std::move(callback.header), std::move(parsed_query.required_privileges),
                        [handler = std::move(callback.fn), pull_plan = std::shared_ptr<PullPlanVector>(nullptr),
@@ -2527,15 +2605,16 @@ PreparedQuery PrepareAuthQuery(ParsedQuery parsed_query, bool in_explicit_transa
 }
 
 PreparedQuery PrepareReplicationQuery(ParsedQuery parsed_query, bool in_explicit_transaction,
-                                      std::vector<Notification> *notifications, dbms::DbmsHandler &dbms_handler,
-                                      CurrentDB &current_db, const InterpreterConfig &config) {
+                                      std::vector<Notification> *notifications,
+                                      ReplicationQueryHandler &replication_query_handler, CurrentDB &current_db,
+                                      const InterpreterConfig &config) {
   if (in_explicit_transaction) {
     throw ReplicationModificationInMulticommandTxException();
   }
 
   auto *replication_query = utils::Downcast<ReplicationQuery>(parsed_query.query);
-  auto callback = HandleReplicationQuery(replication_query, parsed_query.parameters, &dbms_handler, current_db, config,
-                                         notifications);
+  auto callback = HandleReplicationQuery(replication_query, parsed_query.parameters, replication_query_handler,
+                                         current_db, config, notifications);
 
   return PreparedQuery{callback.header, std::move(parsed_query.required_privileges),
                        [callback_fn = std::move(callback.fn), pull_plan = std::shared_ptr<PullPlanVector>{nullptr}](
@@ -2554,8 +2633,10 @@ PreparedQuery PrepareReplicationQuery(ParsedQuery parsed_query, bool in_explicit
   // NOLINTNEXTLINE(clang-analyzer-cplusplus.NewDeleteLeaks)
 }
 
+#ifdef MG_ENTERPRISE
 PreparedQuery PrepareCoordinatorQuery(ParsedQuery parsed_query, bool in_explicit_transaction,
-                                      std::vector<Notification> *notifications, dbms::DbmsHandler &dbms_handler,
+                                      std::vector<Notification> *notifications,
+                                      coordination::CoordinatorState &coordinator_state,
                                       const InterpreterConfig &config) {
   if (in_explicit_transaction) {
     throw CoordinatorModificationInMulticommandTxException();
@@ -2563,7 +2644,7 @@ PreparedQuery PrepareCoordinatorQuery(ParsedQuery parsed_query, bool in_explicit
 
   auto *coordinator_query = utils::Downcast<CoordinatorQuery>(parsed_query.query);
   auto callback =
-      HandleCoordinatorQuery(coordinator_query, parsed_query.parameters, &dbms_handler, config, notifications);
+      HandleCoordinatorQuery(coordinator_query, parsed_query.parameters, &coordinator_state, config, notifications);
 
   return PreparedQuery{callback.header, std::move(parsed_query.required_privileges),
                        [callback_fn = std::move(callback.fn), pull_plan = std::shared_ptr<PullPlanVector>{nullptr}](
@@ -2581,6 +2662,7 @@ PreparedQuery PrepareCoordinatorQuery(ParsedQuery parsed_query, bool in_explicit
   // False positive report for the std::make_shared above
   // NOLINTNEXTLINE(clang-analyzer-cplusplus.NewDeleteLeaks)
 }
+#endif
 
 PreparedQuery PrepareLockPathQuery(ParsedQuery parsed_query, bool in_explicit_transaction, CurrentDB &current_db) {
   if (in_explicit_transaction) {
@@ -3683,7 +3765,8 @@ PreparedQuery PrepareConstraintQuery(ParsedQuery parsed_query, bool in_explicit_
 
 PreparedQuery PrepareMultiDatabaseQuery(ParsedQuery parsed_query, CurrentDB &current_db,
                                         InterpreterContext *interpreter_context,
-                                        std::optional<std::function<void(std::string_view)>> on_change_cb) {
+                                        std::optional<std::function<void(std::string_view)>> on_change_cb,
+                                        Interpreter &interpreter) {
 #ifdef MG_ENTERPRISE
   if (!license::global_license_checker.IsEnterpriseValidFast()) {
     throw QueryException("Trying to use enterprise feature without a valid license.");
@@ -3702,12 +3785,16 @@ PreparedQuery PrepareMultiDatabaseQuery(ParsedQuery parsed_query, CurrentDB &cur
       return PreparedQuery{
           {"STATUS"},
           std::move(parsed_query.required_privileges),
-          [db_name = query->db_name_, db_handler](AnyStream *stream,
-                                                  std::optional<int> n) -> std::optional<QueryHandlerResult> {
+          [db_name = query->db_name_, db_handler, interpreter = &interpreter](
+              AnyStream *stream, std::optional<int> n) -> std::optional<QueryHandlerResult> {
+            if (!interpreter->system_transaction_) {
+              throw QueryException("Expected to be in a system transaction");
+            }
+
             std::vector<std::vector<TypedValue>> status;
             std::string res;
 
-            const auto success = db_handler->New(db_name);
+            const auto success = db_handler->New(db_name, &*interpreter->system_transaction_);
             if (success.HasError()) {
               switch (success.GetError()) {
                 case dbms::NewError::EXISTS:
@@ -3782,16 +3869,20 @@ PreparedQuery PrepareMultiDatabaseQuery(ParsedQuery parsed_query, CurrentDB &cur
       return PreparedQuery{
           {"STATUS"},
           std::move(parsed_query.required_privileges),
-          [db_name = query->db_name_, db_handler, auth = interpreter_context->auth](
+          [db_name = query->db_name_, db_handler, auth = interpreter_context->auth, interpreter = &interpreter](
               AnyStream *stream, std::optional<int> n) -> std::optional<QueryHandlerResult> {
+            if (!interpreter->system_transaction_) {
+              throw QueryException("Expected to be in a system transaction");
+            }
+
             std::vector<std::vector<TypedValue>> status;
 
             try {
               // Remove database
-              auto success = db_handler->TryDelete(db_name);
+              auto success = db_handler->TryDelete(db_name, &*interpreter->system_transaction_);
               if (!success.HasError()) {
                 // Remove from auth
-                if (auth) auth->DeleteDatabase(db_name);
+                if (auth) auth->DeleteDatabase(db_name, &*interpreter->system_transaction_);
               } else {
                 switch (success.GetError()) {
                   case dbms::DeleteError::DEFAULT_DB:
@@ -4042,18 +4133,15 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
                           utils::Downcast<ReplicationQuery>(parsed_query.query);
 
     // TODO Split SHOW REPLICAS (which needs the db) and other replication queries
-    auto system_transaction_guard = std::invoke([&]() -> std::optional<SystemTransactionGuard> {
-      if (system_queries) {
-        // TODO: Ordering between system and data queries
-        // Start a system transaction
-        auto system_unique = std::unique_lock{interpreter_context_->dbms_handler->system_lock_, std::defer_lock};
-        if (!system_unique.try_lock_for(std::chrono::milliseconds(kSystemTxTryMS))) {
-          throw ConcurrentSystemQueriesException("Multiple concurrent system queries are not supported.");
-        }
-        return std::optional<SystemTransactionGuard>{std::in_place, std::move(system_unique),
-                                                     *interpreter_context_->dbms_handler};
+    auto system_transaction = std::invoke([&]() -> std::optional<memgraph::system::Transaction> {
+      if (!system_queries) return std::nullopt;
+
+      // TODO: Ordering between system and data queries
+      auto system_txn = interpreter_context_->system_->TryCreateTransaction(std::chrono::milliseconds(kSystemTxTryMS));
+      if (!system_txn) {
+        throw ConcurrentSystemQueriesException("Multiple concurrent system queries are not supported.");
       }
-      return std::nullopt;
+      return system_txn;
     });
 
     // Some queries do not require a database to be executed (current_db_ won't be passed on to the Prepare*; special
@@ -4119,7 +4207,7 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
       prepared_query = PrepareAnalyzeGraphQuery(std::move(parsed_query), in_explicit_transaction_, current_db_);
     } else if (utils::Downcast<AuthQuery>(parsed_query.query)) {
       /// SYSTEM (Replication) PURE
-      prepared_query = PrepareAuthQuery(std::move(parsed_query), in_explicit_transaction_, interpreter_context_);
+      prepared_query = PrepareAuthQuery(std::move(parsed_query), in_explicit_transaction_, interpreter_context_, *this);
     } else if (utils::Downcast<DatabaseInfoQuery>(parsed_query.query)) {
       prepared_query = PrepareDatabaseInfoQuery(std::move(parsed_query), in_explicit_transaction_, current_db_);
     } else if (utils::Downcast<SystemInfoQuery>(parsed_query.query)) {
@@ -4130,13 +4218,18 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
                                               &query_execution->notifications, current_db_);
     } else if (utils::Downcast<ReplicationQuery>(parsed_query.query)) {
       /// TODO: make replication DB agnostic
-      prepared_query =
-          PrepareReplicationQuery(std::move(parsed_query), in_explicit_transaction_, &query_execution->notifications,
-                                  *interpreter_context_->dbms_handler, current_db_, interpreter_context_->config);
+      prepared_query = PrepareReplicationQuery(
+          std::move(parsed_query), in_explicit_transaction_, &query_execution->notifications,
+          *interpreter_context_->replication_handler_, current_db_, interpreter_context_->config);
+
     } else if (utils::Downcast<CoordinatorQuery>(parsed_query.query)) {
+#ifdef MG_ENTERPRISE
       prepared_query =
           PrepareCoordinatorQuery(std::move(parsed_query), in_explicit_transaction_, &query_execution->notifications,
-                                  *interpreter_context_->dbms_handler, interpreter_context_->config);
+                                  *interpreter_context_->coordinator_state_, interpreter_context_->config);
+#else
+      throw QueryRuntimeException("Coordinator queries are not part of community edition");
+#endif
     } else if (utils::Downcast<LockPathQuery>(parsed_query.query)) {
       prepared_query = PrepareLockPathQuery(std::move(parsed_query), in_explicit_transaction_, current_db_);
     } else if (utils::Downcast<FreeMemoryQuery>(parsed_query.query)) {
@@ -4179,8 +4272,8 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
       }
       /// SYSTEM (Replication) + INTERPRETER
       // DMG_ASSERT(system_guard);
-      prepared_query = PrepareMultiDatabaseQuery(std::move(parsed_query), current_db_, interpreter_context_, on_change_
-                                                 /*, *system_guard*/);
+      prepared_query =
+          PrepareMultiDatabaseQuery(std::move(parsed_query), current_db_, interpreter_context_, on_change_, *this);
     } else if (utils::Downcast<ShowDatabasesQuery>(parsed_query.query)) {
       prepared_query = PrepareShowDatabasesQuery(std::move(parsed_query), interpreter_context_, username_);
     } else if (utils::Downcast<EdgeImportModeQuery>(parsed_query.query)) {
@@ -4212,7 +4305,7 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
     query_execution->summary["db"] = *query_execution->prepared_query->db;
 
     // prepare is done, move system txn guard to be owned by interpreter
-    system_transaction_guard_ = std::move(system_transaction_guard);
+    system_transaction_ = std::move(system_transaction);
     return {query_execution->prepared_query->header, query_execution->prepared_query->privileges, qid,
             query_execution->prepared_query->db};
   } catch (const utils::BasicException &) {
@@ -4362,13 +4455,13 @@ void Interpreter::Commit() {
   current_transaction_.reset();
   if (!current_db_.db_transactional_accessor_ || !current_db_.db_acc_) {
     // No database nor db transaction; check for system transaction
-    if (!system_transaction_guard_) return;
+    if (!system_transaction_) return;
 
     // TODO Distinguish between data and system transaction state
     // Think about updating the status to a struct with bitfield
     // Clean transaction status on exit
     utils::OnScopeExit clean_status([this]() {
-      system_transaction_guard_.reset();
+      system_transaction_.reset();
       // System transactions are not terminable
       // Durability has happened at time of PULL
       // Commit is doing replication and timestamp update
@@ -4386,7 +4479,23 @@ void Interpreter::Commit() {
       }
     });
 
-    system_transaction_guard_->Commit();
+    auto const main_commit = [&](replication::RoleMainData &mainData) {
+    // Only enterprise can do system replication
+#ifdef MG_ENTERPRISE
+      if (license::global_license_checker.IsEnterpriseValidFast()) {
+        return system_transaction_->Commit(memgraph::system::DoReplication{mainData});
+      }
+#endif
+      return system_transaction_->Commit(memgraph::system::DoNothing{});
+    };
+
+    auto const replica_commit = [&](replication::RoleReplicaData &) {
+      return system_transaction_->Commit(memgraph::system::DoNothing{});
+    };
+
+    auto const commit_method = utils::Overloaded{main_commit, replica_commit};
+    [[maybe_unused]] auto sync_result = std::visit(commit_method, interpreter_context_->repl_state->ReplicationData());
+    // TODO: something with sync_result
     return;
   }
   auto *db = current_db_.db_acc_->get();

src/query/interpreter.hpp

@@ -72,6 +72,7 @@ inline constexpr size_t kExecutionPoolMaxBlockSize = 1024UL;  // 2 ^ 10
 
 enum class QueryHandlerResult { COMMIT, ABORT, NOTHING };
 
+#ifdef MG_ENTERPRISE
 class CoordinatorQueryHandler {
  public:
   CoordinatorQueryHandler() = default;
@@ -93,7 +94,6 @@ class CoordinatorQueryHandler {
     ReplicationQuery::ReplicaState state;
   };
 
-#ifdef MG_ENTERPRISE
   struct MainReplicaStatus {
     std::string_view name;
     std::string_view socket_address;
@@ -103,9 +103,7 @@ class CoordinatorQueryHandler {
     MainReplicaStatus(std::string_view name, std::string_view socket_address, bool alive, bool is_main)
         : name{name}, socket_address{socket_address}, alive{alive}, is_main{is_main} {}
   };
-#endif
 
-#ifdef MG_ENTERPRISE
   /// @throw QueryRuntimeException if an error ocurred.
   virtual void RegisterInstance(const std::string &coordinator_socket_address,
                                 const std::string &replication_socket_address,
@@ -117,9 +115,8 @@ class CoordinatorQueryHandler {
 
   /// @throw QueryRuntimeException if an error ocurred.
   virtual std::vector<coordination::CoordinatorInstanceStatus> ShowInstances() const = 0;
-
-#endif
 };
+#endif
 
 class AnalyzeGraphQueryHandler {
  public:
@@ -296,32 +293,12 @@ class Interpreter final {
 
   void SetUser(std::string_view username);
 
-  struct SystemTransactionGuard {
-    explicit SystemTransactionGuard(std::unique_lock<utils::ResourceLock> guard, dbms::DbmsHandler &dbms_handler)
-        : system_guard_(std::move(guard)), dbms_handler_{&dbms_handler} {
-      dbms_handler_->NewSystemTransaction();
-    }
-    SystemTransactionGuard &operator=(SystemTransactionGuard &&) = default;
-    SystemTransactionGuard(SystemTransactionGuard &&) = default;
-
-    ~SystemTransactionGuard() {
-      if (system_guard_.owns_lock()) dbms_handler_->ResetSystemTransaction();
-    }
-
-    dbms::AllSyncReplicaStatus Commit() { return dbms_handler_->Commit(); }
-
-   private:
-    std::unique_lock<utils::ResourceLock> system_guard_;
-    dbms::DbmsHandler *dbms_handler_;
-  };
-
-  std::optional<SystemTransactionGuard> system_transaction_guard_{};
+  std::optional<memgraph::system::Transaction> system_transaction_{};
 
  private:
   void ResetInterpreter() {
     query_executions_.clear();
-    system_guard.reset();
-    system_transaction_guard_.reset();
+    system_transaction_.reset();
     transaction_queries_->clear();
     if (current_db_.db_acc_ && current_db_.db_acc_->is_deleting()) {
       current_db_.db_acc_.reset();
@@ -386,8 +363,6 @@ class Interpreter final {
   // TODO Figure out how this would work for multi-database
   // Exists only during a single transaction (for now should be okay as is)
   std::vector<std::unique_ptr<QueryExecution>> query_executions_;
-  // TODO: our upgradable lock guard for system
-  std::optional<utils::ResourceLockGuard> system_guard;
 
   // all queries that are run as part of the current transaction
   utils::Synchronized<std::vector<std::string>, utils::SpinLock> transaction_queries_;

src/query/interpreter_context.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -12,12 +12,27 @@
 #include "query/interpreter_context.hpp"
 
 #include "query/interpreter.hpp"
+#include "system/include/system/system.hpp"
 namespace memgraph::query {
 
 InterpreterContext::InterpreterContext(InterpreterConfig interpreter_config, dbms::DbmsHandler *dbms_handler,
-                                       replication::ReplicationState *rs, query::AuthQueryHandler *ah,
-                                       query::AuthChecker *ac)
-    : dbms_handler(dbms_handler), config(interpreter_config), repl_state(rs), auth(ah), auth_checker(ac) {}
+                                       replication::ReplicationState *rs, memgraph::system::System &system,
+#ifdef MG_ENTERPRISE
+                                       memgraph::coordination::CoordinatorState *coordinator_state,
+#endif
+                                       AuthQueryHandler *ah, AuthChecker *ac,
+                                       ReplicationQueryHandler *replication_handler)
+    : dbms_handler(dbms_handler),
+      config(interpreter_config),
+      repl_state(rs),
+#ifdef MG_ENTERPRISE
+      coordinator_state_{coordinator_state},
+#endif
+      auth(ah),
+      auth_checker(ac),
+      replication_handler_{replication_handler},
+      system_{&system} {
+}
 
 std::vector<std::vector<TypedValue>> InterpreterContext::TerminateTransactions(
     std::vector<std::string> maybe_kill_transaction_ids, const std::optional<std::string> &username,

src/query/interpreter_context.hpp

@@ -20,14 +20,20 @@
 
 #include "query/config.hpp"
 #include "query/cypher_query_interpreter.hpp"
+#include "query/replication_query_handler.hpp"
 #include "query/typed_value.hpp"
 #include "replication/state.hpp"
 #include "storage/v2/config.hpp"
 #include "storage/v2/transaction.hpp"
+#include "system/state.hpp"
+#include "system/system.hpp"
 #include "utils/gatekeeper.hpp"
 #include "utils/skip_list.hpp"
 #include "utils/spin_lock.hpp"
 #include "utils/synchronized.hpp"
+#ifdef MG_ENTERPRISE
+#include "coordination/coordinator_state.hpp"
+#endif
 
 namespace memgraph::dbms {
 class DbmsHandler;
@@ -48,7 +54,12 @@ class Interpreter;
  */
 struct InterpreterContext {
   InterpreterContext(InterpreterConfig interpreter_config, dbms::DbmsHandler *dbms_handler,
-                     replication::ReplicationState *rs, AuthQueryHandler *ah = nullptr, AuthChecker *ac = nullptr);
+                     replication::ReplicationState *rs, memgraph::system::System &system,
+#ifdef MG_ENTERPRISE
+                     memgraph::coordination::CoordinatorState *coordinator_state,
+#endif
+                     AuthQueryHandler *ah = nullptr, AuthChecker *ac = nullptr,
+                     ReplicationQueryHandler *replication_handler = nullptr);
 
   memgraph::dbms::DbmsHandler *dbms_handler;
 
@@ -59,9 +70,14 @@ struct InterpreterContext {
 
   // GLOBAL
   memgraph::replication::ReplicationState *repl_state;
+#ifdef MG_ENTERPRISE
+  memgraph::coordination::CoordinatorState *coordinator_state_;
+#endif
 
   AuthQueryHandler *auth;
   AuthChecker *auth_checker;
+  ReplicationQueryHandler *replication_handler_;
+  system::System *system_;
 
   // Used to check active transactions
   // TODO: Have a way to read the current database

src/query/replication_query_handler.hpp

@@ -0,0 +1,60 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include "replication_coordination_glue/role.hpp"
+#include "utils/result.hpp"
+
+// BEGIN fwd declares
+namespace memgraph::replication {
+struct ReplicationState;
+struct ReplicationServerConfig;
+struct ReplicationClientConfig;
+}  // namespace memgraph::replication
+
+namespace memgraph::query {
+
+enum class RegisterReplicaError : uint8_t { NAME_EXISTS, ENDPOINT_EXISTS, CONNECTION_FAILED, COULD_NOT_BE_PERSISTED };
+enum class UnregisterReplicaResult : uint8_t {
+  NOT_MAIN,
+  COULD_NOT_BE_PERSISTED,
+  CAN_NOT_UNREGISTER,
+  SUCCESS,
+};
+
+/// A handler type that keep in sync current ReplicationState and the MAIN/REPLICA-ness of Storage
+struct ReplicationQueryHandler {
+  virtual ~ReplicationQueryHandler() = default;
+
+  // as REPLICA, become MAIN
+  virtual bool SetReplicationRoleMain() = 0;
+
+  // as MAIN, become REPLICA
+  virtual bool SetReplicationRoleReplica(const memgraph::replication::ReplicationServerConfig &config) = 0;
+
+  // as MAIN, define and connect to REPLICAs
+  virtual auto TryRegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
+      -> utils::BasicResult<RegisterReplicaError> = 0;
+
+  virtual auto RegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
+      -> utils::BasicResult<RegisterReplicaError> = 0;
+
+  // as MAIN, remove a REPLICA connection
+  virtual auto UnregisterReplica(std::string_view name) -> UnregisterReplicaResult = 0;
+
+  // Helper pass-through (TODO: remove)
+  virtual auto GetRole() const -> memgraph::replication_coordination_glue::ReplicationRole = 0;
+  virtual bool IsMain() const = 0;
+  virtual bool IsReplica() const = 0;
+};
+
+}  // namespace memgraph::query

src/replication/CMakeLists.txt

@@ -6,7 +6,6 @@ target_sources(mg-replication
         include/replication/epoch.hpp
         include/replication/config.hpp
         include/replication/status.hpp
-        include/replication/messages.hpp
         include/replication/replication_client.hpp
         include/replication/replication_server.hpp
 
@@ -15,7 +14,6 @@ target_sources(mg-replication
         epoch.cpp
         config.cpp
         status.cpp
-        messages.cpp
         replication_client.cpp
         replication_server.cpp
 )

src/replication/include/replication/messages.hpp

@@ -1,47 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-
-#pragma once
-
-#include "rpc/messages.hpp"
-#include "slk/serialization.hpp"
-
-namespace memgraph::replication {
-struct SystemHeartbeatReq {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  static void Load(SystemHeartbeatReq *self, memgraph::slk::Reader *reader);
-  static void Save(const SystemHeartbeatReq &self, memgraph::slk::Builder *builder);
-  SystemHeartbeatReq() = default;
-};
-
-struct SystemHeartbeatRes {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  static void Load(SystemHeartbeatRes *self, memgraph::slk::Reader *reader);
-  static void Save(const SystemHeartbeatRes &self, memgraph::slk::Builder *builder);
-  SystemHeartbeatRes() = default;
-  explicit SystemHeartbeatRes(uint64_t system_timestamp) : system_timestamp(system_timestamp) {}
-
-  uint64_t system_timestamp;
-};
-
-using SystemHeartbeatRpc = rpc::RequestResponse<SystemHeartbeatReq, SystemHeartbeatRes>;
-}  // namespace memgraph::replication
-
-namespace memgraph::slk {
-void Save(const memgraph::replication::SystemHeartbeatRes &self, memgraph::slk::Builder *builder);
-void Load(memgraph::replication::SystemHeartbeatRes *self, memgraph::slk::Reader *reader);
-void Save(const memgraph::replication::SystemHeartbeatReq & /*self*/, memgraph::slk::Builder * /*builder*/);
-void Load(memgraph::replication::SystemHeartbeatReq * /*self*/, memgraph::slk::Reader * /*reader*/);
-}  // namespace memgraph::slk

src/replication/include/replication/replication_client.hpp

@@ -41,26 +41,67 @@ struct ReplicationClient {
   void StartFrequentCheck(F &&callback) {
     // Help the user to get the most accurate replica state possible.
     if (replica_check_frequency_ > std::chrono::seconds(0)) {
-      replica_checker_.Run("Replica Checker", replica_check_frequency_,
-                           [this, cb = std::forward<F>(callback), reconnect = false]() mutable {
-                             try {
-                               {
-                                 auto stream{rpc_client_.Stream<memgraph::replication_coordination_glue::FrequentHeartbeatRpc>()};
-                                 stream.AwaitResponse();
-                               }
-                               cb(reconnect, *this);
-                               reconnect = false;
-                             } catch (const rpc::RpcFailedException &) {
-                               // Nothing to do...wait for a reconnect
-                               // NOTE: Here we are communicating with the instance connection.
-                               //       We don't have access to the undelying client; so the only thing we can do it
-                               //       tell the callback that this is a reconnection and to check the state
-                               reconnect = true;
-                             }
-                           });
+      replica_checker_.Run(
+          "Replica Checker", replica_check_frequency_,
+          [this, cb = std::forward<F>(callback), reconnect = false]() mutable {
+            try {
+              {
+                auto stream{rpc_client_.Stream<memgraph::replication_coordination_glue::FrequentHeartbeatRpc>()};
+                stream.AwaitResponse();
+              }
+              cb(reconnect, *this);
+              reconnect = false;
+            } catch (const rpc::RpcFailedException &) {
+              // Nothing to do...wait for a reconnect
+              // NOTE: Here we are communicating with the instance connection.
+              //       We don't have access to the undelying client; so the only thing we can do it
+              //       tell the callback that this is a reconnection and to check the state
+              reconnect = true;
+            }
+          });
     }
   }
 
+  //! \tparam RPC An rpc::RequestResponse
+  //! \tparam Args the args type
+  //! \param client the client to use for rpc communication
+  //! \param check predicate to check response is ok
+  //! \param args arguments to forward to the rpc request
+  //! \return If replica stream is completed or enqueued
+  template <typename RPC, typename... Args>
+  bool SteamAndFinalizeDelta(auto &&check, Args &&...args) {
+    try {
+      auto stream = rpc_client_.template Stream<RPC>(std::forward<Args>(args)...);
+      auto task = [this, check = std::forward<decltype(check)>(check), stream = std::move(stream)]() mutable {
+        if (stream.IsDefunct()) {
+          state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+          return false;
+        }
+        try {
+          if (check(stream.AwaitResponse())) {
+            return true;
+          }
+        } catch (memgraph::rpc::GenericRpcFailedException const &e) {
+          // swallow error, fallthrough to error handling
+        }
+        // This replica needs SYSTEM recovery
+        state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+        return false;
+      };
+
+      if (mode_ == memgraph::replication_coordination_glue::ReplicationMode::ASYNC) {
+        thread_pool_.AddTask([task = utils::CopyMovableFunctionWrapper{std::move(task)}]() mutable { task(); });
+        return true;
+      }
+
+      return task();
+    } catch (memgraph::rpc::GenericRpcFailedException const &e) {
+      // This replica needs SYSTEM recovery
+      state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+      return false;
+    }
+  };
+
   std::string name_;
   communication::ClientContext rpc_context_;
   rpc::Client rpc_client_;

src/replication/messages.cpp

@@ -1,52 +0,0 @@
-// Copyright 2024 Memgraph Ltd.
-//
-// Use of this software is governed by the Business Source License
-// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-// License, and you may not use this file except in compliance with the Business Source License.
-//
-// As of the Change Date specified in that file, in accordance with
-// the Business Source License, use of this software will be governed
-// by the Apache License, Version 2.0, included in the file
-// licenses/APL.txt.
-#include "replication/messages.hpp"
-
-namespace memgraph::replication {
-
-constexpr utils::TypeInfo SystemHeartbeatReq::kType{utils::TypeId::REP_SYSTEM_HEARTBEAT_REQ, "SystemHeartbeatReq",
-                                                    nullptr};
-
-constexpr utils::TypeInfo SystemHeartbeatRes::kType{utils::TypeId::REP_SYSTEM_HEARTBEAT_RES, "SystemHeartbeatRes",
-                                                    nullptr};
-
-void SystemHeartbeatReq::Save(const SystemHeartbeatReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void SystemHeartbeatReq::Load(SystemHeartbeatReq *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(self, reader);
-}
-void SystemHeartbeatRes::Save(const SystemHeartbeatRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void SystemHeartbeatRes::Load(SystemHeartbeatRes *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(self, reader);
-}
-
-}  // namespace memgraph::replication
-
-namespace memgraph::slk {
-// Serialize code for SystemHeartbeatRes
-void Save(const memgraph::replication::SystemHeartbeatRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self.system_timestamp, builder);
-}
-void Load(memgraph::replication::SystemHeartbeatRes *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(&self->system_timestamp, reader);
-}
-
-// Serialize code for SystemHeartbeatReq
-void Save(const memgraph::replication::SystemHeartbeatReq & /*self*/, memgraph::slk::Builder * /*builder*/) {
-  /* Nothing to serialize */
-}
-void Load(memgraph::replication::SystemHeartbeatReq * /*self*/, memgraph::slk::Reader * /*reader*/) {
-  /* Nothing to serialize */
-}
-}  // namespace memgraph::slk

src/replication_handler/CMakeLists.txt

@@ -0,0 +1,17 @@
+add_library(mg-replication_handler STATIC)
+add_library(mg::replication_handler ALIAS mg-replication_handler)
+target_sources(mg-replication_handler
+        PUBLIC
+        include/replication_handler/replication_handler.hpp
+        include/replication_handler/system_replication.hpp
+        include/replication_handler/system_rpc.hpp
+
+        PRIVATE
+        replication_handler.cpp
+        system_replication.cpp
+        system_rpc.cpp
+        )
+target_include_directories(mg-replication_handler PUBLIC include)
+
+target_link_libraries(mg-replication_handler
+        PUBLIC mg-auth mg-dbms mg-replication)

src/replication_handler/include/replication_handler/replication_handler.hpp

@@ -0,0 +1,220 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+#pragma once
+
+#include "auth/auth.hpp"
+#include "dbms/dbms_handler.hpp"
+#include "replication/include/replication/state.hpp"
+#include "replication_handler/system_rpc.hpp"
+#include "utils/result.hpp"
+
+namespace memgraph::replication {
+
+inline std::optional<query::RegisterReplicaError> HandleRegisterReplicaStatus(
+    utils::BasicResult<replication::RegisterReplicaError, replication::ReplicationClient *> &instance_client);
+
+#ifdef MG_ENTERPRISE
+void StartReplicaClient(replication::ReplicationClient &client, system::System *system, dbms::DbmsHandler &dbms_handler,
+                        auth::SynchedAuth &auth);
+#else
+void StartReplicaClient(replication::ReplicationClient &client, dbms::DbmsHandler &dbms_handler);
+#endif
+
+#ifdef MG_ENTERPRISE
+// TODO: Split into 2 functions: dbms and auth
+// When being called by interpreter no need to gain lock, it should already be under a system transaction
+// But concurrently the FrequentCheck is running and will need to lock before reading last_committed_system_timestamp_
+template <bool REQUIRE_LOCK = false>
+void SystemRestore(replication::ReplicationClient &client, system::System *system, dbms::DbmsHandler &dbms_handler,
+                   auth::SynchedAuth &auth) {
+  // Check if system is up to date
+  if (client.state_.WithLock(
+          [](auto &state) { return state == memgraph::replication::ReplicationClient::State::READY; }))
+    return;
+
+  // Try to recover...
+  {
+    struct DbInfo {
+      std::vector<storage::SalientConfig> configs;
+      uint64_t last_committed_timestamp;
+    };
+    DbInfo db_info = std::invoke([&] {
+      auto guard = std::invoke([&]() -> std::optional<memgraph::system::TransactionGuard> {
+        if constexpr (REQUIRE_LOCK) {
+          return system->GenTransactionGuard();
+        }
+        return std::nullopt;
+      });
+
+      if (license::global_license_checker.IsEnterpriseValidFast()) {
+        auto configs = std::vector<storage::SalientConfig>{};
+        dbms_handler.ForEach([&configs](dbms::DatabaseAccess acc) { configs.emplace_back(acc->config().salient); });
+        // TODO: This is `SystemRestore` maybe DbInfo is incorrect as it will need Auth also
+        return DbInfo{configs, system->LastCommittedSystemTimestamp()};
+      }
+
+      // No license -> send only default config
+      return DbInfo{{dbms_handler.Get()->config().salient}, system->LastCommittedSystemTimestamp()};
+    });
+    try {
+      auto stream = std::invoke([&]() {
+        // Handle only default database is no license
+        if (!license::global_license_checker.IsEnterpriseValidFast()) {
+          return client.rpc_client_.Stream<replication::SystemRecoveryRpc>(
+              db_info.last_committed_timestamp, std::move(db_info.configs), auth::Auth::Config{},
+              std::vector<auth::User>{}, std::vector<auth::Role>{});
+        }
+        return auth.WithLock([&](auto &locked_auth) {
+          return client.rpc_client_.Stream<replication::SystemRecoveryRpc>(
+              db_info.last_committed_timestamp, std::move(db_info.configs), locked_auth.GetConfig(),
+              locked_auth.AllUsers(), locked_auth.AllRoles());
+        });
+      });
+      const auto response = stream.AwaitResponse();
+      if (response.result == replication::SystemRecoveryRes::Result::FAILURE) {
+        client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+        return;
+      }
+    } catch (memgraph::rpc::GenericRpcFailedException const &e) {
+      client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+      return;
+    }
+  }
+
+  // Successfully recovered
+  client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::READY; });
+}
+#endif
+
+/// A handler type that keep in sync current ReplicationState and the MAIN/REPLICA-ness of Storage
+struct ReplicationHandler : public memgraph::query::ReplicationQueryHandler {
+#ifdef MG_ENTERPRISE
+  explicit ReplicationHandler(memgraph::replication::ReplicationState &repl_state,
+                              memgraph::dbms::DbmsHandler &dbms_handler, memgraph::system::System *system,
+                              memgraph::auth::SynchedAuth &auth);
+#else
+  explicit ReplicationHandler(memgraph::replication::ReplicationState &repl_state,
+                              memgraph::dbms::DbmsHandler &dbms_handler);
+#endif
+
+  // as REPLICA, become MAIN
+  bool SetReplicationRoleMain() override;
+
+  // as MAIN, become REPLICA
+  bool SetReplicationRoleReplica(const memgraph::replication::ReplicationServerConfig &config) override;
+
+  // as MAIN, define and connect to REPLICAs
+  auto TryRegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
+      -> memgraph::utils::BasicResult<memgraph::query::RegisterReplicaError> override;
+
+  auto RegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
+      -> memgraph::utils::BasicResult<memgraph::query::RegisterReplicaError> override;
+
+  // as MAIN, remove a REPLICA connection
+  auto UnregisterReplica(std::string_view name) -> memgraph::query::UnregisterReplicaResult override;
+
+  bool DoReplicaToMainPromotion();
+
+  // Helper pass-through (TODO: remove)
+  auto GetRole() const -> memgraph::replication_coordination_glue::ReplicationRole override;
+  bool IsMain() const override;
+  bool IsReplica() const override;
+
+ private:
+  template <bool HandleFailure>
+  auto RegisterReplica_(const memgraph::replication::ReplicationClientConfig &config)
+      -> memgraph::utils::BasicResult<memgraph::query::RegisterReplicaError> {
+    MG_ASSERT(repl_state_.IsMain(), "Only main instance can register a replica!");
+
+    auto maybe_client = repl_state_.RegisterReplica(config);
+    if (maybe_client.HasError()) {
+      switch (maybe_client.GetError()) {
+        case memgraph::replication::RegisterReplicaError::NOT_MAIN:
+          MG_ASSERT(false, "Only main instance can register a replica!");
+          return {};
+        case memgraph::replication::RegisterReplicaError::NAME_EXISTS:
+          return memgraph::query::RegisterReplicaError::NAME_EXISTS;
+        case memgraph::replication::RegisterReplicaError::ENDPOINT_EXISTS:
+          return memgraph::query::RegisterReplicaError::ENDPOINT_EXISTS;
+        case memgraph::replication::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
+          return memgraph::query::RegisterReplicaError::COULD_NOT_BE_PERSISTED;
+        case memgraph::replication::RegisterReplicaError::SUCCESS:
+          break;
+      }
+    }
+
+    if (!memgraph::dbms::allow_mt_repl && dbms_handler_.All().size() > 1) {
+      spdlog::warn("Multi-tenant replication is currently not supported!");
+    }
+
+#ifdef MG_ENTERPRISE
+    // Update system before enabling individual storage <-> replica clients
+    SystemRestore(*maybe_client.GetValue(), system_, dbms_handler_, auth_);
+#endif
+
+    const auto dbms_error = HandleRegisterReplicaStatus(maybe_client);
+    if (dbms_error.has_value()) {
+      return *dbms_error;
+    }
+    auto &instance_client_ptr = maybe_client.GetValue();
+
+    bool all_clients_good = true;
+    // Add database specific clients (NOTE Currently all databases are connected to each replica)
+    dbms_handler_.ForEach([&](dbms::DatabaseAccess db_acc) {
+      auto *storage = db_acc->storage();
+      if (!dbms::allow_mt_repl && storage->name() != dbms::kDefaultDB) {
+        return;
+      }
+      // TODO: ATM only IN_MEMORY_TRANSACTIONAL, fix other modes
+      if (storage->storage_mode_ != storage::StorageMode::IN_MEMORY_TRANSACTIONAL) return;
+
+      all_clients_good &= storage->repl_storage_state_.replication_clients_.WithLock(
+          [storage, &instance_client_ptr, db_acc = std::move(db_acc)](auto &storage_clients) mutable {  // NOLINT
+            auto client = std::make_unique<storage::ReplicationStorageClient>(*instance_client_ptr);
+            // All good, start replica client
+            client->Start(storage, std::move(db_acc));
+            // After start the storage <-> replica state should be READY or RECOVERING (if correctly started)
+            // MAYBE_BEHIND isn't a statement of the current state, this is the default value
+            // Failed to start due an error like branching of MAIN and REPLICA
+            const bool success = client->State() != storage::replication::ReplicaState::MAYBE_BEHIND;
+            if (HandleFailure || success) {
+              storage_clients.push_back(std::move(client));
+            }
+            return success;
+          });
+    });
+
+    // NOTE Currently if any databases fails, we revert back
+    if (!HandleFailure && !all_clients_good) {
+      spdlog::error("Failed to register all databases on the REPLICA \"{}\"", config.name);
+      UnregisterReplica(config.name);
+      return memgraph::query::RegisterReplicaError::CONNECTION_FAILED;
+    }
+
+    // No client error, start instance level client
+#ifdef MG_ENTERPRISE
+    StartReplicaClient(*instance_client_ptr, system_, dbms_handler_, auth_);
+#else
+    StartReplicaClient(*instance_client_ptr, dbms_handler_);
+#endif
+    return {};
+  }
+
+  memgraph::replication::ReplicationState &repl_state_;
+  memgraph::dbms::DbmsHandler &dbms_handler_;
+
+#ifdef MG_ENTERPRISE
+  memgraph::system::System *system_;
+  memgraph::auth::SynchedAuth &auth_;
+#endif
+};
+
+}  // namespace memgraph::replication

src/replication_handler/include/replication_handler/system_replication.hpp

@@ -0,0 +1,31 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include "auth/auth.hpp"
+#include "dbms/dbms_handler.hpp"
+#include "slk/streams.hpp"
+#include "system/state.hpp"
+
+namespace memgraph::replication {
+#ifdef MG_ENTERPRISE
+void SystemHeartbeatHandler(uint64_t ts, slk::Reader *req_reader, slk::Builder *res_builder);
+void SystemRecoveryHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                           dbms::DbmsHandler &dbms_handler, auth::SynchedAuth &auth, slk::Reader *req_reader,
+                           slk::Builder *res_builder);
+void Register(replication::RoleReplicaData const &data, dbms::DbmsHandler &dbms_handler, auth::SynchedAuth &auth);
+bool StartRpcServer(dbms::DbmsHandler &dbms_handler, const replication::RoleReplicaData &data, auth::SynchedAuth &auth);
+#else
+bool StartRpcServer(dbms::DbmsHandler &dbms_handler, const replication::RoleReplicaData &data);
+#endif
+
+}  // namespace memgraph::replication

src/replication_handler/include/replication_handler/system_rpc.hpp

@@ -0,0 +1,95 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include <cstdint>
+#include <vector>
+
+#include "auth/auth.hpp"
+#include "auth/models.hpp"
+#include "rpc/messages.hpp"
+#include "storage/v2/config.hpp"
+
+namespace memgraph::replication {
+struct SystemHeartbeatReq {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(SystemHeartbeatReq *self, memgraph::slk::Reader *reader);
+  static void Save(const SystemHeartbeatReq &self, memgraph::slk::Builder *builder);
+  SystemHeartbeatReq() = default;
+};
+
+struct SystemHeartbeatRes {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(SystemHeartbeatRes *self, memgraph::slk::Reader *reader);
+  static void Save(const SystemHeartbeatRes &self, memgraph::slk::Builder *builder);
+  SystemHeartbeatRes() = default;
+  explicit SystemHeartbeatRes(uint64_t system_timestamp) : system_timestamp(system_timestamp) {}
+
+  uint64_t system_timestamp;
+};
+
+using SystemHeartbeatRpc = rpc::RequestResponse<SystemHeartbeatReq, SystemHeartbeatRes>;
+
+struct SystemRecoveryReq {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  static void Load(SystemRecoveryReq *self, memgraph::slk::Reader *reader);
+  static void Save(const SystemRecoveryReq &self, memgraph::slk::Builder *builder);
+  SystemRecoveryReq() = default;
+  SystemRecoveryReq(uint64_t forced_group_timestamp, std::vector<storage::SalientConfig> database_configs,
+                    auth::Auth::Config auth_config, std::vector<auth::User> users, std::vector<auth::Role> roles)
+      : forced_group_timestamp{forced_group_timestamp},
+        database_configs(std::move(database_configs)),
+        auth_config(std::move(auth_config)),
+        users{std::move(users)},
+        roles{std::move(roles)} {}
+
+  uint64_t forced_group_timestamp;
+  std::vector<storage::SalientConfig> database_configs;
+  auth::Auth::Config auth_config;
+  std::vector<auth::User> users;
+  std::vector<auth::Role> roles;
+};
+
+struct SystemRecoveryRes {
+  static const utils::TypeInfo kType;
+  static const utils::TypeInfo &GetTypeInfo() { return kType; }
+
+  enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
+
+  static void Load(SystemRecoveryRes *self, memgraph::slk::Reader *reader);
+  static void Save(const SystemRecoveryRes &self, memgraph::slk::Builder *builder);
+  SystemRecoveryRes() = default;
+  explicit SystemRecoveryRes(Result res) : result(res) {}
+
+  Result result;
+};
+
+using SystemRecoveryRpc = rpc::RequestResponse<SystemRecoveryReq, SystemRecoveryRes>;
+
+}  // namespace memgraph::replication
+
+namespace memgraph::slk {
+void Save(const memgraph::replication::SystemHeartbeatRes &self, memgraph::slk::Builder *builder);
+void Load(memgraph::replication::SystemHeartbeatRes *self, memgraph::slk::Reader *reader);
+void Save(const memgraph::replication::SystemHeartbeatReq & /*self*/, memgraph::slk::Builder * /*builder*/);
+void Load(memgraph::replication::SystemHeartbeatReq * /*self*/, memgraph::slk::Reader * /*reader*/);
+void Save(const memgraph::replication::SystemRecoveryReq &self, memgraph::slk::Builder *builder);
+void Load(memgraph::replication::SystemRecoveryReq *self, memgraph::slk::Reader *reader);
+void Save(const memgraph::replication::SystemRecoveryRes &self, memgraph::slk::Builder *builder);
+void Load(memgraph::replication::SystemRecoveryRes *self, memgraph::slk::Reader *reader);
+}  // namespace memgraph::slk

src/replication_handler/replication_handler.cpp

@@ -0,0 +1,291 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "replication_handler/replication_handler.hpp"
+#include "dbms/dbms_handler.hpp"
+#include "replication_handler/system_replication.hpp"
+
+namespace memgraph::replication {
+
+namespace {
+#ifdef MG_ENTERPRISE
+void RecoverReplication(memgraph::replication::ReplicationState &repl_state, memgraph::system::System *system,
+                        memgraph::dbms::DbmsHandler &dbms_handler, memgraph::auth::SynchedAuth &auth) {
+  /*
+   * REPLICATION RECOVERY AND STARTUP
+   */
+
+  // Startup replication state (if recovered at startup)
+  auto replica = [&dbms_handler, &auth](memgraph::replication::RoleReplicaData const &data) {
+    return memgraph::replication::StartRpcServer(dbms_handler, data, auth);
+  };
+
+  // Replication recovery and frequent check start
+  auto main = [system, &dbms_handler, &auth](memgraph::replication::RoleMainData &mainData) {
+    for (auto &client : mainData.registered_replicas_) {
+      memgraph::replication::SystemRestore(client, system, dbms_handler, auth);
+    }
+    // DBMS here
+    dbms_handler.ForEach([&mainData](memgraph::dbms::DatabaseAccess db_acc) {
+      dbms::DbmsHandler::RecoverStorageReplication(std::move(db_acc), mainData);
+    });
+
+    for (auto &client : mainData.registered_replicas_) {
+      memgraph::replication::StartReplicaClient(client, system, dbms_handler, auth);
+    }
+
+    // Warning
+    if (dbms_handler.default_config().durability.snapshot_wal_mode ==
+        memgraph::storage::Config::Durability::SnapshotWalMode::DISABLED) {
+      spdlog::warn(
+          "The instance has the MAIN replication role, but durability logs and snapshots are disabled. Please "
+          "consider "
+          "enabling durability by using --storage-snapshot-interval-sec and --storage-wal-enabled flags because "
+          "without write-ahead logs this instance is not replicating any data.");
+    }
+
+    return true;
+  };
+
+  auto result = std::visit(memgraph::utils::Overloaded{replica, main}, repl_state.ReplicationData());
+  MG_ASSERT(result, "Replica recovery failure!");
+}
+#else
+void RecoverReplication(memgraph::replication::ReplicationState &repl_state,
+                        memgraph::dbms::DbmsHandler &dbms_handler) {
+  // Startup replication state (if recovered at startup)
+  auto replica = [&dbms_handler](memgraph::replication::RoleReplicaData const &data) {
+    return memgraph::replication::StartRpcServer(dbms_handler, data);
+  };
+
+  // Replication recovery and frequent check start
+  auto main = [&dbms_handler](memgraph::replication::RoleMainData &mainData) {
+    dbms::DbmsHandler::RecoverStorageReplication(dbms_handler.Get(), mainData);
+
+    for (auto &client : mainData.registered_replicas_) {
+      memgraph::replication::StartReplicaClient(client, dbms_handler);
+    }
+
+    // Warning
+    if (dbms_handler.default_config().durability.snapshot_wal_mode ==
+        memgraph::storage::Config::Durability::SnapshotWalMode::DISABLED) {
+      spdlog::warn(
+          "The instance has the MAIN replication role, but durability logs and snapshots are disabled. Please "
+          "consider "
+          "enabling durability by using --storage-snapshot-interval-sec and --storage-wal-enabled flags because "
+          "without write-ahead logs this instance is not replicating any data.");
+    }
+
+    return true;
+  };
+
+  auto result = std::visit(memgraph::utils::Overloaded{replica, main}, repl_state.ReplicationData());
+  MG_ASSERT(result, "Replica recovery failure!");
+}
+#endif
+}  // namespace
+
+inline std::optional<query::RegisterReplicaError> HandleRegisterReplicaStatus(
+    utils::BasicResult<replication::RegisterReplicaError, replication::ReplicationClient *> &instance_client) {
+  if (instance_client.HasError()) switch (instance_client.GetError()) {
+      case replication::RegisterReplicaError::NOT_MAIN:
+        MG_ASSERT(false, "Only main instance can register a replica!");
+        return {};
+      case replication::RegisterReplicaError::NAME_EXISTS:
+        return query::RegisterReplicaError::NAME_EXISTS;
+      case replication::RegisterReplicaError::ENDPOINT_EXISTS:
+        return query::RegisterReplicaError::ENDPOINT_EXISTS;
+      case replication::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
+        return query::RegisterReplicaError::COULD_NOT_BE_PERSISTED;
+      case replication::RegisterReplicaError::SUCCESS:
+        break;
+    }
+  return {};
+}
+
+#ifdef MG_ENTERPRISE
+void StartReplicaClient(replication::ReplicationClient &client, system::System *system, dbms::DbmsHandler &dbms_handler,
+                        auth::SynchedAuth &auth) {
+#else
+void StartReplicaClient(replication::ReplicationClient &client, dbms::DbmsHandler &dbms_handler) {
+#endif
+  // No client error, start instance level client
+  auto const &endpoint = client.rpc_client_.Endpoint();
+  spdlog::trace("Replication client started at: {}:{}", endpoint.address, endpoint.port);
+  client.StartFrequentCheck([&,
+#ifdef MG_ENTERPRISE
+                             system = system,
+#endif
+                             license = license::global_license_checker.IsEnterpriseValidFast()](
+                                bool reconnect, replication::ReplicationClient &client) mutable {
+    // Working connection
+    // Check if system needs restoration
+    if (reconnect) {
+      client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+    }
+    // Check if license has changed
+    const auto new_license = license::global_license_checker.IsEnterpriseValidFast();
+    if (new_license != license) {
+      license = new_license;
+      client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
+    }
+#ifdef MG_ENTERPRISE
+    SystemRestore<true>(client, system, dbms_handler, auth);
+#endif
+    // Check if any database has been left behind
+    dbms_handler.ForEach([&name = client.name_, reconnect](dbms::DatabaseAccess db_acc) {
+      // Specific database <-> replica client
+      db_acc->storage()->repl_storage_state_.WithClient(name, [&](storage::ReplicationStorageClient *client) {
+        if (reconnect || client->State() == storage::replication::ReplicaState::MAYBE_BEHIND) {
+          // Database <-> replica might be behind, check and recover
+          client->TryCheckReplicaStateAsync(db_acc->storage(), db_acc);
+        }
+      });
+    });
+  });
+}
+
+#ifdef MG_ENTERPRISE
+ReplicationHandler::ReplicationHandler(memgraph::replication::ReplicationState &repl_state,
+                                       memgraph::dbms::DbmsHandler &dbms_handler, memgraph::system::System *system,
+                                       memgraph::auth::SynchedAuth &auth)
+    : repl_state_{repl_state}, dbms_handler_{dbms_handler}, system_{system}, auth_{auth} {
+  RecoverReplication(repl_state_, system_, dbms_handler_, auth_);
+}
+#else
+ReplicationHandler::ReplicationHandler(replication::ReplicationState &repl_state, dbms::DbmsHandler &dbms_handler)
+    : repl_state_{repl_state}, dbms_handler_{dbms_handler} {
+  RecoverReplication(repl_state_, dbms_handler_);
+}
+#endif
+
+bool ReplicationHandler::SetReplicationRoleMain() {
+  auto const main_handler = [](memgraph::replication::RoleMainData &) {
+    // If we are already MAIN, we don't want to change anything
+    return false;
+  };
+
+  auto const replica_handler = [this](memgraph::replication::RoleReplicaData const &) {
+    return DoReplicaToMainPromotion();
+  };
+
+  // TODO: under lock
+  return std::visit(memgraph::utils::Overloaded{main_handler, replica_handler}, repl_state_.ReplicationData());
+}
+
+bool ReplicationHandler::SetReplicationRoleReplica(const memgraph::replication::ReplicationServerConfig &config) {
+  // We don't want to restart the server if we're already a REPLICA
+  if (repl_state_.IsReplica()) {
+    return false;
+  }
+
+  // TODO StorageState needs to be synched. Could have a dangling reference if someone adds a database as we are
+  //      deleting the replica.
+  // Remove database specific clients
+  dbms_handler_.ForEach([&](memgraph::dbms::DatabaseAccess db_acc) {
+    auto *storage = db_acc->storage();
+    storage->repl_storage_state_.replication_clients_.WithLock([](auto &clients) { clients.clear(); });
+  });
+  // Remove instance level clients
+  std::get<memgraph::replication::RoleMainData>(repl_state_.ReplicationData()).registered_replicas_.clear();
+
+  // Creates the server
+  repl_state_.SetReplicationRoleReplica(config);
+
+  // Start
+  const auto success =
+      std::visit(memgraph::utils::Overloaded{[](memgraph::replication::RoleMainData const &) {
+                                               // ASSERT
+                                               return false;
+                                             },
+                                             [this](memgraph::replication::RoleReplicaData const &data) {
+#ifdef MG_ENTERPRISE
+                                               return StartRpcServer(dbms_handler_, data, auth_);
+#else
+                                               return StartRpcServer(dbms_handler_, data);
+#endif
+                                             }},
+                 repl_state_.ReplicationData());
+  // TODO Handle error (restore to main?)
+  return success;
+}
+
+bool ReplicationHandler::DoReplicaToMainPromotion() {
+  // STEP 1) bring down all REPLICA servers
+  dbms_handler_.ForEach([](dbms::DatabaseAccess db_acc) {
+    auto *storage = db_acc->storage();
+    // Remember old epoch + storage timestamp association
+    storage->PrepareForNewEpoch();
+  });
+
+  // STEP 2) Change to MAIN
+  // TODO: restore replication servers if false?
+  if (!repl_state_.SetReplicationRoleMain()) {
+    // TODO: Handle recovery on failure???
+    return false;
+  }
+
+  // STEP 3) We are now MAIN, update storage local epoch
+  const auto &epoch = std::get<replication::RoleMainData>(std::as_const(repl_state_).ReplicationData()).epoch_;
+  dbms_handler_.ForEach([&](dbms::DatabaseAccess db_acc) {
+    auto *storage = db_acc->storage();
+    storage->repl_storage_state_.epoch_ = epoch;
+  });
+
+  return true;
+};
+
+// as MAIN, define and connect to REPLICAs
+auto ReplicationHandler::TryRegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
+    -> memgraph::utils::BasicResult<memgraph::query::RegisterReplicaError> {
+  return RegisterReplica_<false>(config);
+}
+
+auto ReplicationHandler::RegisterReplica(const memgraph::replication::ReplicationClientConfig &config)
+    -> memgraph::utils::BasicResult<memgraph::query::RegisterReplicaError> {
+  return RegisterReplica_<true>(config);
+}
+
+auto ReplicationHandler::UnregisterReplica(std::string_view name) -> memgraph::query::UnregisterReplicaResult {
+  auto const replica_handler =
+      [](memgraph::replication::RoleReplicaData const &) -> memgraph::query::UnregisterReplicaResult {
+    return memgraph::query::UnregisterReplicaResult::NOT_MAIN;
+  };
+  auto const main_handler =
+      [this, name](memgraph::replication::RoleMainData &mainData) -> memgraph::query::UnregisterReplicaResult {
+    if (!repl_state_.TryPersistUnregisterReplica(name)) {
+      return memgraph::query::UnregisterReplicaResult::COULD_NOT_BE_PERSISTED;
+    }
+    // Remove database specific clients
+    dbms_handler_.ForEach([name](memgraph::dbms::DatabaseAccess db_acc) {
+      db_acc->storage()->repl_storage_state_.replication_clients_.WithLock([&name](auto &clients) {
+        std::erase_if(clients, [name](const auto &client) { return client->Name() == name; });
+      });
+    });
+    // Remove instance level clients
+    auto const n_unregistered =
+        std::erase_if(mainData.registered_replicas_, [name](auto const &client) { return client.name_ == name; });
+    return n_unregistered != 0 ? memgraph::query::UnregisterReplicaResult::SUCCESS
+                               : memgraph::query::UnregisterReplicaResult::CAN_NOT_UNREGISTER;
+  };
+
+  return std::visit(memgraph::utils::Overloaded{main_handler, replica_handler}, repl_state_.ReplicationData());
+}
+
+auto ReplicationHandler::GetRole() const -> memgraph::replication_coordination_glue::ReplicationRole {
+  return repl_state_.GetRole();
+}
+
+bool ReplicationHandler::IsMain() const { return repl_state_.IsMain(); }
+
+bool ReplicationHandler::IsReplica() const { return repl_state_.IsReplica(); }
+
+}  // namespace memgraph::replication

src/replication_handler/system_replication.cpp

@@ -0,0 +1,115 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "replication_handler/system_replication.hpp"
+
+#include <spdlog/spdlog.h>
+
+#include "auth/replication_handlers.hpp"
+#include "dbms/replication_handlers.hpp"
+#include "license/license.hpp"
+#include "replication_handler/system_rpc.hpp"
+
+namespace memgraph::replication {
+
+#ifdef MG_ENTERPRISE
+void SystemHeartbeatHandler(const uint64_t ts, slk::Reader *req_reader, slk::Builder *res_builder) {
+  replication::SystemHeartbeatRes res{0};
+
+  // Ignore if no license
+  if (!license::global_license_checker.IsEnterpriseValidFast()) {
+    spdlog::error("Handling SystemHeartbeat, an enterprise RPC message, without license.");
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  replication::SystemHeartbeatReq req;
+  replication::SystemHeartbeatReq::Load(&req, req_reader);
+
+  res = replication::SystemHeartbeatRes{ts};
+  memgraph::slk::Save(res, res_builder);
+}
+
+void SystemRecoveryHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                           dbms::DbmsHandler &dbms_handler, auth::SynchedAuth &auth, slk::Reader *req_reader,
+                           slk::Builder *res_builder) {
+  using memgraph::replication::SystemRecoveryRes;
+  SystemRecoveryRes res(SystemRecoveryRes::Result::FAILURE);
+
+  utils::OnScopeExit send_on_exit([&]() { memgraph::slk::Save(res, res_builder); });
+
+  memgraph::replication::SystemRecoveryReq req;
+  memgraph::slk::Load(&req, req_reader);
+
+  /*
+   * DBMS
+   */
+  if (!dbms::SystemRecoveryHandler(dbms_handler, req.database_configs)) return;  // Failure sent on exit
+
+  /*
+   * AUTH
+   */
+  if (!auth::SystemRecoveryHandler(auth, req.auth_config, req.users, req.roles)) return;  // Failure sent on exit
+
+  /*
+   * SUCCESSFUL RECOVERY
+   */
+  system_state_access.SetLastCommitedTS(req.forced_group_timestamp);
+  spdlog::debug("SystemRecoveryHandler: SUCCESS updated LCTS to {}", req.forced_group_timestamp);
+  res = SystemRecoveryRes(SystemRecoveryRes::Result::SUCCESS);
+}
+
+void Register(replication::RoleReplicaData const &data, dbms::DbmsHandler &dbms_handler, auth::SynchedAuth &auth) {
+  // NOTE: Register even without license as the user could add a license at run-time
+  // TODO: fix Register when system is removed from DbmsHandler
+
+  auto system_state_access = dbms_handler.system_->CreateSystemStateAccess();
+
+  // System
+  data.server->rpc_server_.Register<replication::SystemHeartbeatRpc>(
+      [system_state_access](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received SystemHeartbeatRpc");
+        SystemHeartbeatHandler(system_state_access.LastCommitedTS(), req_reader, res_builder);
+      });
+  data.server->rpc_server_.Register<replication::SystemRecoveryRpc>(
+      [system_state_access, &dbms_handler, &auth](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received SystemRecoveryRpc");
+        SystemRecoveryHandler(system_state_access, dbms_handler, auth, req_reader, res_builder);
+      });
+
+  // DBMS
+  dbms::Register(data, system_state_access, dbms_handler);
+
+  // Auth
+  auth::Register(data, system_state_access, auth);
+}
+#endif
+
+#ifdef MG_ENTERPRISE
+bool StartRpcServer(dbms::DbmsHandler &dbms_handler, const replication::RoleReplicaData &data,
+                    auth::SynchedAuth &auth) {
+#else
+bool StartRpcServer(dbms::DbmsHandler &dbms_handler, const replication::RoleReplicaData &data) {
+#endif
+  // Register storage handlers
+  dbms::InMemoryReplicationHandlers::Register(&dbms_handler, *data.server);
+#ifdef MG_ENTERPRISE
+  // Register system handlers
+  Register(data, dbms_handler, auth);
+#endif
+  // Start server
+  if (!data.server->Start()) {
+    spdlog::error("Unable to start the replication server.");
+    return false;
+  }
+  return true;
+}
+}  // namespace memgraph::replication

src/replication_handler/system_rpc.cpp

@@ -0,0 +1,111 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "replication_handler/system_rpc.hpp"
+
+#include <json/json.hpp>
+
+#include "auth/rpc.hpp"
+#include "slk/serialization.hpp"
+#include "slk/streams.hpp"
+#include "storage/v2/replication/rpc.hpp"
+#include "utils/enum.hpp"
+
+namespace memgraph::slk {
+// Serialize code for SystemHeartbeatRes
+void Save(const memgraph::replication::SystemHeartbeatRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.system_timestamp, builder);
+}
+void Load(memgraph::replication::SystemHeartbeatRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->system_timestamp, reader);
+}
+
+// Serialize code for SystemHeartbeatReq
+void Save(const memgraph::replication::SystemHeartbeatReq & /*self*/, memgraph::slk::Builder * /*builder*/) {
+  /* Nothing to serialize */
+}
+void Load(memgraph::replication::SystemHeartbeatReq * /*self*/, memgraph::slk::Reader * /*reader*/) {
+  /* Nothing to serialize */
+}
+
+// Serialize code for SystemRecoveryReq
+void Save(const memgraph::replication::SystemRecoveryReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.forced_group_timestamp, builder);
+  memgraph::slk::Save(self.database_configs, builder);
+  memgraph::slk::Save(self.auth_config, builder);
+  memgraph::slk::Save(self.users, builder);
+  memgraph::slk::Save(self.roles, builder);
+}
+
+void Load(memgraph::replication::SystemRecoveryReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->forced_group_timestamp, reader);
+  memgraph::slk::Load(&self->database_configs, reader);
+  memgraph::slk::Load(&self->auth_config, reader);
+  memgraph::slk::Load(&self->users, reader);
+  memgraph::slk::Load(&self->roles, reader);
+}
+
+// Serialize code for SystemRecoveryRes
+void Save(const memgraph::replication::SystemRecoveryRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(utils::EnumToNum<uint8_t>(self.result), builder);
+}
+
+void Load(memgraph::replication::SystemRecoveryRes *self, memgraph::slk::Reader *reader) {
+  uint8_t res = 0;
+  memgraph::slk::Load(&res, reader);
+  if (!utils::NumToEnum(res, self->result)) {
+    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
+  }
+}
+
+}  // namespace memgraph::slk
+
+namespace memgraph::replication {
+
+constexpr utils::TypeInfo SystemHeartbeatReq::kType{utils::TypeId::REP_SYSTEM_HEARTBEAT_REQ, "SystemHeartbeatReq",
+                                                    nullptr};
+
+constexpr utils::TypeInfo SystemHeartbeatRes::kType{utils::TypeId::REP_SYSTEM_HEARTBEAT_RES, "SystemHeartbeatRes",
+                                                    nullptr};
+
+constexpr utils::TypeInfo SystemRecoveryReq::kType{utils::TypeId::REP_SYSTEM_RECOVERY_REQ, "SystemRecoveryReq",
+                                                   nullptr};
+
+constexpr utils::TypeInfo SystemRecoveryRes::kType{utils::TypeId::REP_SYSTEM_RECOVERY_RES, "SystemRecoveryRes",
+                                                   nullptr};
+
+void SystemHeartbeatReq::Save(const SystemHeartbeatReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void SystemHeartbeatReq::Load(SystemHeartbeatReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+void SystemHeartbeatRes::Save(const SystemHeartbeatRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void SystemHeartbeatRes::Load(SystemHeartbeatRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+
+void SystemRecoveryReq::Save(const SystemRecoveryReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void SystemRecoveryReq::Load(SystemRecoveryReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+void SystemRecoveryRes::Save(const SystemRecoveryRes &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self, builder);
+}
+void SystemRecoveryRes::Load(SystemRecoveryRes *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(self, reader);
+}
+
+}  // namespace memgraph::replication

src/storage/v2/config.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -15,6 +15,7 @@
 #include <cstdint>
 #include <filesystem>
 
+#include "flags/replication.hpp"
 #include "storage/v2/isolation_level.hpp"
 #include "storage/v2/storage_mode.hpp"
 #include "utils/exceptions.hpp"
@@ -128,10 +129,15 @@ struct Config {
 };
 
 inline auto ReplicationStateRootPath(memgraph::storage::Config const &config) -> std::optional<std::filesystem::path> {
-  if (!config.durability.restore_replication_state_on_startup) {
+  if (!config.durability.restore_replication_state_on_startup
+#ifdef MG_ENTERPRISE
+      && !FLAGS_coordinator_server_port
+#endif
+  ) {
     spdlog::warn(
         "Replication configuration will NOT be stored. When the server restarts, replication state will be "
         "forgotten.");
+
     return std::nullopt;
   }
   return {config.durability.storage_directory};

src/storage/v2/inmemory/storage.cpp

@@ -176,9 +176,9 @@ InMemoryStorage::~InMemoryStorage() {
   committed_transactions_.WithLock([](auto &transactions) { transactions.clear(); });
 }
 
-InMemoryStorage::InMemoryAccessor::InMemoryAccessor(auto tag, InMemoryStorage *storage, IsolationLevel isolation_level,
-                                                    StorageMode storage_mode,
-                                                    memgraph::replication_coordination_glue::ReplicationRole replication_role)
+InMemoryStorage::InMemoryAccessor::InMemoryAccessor(
+    auto tag, InMemoryStorage *storage, IsolationLevel isolation_level, StorageMode storage_mode,
+    memgraph::replication_coordination_glue::ReplicationRole replication_role)
     : Accessor(tag, storage, isolation_level, storage_mode, replication_role),
       config_(storage->config_.salient.items) {}
 InMemoryStorage::InMemoryAccessor::InMemoryAccessor(InMemoryAccessor &&other) noexcept
@@ -882,6 +882,9 @@ void InMemoryStorage::InMemoryAccessor::Abort() {
   std::list<Gid> my_deleted_vertices;
   std::list<Gid> my_deleted_edges;
 
+  std::unordered_map<Gid, uint64_t> added_in_edges;
+  std::unordered_map<Gid, uint64_t> added_out_edges;
+
   std::map<LabelId, std::vector<Vertex *>> label_cleanup;
   std::map<LabelId, std::vector<std::pair<PropertyValue, Vertex *>>> label_property_cleanup;
   std::map<PropertyId, std::vector<std::pair<PropertyValue, Vertex *>>> property_cleanup;
@@ -904,6 +907,10 @@ void InMemoryStorage::InMemoryAccessor::Abort() {
         auto *vertex = prev.vertex;
         auto guard = std::unique_lock{vertex->lock};
         Delta *current = vertex->delta;
+
+        added_in_edges.clear();
+        added_out_edges.clear();
+
         while (current != nullptr &&
                current->timestamp->load(std::memory_order_acquire) == transaction_.transaction_id) {
           switch (current->action) {
@@ -955,19 +962,43 @@ void InMemoryStorage::InMemoryAccessor::Abort() {
               break;
             }
             case Delta::Action::ADD_IN_EDGE: {
+              if (added_in_edges.empty()) {
+                uint64_t idx = 0;
+                for (auto &[type, opposing_vertex, ref] : vertex->in_edges) {
+                  auto gid = config_.properties_on_edges ? ref.ptr->gid : ref.gid;
+                  added_in_edges.emplace(gid, idx);
+                  idx++;
+                }
+              }
+
+              auto wanted_gid =
+                  config_.properties_on_edges ? current->vertex_edge.edge.ptr->gid : current->vertex_edge.edge.gid;
+              MG_ASSERT(!added_in_edges.contains(wanted_gid), "Invalid database state!");
+
               std::tuple<EdgeTypeId, Vertex *, EdgeRef> link{current->vertex_edge.edge_type,
                                                              current->vertex_edge.vertex, current->vertex_edge.edge};
-              auto it = std::find(vertex->in_edges.begin(), vertex->in_edges.end(), link);
-              MG_ASSERT(it == vertex->in_edges.end(), "Invalid database state!");
               vertex->in_edges.push_back(link);
+              added_in_edges.emplace(wanted_gid, vertex->in_edges.size() - 1);
               break;
             }
             case Delta::Action::ADD_OUT_EDGE: {
+              if (added_out_edges.empty()) {
+                uint64_t idx = 0;
+                for (auto &[type, opposing_vertex, ref] : vertex->out_edges) {
+                  auto gid = config_.properties_on_edges ? ref.ptr->gid : ref.gid;
+                  added_out_edges.emplace(gid, idx);
+                  idx++;
+                }
+              }
+
+              auto wanted_gid =
+                  config_.properties_on_edges ? current->vertex_edge.edge.ptr->gid : current->vertex_edge.edge.gid;
+              MG_ASSERT(!added_out_edges.contains(wanted_gid), "Invalid database state!");
+
               std::tuple<EdgeTypeId, Vertex *, EdgeRef> link{current->vertex_edge.edge_type,
                                                              current->vertex_edge.vertex, current->vertex_edge.edge};
-              auto it = std::find(vertex->out_edges.begin(), vertex->out_edges.end(), link);
-              MG_ASSERT(it == vertex->out_edges.end(), "Invalid database state!");
               vertex->out_edges.push_back(link);
+              added_out_edges.emplace(wanted_gid, vertex->out_edges.size() - 1);
               // Increment edge count. We only increment the count here because
               // the information in `ADD_IN_EDGE` and `Edge/RECREATE_OBJECT` is
               // redundant. Also, `Edge/RECREATE_OBJECT` isn't available when
@@ -976,21 +1007,55 @@ void InMemoryStorage::InMemoryAccessor::Abort() {
               break;
             }
             case Delta::Action::REMOVE_IN_EDGE: {
-              std::tuple<EdgeTypeId, Vertex *, EdgeRef> link{current->vertex_edge.edge_type,
-                                                             current->vertex_edge.vertex, current->vertex_edge.edge};
-              auto it = std::find(vertex->in_edges.begin(), vertex->in_edges.end(), link);
-              MG_ASSERT(it != vertex->in_edges.end(), "Invalid database state!");
-              std::swap(*it, *vertex->in_edges.rbegin());
+              if (added_in_edges.empty()) {
+                uint64_t idx = 0;
+                for (auto &[type, opposing_vertex, ref] : vertex->in_edges) {
+                  auto gid = config_.properties_on_edges ? ref.ptr->gid : ref.gid;
+                  added_in_edges.emplace(gid, idx);
+                  idx++;
+                }
+              }
+
+              auto wanted_gid =
+                  config_.properties_on_edges ? current->vertex_edge.edge.ptr->gid : current->vertex_edge.edge.gid;
+
+              MG_ASSERT(added_in_edges.count(wanted_gid), "Invalid database state!");
+
+              auto pop_index = added_in_edges[wanted_gid];
+
+              std::swap(vertex->in_edges[pop_index], *vertex->in_edges.rbegin());
               vertex->in_edges.pop_back();
+              added_in_edges.erase(wanted_gid);
+
+              auto switched_vertex = std::get<2>(vertex->in_edges[pop_index]);
+              auto switch_gid = config_.properties_on_edges ? switched_vertex.ptr->gid : switched_vertex.gid;
+              added_in_edges[switch_gid] = pop_index;
               break;
             }
             case Delta::Action::REMOVE_OUT_EDGE: {
-              std::tuple<EdgeTypeId, Vertex *, EdgeRef> link{current->vertex_edge.edge_type,
-                                                             current->vertex_edge.vertex, current->vertex_edge.edge};
-              auto it = std::find(vertex->out_edges.begin(), vertex->out_edges.end(), link);
-              MG_ASSERT(it != vertex->out_edges.end(), "Invalid database state!");
-              std::swap(*it, *vertex->out_edges.rbegin());
+              if (added_out_edges.empty()) {
+                uint64_t idx = 0;
+                for (auto &[type, opposing_vertex, ref] : vertex->out_edges) {
+                  auto gid = config_.properties_on_edges ? ref.ptr->gid : ref.gid;
+                  added_out_edges.emplace(gid, idx);
+                  idx++;
+                }
+              }
+
+              auto wanted_gid =
+                  config_.properties_on_edges ? current->vertex_edge.edge.ptr->gid : current->vertex_edge.edge.gid;
+
+              MG_ASSERT(added_out_edges.count(wanted_gid), "Invalid database state!");
+
+              auto pop_index = added_out_edges[wanted_gid];
+
+              std::swap(vertex->out_edges[pop_index], *vertex->out_edges.rbegin());
               vertex->out_edges.pop_back();
+              added_out_edges.erase(wanted_gid);
+
+              auto switched_vertex = std::get<2>(vertex->out_edges[pop_index]);
+              auto switch_gid = config_.properties_on_edges ? switched_vertex.ptr->gid : switched_vertex.gid;
+              added_out_edges[switch_gid] = pop_index;
               // Decrement edge count. We only decrement the count here because
               // the information in `REMOVE_IN_EDGE` and `Edge/DELETE_OBJECT` is
               // redundant. Also, `Edge/DELETE_OBJECT` isn't available when edge

src/storage/v2/replication/replication_client.cpp

@@ -35,6 +35,7 @@ void ReplicationStorageClient::UpdateReplicaState(Storage *storage, DatabaseAcce
 
   auto hb_stream{client_.rpc_client_.Stream<replication::HeartbeatRpc>(
       storage->uuid(), replStorageState.last_commit_timestamp_, std::string{replStorageState.epoch_.id()})};
+
   const auto replica = hb_stream.AwaitResponse();
 
 #ifdef MG_ENTERPRISE       // Multi-tenancy is only supported in enterprise
@@ -67,7 +68,6 @@ void ReplicationStorageClient::UpdateReplicaState(Storage *storage, DatabaseAcce
         "now hold unique data. Please resolve data conflicts and start the "
         "replication on a clean instance.",
         client_.name_, client_.name_, client_.name_);
-    // TODO: (andi) Talk about renaming MAYBE_BEHIND to branching
     // State not updated, hence in MAYBE_BEHIND state
     return;
   }

src/storage/v2/replication/rpc.cpp

@@ -59,39 +59,6 @@ void TimestampRes::Save(const TimestampRes &self, memgraph::slk::Builder *builde
   memgraph::slk::Save(self, builder);
 }
 void TimestampRes::Load(TimestampRes *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
-void CreateDatabaseReq::Save(const CreateDatabaseReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void CreateDatabaseReq::Load(CreateDatabaseReq *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(self, reader);
-}
-void CreateDatabaseRes::Save(const CreateDatabaseRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void CreateDatabaseRes::Load(CreateDatabaseRes *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(self, reader);
-}
-void DropDatabaseReq::Save(const DropDatabaseReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void DropDatabaseReq::Load(DropDatabaseReq *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
-void DropDatabaseRes::Save(const DropDatabaseRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void DropDatabaseRes::Load(DropDatabaseRes *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
-void SystemRecoveryReq::Save(const SystemRecoveryReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void SystemRecoveryReq::Load(SystemRecoveryReq *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(self, reader);
-}
-void SystemRecoveryRes::Save(const SystemRecoveryRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self, builder);
-}
-void SystemRecoveryRes::Load(SystemRecoveryRes *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(self, reader);
-}
-
 }  // namespace storage::replication
 
 constexpr utils::TypeInfo storage::replication::AppendDeltasReq::kType{utils::TypeId::REP_APPEND_DELTAS_REQ,
@@ -130,24 +97,6 @@ constexpr utils::TypeInfo storage::replication::TimestampReq::kType{utils::TypeI
 constexpr utils::TypeInfo storage::replication::TimestampRes::kType{utils::TypeId::REP_TIMESTAMP_RES, "TimestampRes",
                                                                     nullptr};
 
-constexpr utils::TypeInfo storage::replication::CreateDatabaseReq::kType{utils::TypeId::REP_CREATE_DATABASE_REQ,
-                                                                         "CreateDatabaseReq", nullptr};
-
-constexpr utils::TypeInfo storage::replication::CreateDatabaseRes::kType{utils::TypeId::REP_CREATE_DATABASE_RES,
-                                                                         "CreateDatabaseRes", nullptr};
-
-constexpr utils::TypeInfo storage::replication::DropDatabaseReq::kType{utils::TypeId::REP_DROP_DATABASE_REQ,
-                                                                       "DropDatabaseReq", nullptr};
-
-constexpr utils::TypeInfo storage::replication::DropDatabaseRes::kType{utils::TypeId::REP_DROP_DATABASE_RES,
-                                                                       "DropDatabaseRes", nullptr};
-
-constexpr utils::TypeInfo storage::replication::SystemRecoveryReq::kType{utils::TypeId::REP_SYSTEM_RECOVERY_REQ,
-                                                                         "SystemRecoveryReq", nullptr};
-
-constexpr utils::TypeInfo storage::replication::SystemRecoveryRes::kType{utils::TypeId::REP_SYSTEM_RECOVERY_RES,
-                                                                         "SystemRecoveryRes", nullptr};
-
 // Autogenerated SLK serialization code
 namespace slk {
 // Serialize code for TimestampRes
@@ -316,91 +265,5 @@ void Load(memgraph::storage::SalientConfig *self, memgraph::slk::Reader *reader)
   memgraph::slk::Load(&self->items.enable_schema_metadata, reader);
 }
 
-// Serialize code for CreateDatabaseReq
-
-void Save(const memgraph::storage::replication::CreateDatabaseReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self.epoch_id, builder);
-  memgraph::slk::Save(self.expected_group_timestamp, builder);
-  memgraph::slk::Save(self.new_group_timestamp, builder);
-  memgraph::slk::Save(self.config, builder);
-}
-
-void Load(memgraph::storage::replication::CreateDatabaseReq *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(&self->epoch_id, reader);
-  memgraph::slk::Load(&self->expected_group_timestamp, reader);
-  memgraph::slk::Load(&self->new_group_timestamp, reader);
-  memgraph::slk::Load(&self->config, reader);
-}
-
-// Serialize code for CreateDatabaseRes
-
-void Save(const memgraph::storage::replication::CreateDatabaseRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(utils::EnumToNum<uint8_t>(self.result), builder);
-}
-
-void Load(memgraph::storage::replication::CreateDatabaseRes *self, memgraph::slk::Reader *reader) {
-  uint8_t res = 0;
-  memgraph::slk::Load(&res, reader);
-  if (!utils::NumToEnum(res, self->result)) {
-    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
-  }
-}
-
-// Serialize code for DropDatabaseReq
-
-void Save(const memgraph::storage::replication::DropDatabaseReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self.epoch_id, builder);
-  memgraph::slk::Save(self.expected_group_timestamp, builder);
-  memgraph::slk::Save(self.new_group_timestamp, builder);
-  memgraph::slk::Save(self.uuid, builder);
-}
-
-void Load(memgraph::storage::replication::DropDatabaseReq *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(&self->epoch_id, reader);
-  memgraph::slk::Load(&self->expected_group_timestamp, reader);
-  memgraph::slk::Load(&self->new_group_timestamp, reader);
-  memgraph::slk::Load(&self->uuid, reader);
-}
-
-// Serialize code for DropDatabaseRes
-
-void Save(const memgraph::storage::replication::DropDatabaseRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(utils::EnumToNum<uint8_t>(self.result), builder);
-}
-
-void Load(memgraph::storage::replication::DropDatabaseRes *self, memgraph::slk::Reader *reader) {
-  uint8_t res = 0;
-  memgraph::slk::Load(&res, reader);
-  if (!utils::NumToEnum(res, self->result)) {
-    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
-  }
-}
-
-// Serialize code for SystemRecoveryReq
-
-void Save(const memgraph::storage::replication::SystemRecoveryReq &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(self.forced_group_timestamp, builder);
-  memgraph::slk::Save(self.database_configs, builder);
-}
-
-void Load(memgraph::storage::replication::SystemRecoveryReq *self, memgraph::slk::Reader *reader) {
-  memgraph::slk::Load(&self->forced_group_timestamp, reader);
-  memgraph::slk::Load(&self->database_configs, reader);
-}
-
-// Serialize code for SystemRecoveryRes
-
-void Save(const memgraph::storage::replication::SystemRecoveryRes &self, memgraph::slk::Builder *builder) {
-  memgraph::slk::Save(utils::EnumToNum<uint8_t>(self.result), builder);
-}
-
-void Load(memgraph::storage::replication::SystemRecoveryRes *self, memgraph::slk::Reader *reader) {
-  uint8_t res = 0;
-  memgraph::slk::Load(&res, reader);
-  if (!utils::NumToEnum(res, self->result)) {
-    throw SlkReaderException("Unexpected result line:{}!", __LINE__);
-  }
-}
-
 }  // namespace slk
 }  // namespace memgraph

src/storage/v2/replication/rpc.hpp

@@ -201,108 +201,6 @@ struct TimestampRes {
 
 using TimestampRpc = rpc::RequestResponse<TimestampReq, TimestampRes>;
 
-struct CreateDatabaseReq {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  static void Load(CreateDatabaseReq *self, memgraph::slk::Reader *reader);
-  static void Save(const CreateDatabaseReq &self, memgraph::slk::Builder *builder);
-  CreateDatabaseReq() = default;
-  CreateDatabaseReq(std::string epoch_id, uint64_t expected_group_timestamp, uint64_t new_group_timestamp,
-                    storage::SalientConfig config)
-      : epoch_id(std::move(epoch_id)),
-        expected_group_timestamp{expected_group_timestamp},
-        new_group_timestamp(new_group_timestamp),
-        config(std::move(config)) {}
-
-  std::string epoch_id;
-  uint64_t expected_group_timestamp;
-  uint64_t new_group_timestamp;
-  storage::SalientConfig config;
-};
-
-struct CreateDatabaseRes {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
-
-  static void Load(CreateDatabaseRes *self, memgraph::slk::Reader *reader);
-  static void Save(const CreateDatabaseRes &self, memgraph::slk::Builder *builder);
-  CreateDatabaseRes() = default;
-  explicit CreateDatabaseRes(Result res) : result(res) {}
-
-  Result result;
-};
-
-using CreateDatabaseRpc = rpc::RequestResponse<CreateDatabaseReq, CreateDatabaseRes>;
-
-struct DropDatabaseReq {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  static void Load(DropDatabaseReq *self, memgraph::slk::Reader *reader);
-  static void Save(const DropDatabaseReq &self, memgraph::slk::Builder *builder);
-  DropDatabaseReq() = default;
-  DropDatabaseReq(std::string epoch_id, uint64_t expected_group_timestamp, uint64_t new_group_timestamp,
-                  const utils::UUID &uuid)
-      : epoch_id(std::move(epoch_id)),
-        expected_group_timestamp{expected_group_timestamp},
-        new_group_timestamp(new_group_timestamp),
-        uuid(uuid) {}
-
-  std::string epoch_id;
-  uint64_t expected_group_timestamp;
-  uint64_t new_group_timestamp;
-  utils::UUID uuid;
-};
-
-struct DropDatabaseRes {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
-
-  static void Load(DropDatabaseRes *self, memgraph::slk::Reader *reader);
-  static void Save(const DropDatabaseRes &self, memgraph::slk::Builder *builder);
-  DropDatabaseRes() = default;
-  explicit DropDatabaseRes(Result res) : result(res) {}
-
-  Result result;
-};
-
-using DropDatabaseRpc = rpc::RequestResponse<DropDatabaseReq, DropDatabaseRes>;
-
-struct SystemRecoveryReq {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  static void Load(SystemRecoveryReq *self, memgraph::slk::Reader *reader);
-  static void Save(const SystemRecoveryReq &self, memgraph::slk::Builder *builder);
-  SystemRecoveryReq() = default;
-  SystemRecoveryReq(uint64_t forced_group_timestamp, std::vector<storage::SalientConfig> database_configs)
-      : forced_group_timestamp{forced_group_timestamp}, database_configs(std::move(database_configs)) {}
-
-  uint64_t forced_group_timestamp;
-  std::vector<storage::SalientConfig> database_configs;
-};
-
-struct SystemRecoveryRes {
-  static const utils::TypeInfo kType;
-  static const utils::TypeInfo &GetTypeInfo() { return kType; }
-
-  enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
-
-  static void Load(SystemRecoveryRes *self, memgraph::slk::Reader *reader);
-  static void Save(const SystemRecoveryRes &self, memgraph::slk::Builder *builder);
-  SystemRecoveryRes() = default;
-  explicit SystemRecoveryRes(Result res) : result(res) {}
-
-  Result result;
-};
-
-using SystemRecoveryRpc = rpc::RequestResponse<SystemRecoveryReq, SystemRecoveryRes>;
-
 }  // namespace memgraph::storage::replication
 
 // SLK serialization declarations
@@ -356,28 +254,8 @@ void Save(const memgraph::storage::replication::AppendDeltasReq &self, memgraph:
 
 void Load(memgraph::storage::replication::AppendDeltasReq *self, memgraph::slk::Reader *reader);
 
-void Save(const memgraph::storage::replication::CreateDatabaseReq &self, memgraph::slk::Builder *builder);
+void Save(const memgraph::storage::SalientConfig &self, memgraph::slk::Builder *builder);
 
-void Load(memgraph::storage::replication::CreateDatabaseReq *self, memgraph::slk::Reader *reader);
-
-void Save(const memgraph::storage::replication::CreateDatabaseRes &self, memgraph::slk::Builder *builder);
-
-void Load(memgraph::storage::replication::CreateDatabaseRes *self, memgraph::slk::Reader *reader);
-
-void Save(const memgraph::storage::replication::DropDatabaseReq &self, memgraph::slk::Builder *builder);
-
-void Load(memgraph::storage::replication::DropDatabaseReq *self, memgraph::slk::Reader *reader);
-
-void Save(const memgraph::storage::replication::DropDatabaseRes &self, memgraph::slk::Builder *builder);
-
-void Load(memgraph::storage::replication::DropDatabaseRes *self, memgraph::slk::Reader *reader);
-
-void Save(const memgraph::storage::replication::SystemRecoveryReq &self, memgraph::slk::Builder *builder);
-
-void Load(memgraph::storage::replication::SystemRecoveryReq *self, memgraph::slk::Reader *reader);
-
-void Save(const memgraph::storage::replication::SystemRecoveryRes &self, memgraph::slk::Builder *builder);
-
-void Load(memgraph::storage::replication::SystemRecoveryRes *self, memgraph::slk::Reader *reader);
+void Load(memgraph::storage::SalientConfig *self, memgraph::slk::Reader *reader);
 
 }  // namespace memgraph::slk

src/system/CMakeLists.txt

@@ -0,0 +1,23 @@
+add_library(mg-system STATIC)
+add_library(mg::system ALIAS mg-system)
+target_sources(mg-system
+        PUBLIC
+        include/system/action.hpp
+        include/system/system.hpp
+        include/system/transaction.hpp
+        include/system/state.hpp
+
+        PRIVATE
+        action.cpp
+        system.cpp
+        transaction.cpp
+        state.cpp
+
+)
+target_include_directories(mg-system PUBLIC include)
+
+target_link_libraries(mg-system
+        PUBLIC
+        mg::replication
+
+)

src/system/action.cpp

@@ -8,14 +8,4 @@
 // the Business Source License, use of this software will be governed
 // by the Apache License, Version 2.0, included in the file
 // licenses/APL.txt.
-
-#pragma once
-
-#ifdef MG_ENTERPRISE
-
-#include <cstdint>
-
-namespace memgraph::coordination {
-enum class DoFailoverStatus : uint8_t { SUCCESS, ALL_REPLICAS_DOWN, MAIN_ALIVE, RPC_FAILED };
-}  // namespace memgraph::coordination
-#endif
+#include "system/include/system/action.hpp"

src/system/include/system/action.hpp

@@ -0,0 +1,38 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include "replication/epoch.hpp"
+#include "replication/replication_client.hpp"
+#include "replication/state.hpp"
+
+namespace memgraph::system {
+
+struct Transaction;
+
+/// The system action interface that subsystems will implement. This OO-style separation is needed so that one common
+/// mechanism can be used for all subsystem replication within a system transaction, without the need for System to
+/// know about all the subsystems.
+struct ISystemAction {
+  /// Durability step which is defered until commit time
+  virtual void DoDurability() = 0;
+
+  /// Prepare the RPC payload that will be sent to all replicas clients
+  virtual bool DoReplication(memgraph::replication::ReplicationClient &client,
+                             memgraph::replication::ReplicationEpoch const &epoch,
+                             Transaction const &system_tx) const = 0;
+
+  virtual void PostReplication(memgraph::replication::RoleMainData &main_data) const = 0;
+
+  virtual ~ISystemAction() = default;
+};
+}  // namespace memgraph::system

src/system/include/system/state.hpp

@@ -0,0 +1,57 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include <atomic>
+#include <cstdint>
+
+#include "kvstore/kvstore.hpp"
+#include "utils/file.hpp"
+
+namespace memgraph::system {
+
+namespace {
+constexpr std::string_view kLastCommitedSystemTsKey = "last_committed_system_ts";  // Key for timestamp durability
+}
+
+struct State {
+  explicit State(std::optional<std::filesystem::path> storage, bool recovery_on_startup);
+
+  void FinalizeTransaction(std::uint64_t timestamp) {
+    if (durability_) {
+      durability_->Put(kLastCommitedSystemTsKey, std::to_string(timestamp));
+    }
+    last_committed_system_timestamp_.store(timestamp);
+  }
+
+  auto LastCommittedSystemTimestamp() -> uint64_t { return last_committed_system_timestamp_.load(); }
+
+ private:
+  friend struct ReplicaHandlerAccessToState;
+  friend struct Transaction;
+
+  std::optional<kvstore::KVStore> durability_;
+  std::atomic_uint64_t last_committed_system_timestamp_{};
+};
+
+struct ReplicaHandlerAccessToState {
+  explicit ReplicaHandlerAccessToState(memgraph::system::State &state) : state_{&state} {}
+
+  auto LastCommitedTS() const -> uint64_t { return state_->last_committed_system_timestamp_.load(); }
+
+  void SetLastCommitedTS(uint64_t new_timestamp) { state_->last_committed_system_timestamp_.store(new_timestamp); }
+
+ private:
+  State *state_;
+};
+
+}  // namespace memgraph::system

src/system/include/system/system.hpp

@@ -0,0 +1,52 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include "system/state.hpp"
+#include "system/transaction.hpp"
+
+namespace memgraph::system {
+
+struct TransactionGuard {
+  explicit TransactionGuard(std::unique_lock<std::timed_mutex> guard) : guard_(std::move(guard)) {}
+
+ private:
+  std::unique_lock<std::timed_mutex> guard_;
+};
+
+struct System {
+  // NOTE: default arguments to make testing easier.
+  System(std::optional<std::filesystem::path> storage = std::nullopt, bool recovery_on_startup = false)
+      : state_(std::move(storage), recovery_on_startup), timestamp_{state_.LastCommittedSystemTimestamp()} {}
+
+  auto TryCreateTransaction(std::chrono::microseconds try_time = std::chrono::milliseconds{100})
+      -> std::optional<Transaction> {
+    auto system_unique = std::unique_lock{mtx_, std::defer_lock};
+    if (!system_unique.try_lock_for(try_time)) {
+      return std::nullopt;
+    }
+    return Transaction{state_, std::move(system_unique), timestamp_++};
+  }
+
+  // TODO: this and LastCommittedSystemTimestamp maybe not needed
+  auto GenTransactionGuard() -> TransactionGuard { return TransactionGuard{std::unique_lock{mtx_}}; }
+  auto LastCommittedSystemTimestamp() -> uint64_t { return state_.LastCommittedSystemTimestamp(); }
+
+  auto CreateSystemStateAccess() -> ReplicaHandlerAccessToState { return ReplicaHandlerAccessToState{state_}; }
+
+ private:
+  State state_;
+  std::timed_mutex mtx_{};
+  std::uint64_t timestamp_{};
+};
+
+}  // namespace memgraph::system

src/system/include/system/transaction.hpp

@@ -0,0 +1,124 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#pragma once
+
+#include <chrono>
+#include <list>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include "replication/state.hpp"
+#include "system/action.hpp"
+#include "system/state.hpp"
+
+namespace memgraph::system {
+
+enum class AllSyncReplicaStatus : std::uint8_t {
+  AllCommitsConfirmed,
+  SomeCommitsUnconfirmed,
+};
+
+struct Transaction;
+
+template <typename T>
+concept ReplicationPolicy = requires(T handler, ISystemAction const &action, Transaction const &txn) {
+  { handler.ApplyAction(action, txn) } -> std::same_as<AllSyncReplicaStatus>;
+};
+
+struct System;
+
+struct Transaction {
+  template <std::derived_from<ISystemAction> TAction, typename... Args>
+  requires std::constructible_from<TAction, Args...>
+  void AddAction(Args &&...args) { actions_.emplace_back(std::make_unique<TAction>(std::forward<Args>(args)...)); }
+
+  template <ReplicationPolicy Handler>
+  auto Commit(Handler handler) -> AllSyncReplicaStatus {
+    if (!lock_.owns_lock() || actions_.empty()) {
+      // If no actions, we do not increment the last commited ts, since there is no delta to send to the REPLICA
+      Abort();
+      return AllSyncReplicaStatus::AllCommitsConfirmed;  // TODO: some kind of error
+    }
+
+    auto sync_status = AllSyncReplicaStatus::AllCommitsConfirmed;
+
+    while (!actions_.empty()) {
+      auto &action = actions_.front();
+
+      /// durability
+      action->DoDurability();
+
+      /// replication prep
+      auto action_sync_status = handler.ApplyAction(*action, *this);
+      if (action_sync_status != AllSyncReplicaStatus::AllCommitsConfirmed) {
+        sync_status = AllSyncReplicaStatus::SomeCommitsUnconfirmed;
+      }
+
+      actions_.pop_front();
+    }
+
+    state_->FinalizeTransaction(timestamp_);
+    lock_.unlock();
+
+    return sync_status;
+  }
+
+  void Abort() {
+    if (lock_.owns_lock()) {
+      lock_.unlock();
+    }
+    actions_.clear();
+  }
+
+  auto last_committed_system_timestamp() const -> uint64_t { return state_->last_committed_system_timestamp_.load(); }
+  auto timestamp() const -> uint64_t { return timestamp_; }
+
+ private:
+  friend struct System;
+  Transaction(State &state, std::unique_lock<std::timed_mutex> lock, std::uint64_t timestamp)
+      : state_{std::addressof(state)}, lock_(std::move(lock)), timestamp_{timestamp} {}
+
+  State *state_;
+  std::unique_lock<std::timed_mutex> lock_;
+  std::uint64_t timestamp_;
+  std::list<std::unique_ptr<ISystemAction>> actions_;
+};
+
+struct DoReplication {
+  explicit DoReplication(replication::RoleMainData &main_data) : main_data_{main_data} {}
+  auto ApplyAction(ISystemAction const &action, Transaction const &system_tx) -> AllSyncReplicaStatus {
+    auto sync_status = AllSyncReplicaStatus::AllCommitsConfirmed;
+
+    for (auto &client : main_data_.registered_replicas_) {
+      bool completed = action.DoReplication(client, main_data_.epoch_, system_tx);
+      if (!completed && client.mode_ == replication_coordination_glue::ReplicationMode::SYNC) {
+        sync_status = AllSyncReplicaStatus::SomeCommitsUnconfirmed;
+      }
+    }
+
+    action.PostReplication(main_data_);
+    return sync_status;
+  }
+
+ private:
+  replication::RoleMainData &main_data_;
+};
+static_assert(ReplicationPolicy<DoReplication>);
+
+struct DoNothing {
+  auto ApplyAction(ISystemAction const & /*action*/, Transaction const & /*system_tx*/) -> AllSyncReplicaStatus {
+    return AllSyncReplicaStatus::AllCommitsConfirmed;
+  }
+};
+static_assert(ReplicationPolicy<DoNothing>);
+
+}  // namespace memgraph::system

src/system/state.cpp

@@ -0,0 +1,57 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+#include "system/state.hpp"
+
+namespace memgraph::system {
+
+namespace {
+
+constexpr std::string_view kSystemDir = ".system";
+constexpr std::string_view kVersion = "version";  // Key for version durability
+constexpr std::string_view kVersionV1 = "V1";     // Value for version 1
+
+auto InitializeSystemDurability(std::optional<std::filesystem::path> storage, bool recovery_on_startup)
+    -> std::optional<memgraph::kvstore::KVStore> {
+  if (!storage) return std::nullopt;
+
+  auto const &path = *storage;
+  memgraph::utils::EnsureDir(path);
+  auto system_dir = path / kSystemDir;
+  memgraph::utils::EnsureDir(system_dir);
+  auto durability = memgraph::kvstore::KVStore{std::move(system_dir)};
+
+  auto version = durability.Get(kVersion);
+  // TODO: migration schemes here in the future
+  if (!version || *version != kVersionV1) {
+    // ensure we start out with V1
+    durability.Put(kVersion, kVersionV1);
+  }
+
+  if (!recovery_on_startup) {
+    // reset last_committed_system_ts
+    durability.Delete(kLastCommitedSystemTsKey);
+  }
+
+  return durability;
+}
+
+auto LoadLastCommittedSystemTimestamp(std::optional<kvstore::KVStore> const &store) -> uint64_t {
+  auto lcst = store ? store->Get(kLastCommitedSystemTsKey) : std::nullopt;
+  return lcst ? std::stoul(*lcst) : 0U;
+}
+
+}  // namespace
+
+State::State(std::optional<std::filesystem::path> storage, bool recovery_on_startup)
+    : durability_{InitializeSystemDurability(std::move(storage), recovery_on_startup)},
+      last_committed_system_timestamp_{LoadLastCommittedSystemTimestamp(durability_)} {}
+}  // namespace memgraph::system

src/system/system.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -8,14 +8,4 @@
 // the Business Source License, use of this software will be governed
 // by the Apache License, Version 2.0, included in the file
 // licenses/APL.txt.
-
-#pragma once
-
-#include "dbms/dbms_handler.hpp"
-#include "replication/replication_client.hpp"
-
-namespace memgraph::dbms {
-
-void StartReplicaClient(DbmsHandler &dbms_handler, replication::ReplicationClient &client);
-
-}  // namespace memgraph::dbms
+#include "system/include/system/system.hpp"

src/system/transaction.cpp

@@ -0,0 +1,11 @@
+// Copyright 2024 Memgraph Ltd.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
+// License, and you may not use this file except in compliance with the Business Source License.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+#include "system/include/system/transaction.hpp"

src/telemetry/telemetry.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -149,9 +149,9 @@ void Telemetry::AddClientCollector() {
 }
 
 #ifdef MG_ENTERPRISE
-void Telemetry::AddDatabaseCollector(dbms::DbmsHandler &dbms_handler) {
-  AddCollector("database", [&dbms_handler]() -> nlohmann::json {
-    const auto &infos = dbms_handler.Info();
+void Telemetry::AddDatabaseCollector(dbms::DbmsHandler &dbms_handler, replication::ReplicationState &repl_state) {
+  AddCollector("database", [&dbms_handler, &repl_state]() -> nlohmann::json {
+    const auto &infos = dbms_handler.Info(repl_state.GetRole());
     auto dbs = nlohmann::json::array();
     for (const auto &db_info : infos) {
       dbs.push_back(memgraph::dbms::ToJson(db_info));
@@ -162,11 +162,10 @@ void Telemetry::AddDatabaseCollector(dbms::DbmsHandler &dbms_handler) {
 #else
 #endif
 
-void Telemetry::AddStorageCollector(
-    dbms::DbmsHandler &dbms_handler,
-    memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> &auth) {
-  AddCollector("storage", [&dbms_handler, &auth]() -> nlohmann::json {
-    auto stats = dbms_handler.Stats();
+void Telemetry::AddStorageCollector(dbms::DbmsHandler &dbms_handler, memgraph::auth::SynchedAuth &auth,
+                                    memgraph::replication::ReplicationState &repl_state) {
+  AddCollector("storage", [&dbms_handler, &auth, &repl_state]() -> nlohmann::json {
+    auto stats = dbms_handler.Stats(repl_state.GetRole());
     stats.users = auth->AllUsers().size();
     return ToJson(stats);
   });

src/telemetry/telemetry.hpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -43,12 +43,11 @@ class Telemetry final {
   void AddCollector(const std::string &name, const std::function<const nlohmann::json(void)> &func);
 
   // Specialized collectors
-  void AddStorageCollector(
-      dbms::DbmsHandler &dbms_handler,
-      memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> &auth);
+  void AddStorageCollector(dbms::DbmsHandler &dbms_handler, memgraph::auth::SynchedAuth &auth,
+                           memgraph::replication::ReplicationState &repl_state);
 
 #ifdef MG_ENTERPRISE
-  void AddDatabaseCollector(dbms::DbmsHandler &dbms_handler);
+  void AddDatabaseCollector(dbms::DbmsHandler &dbms_handler, replication::ReplicationState &repl_state);
 #else
   void AddDatabaseCollector() {
     AddCollector("database", []() -> nlohmann::json { return nlohmann::json::array(); });

src/utils/gatekeeper.hpp

@@ -161,10 +161,22 @@ struct Gatekeeper {
 
     ~Accessor() { reset(); }
 
-    auto get() -> T * { return std::addressof(*owner_->value_); }
-    auto get() const -> const T * { return std::addressof(*owner_->value_); }
-    T *operator->() { return std::addressof(*owner_->value_); }
-    const T *operator->() const { return std::addressof(*owner_->value_); }
+    auto get() -> T * {
+      if (owner_ == nullptr) return nullptr;
+      return std::addressof(*owner_->value_);
+    }
+    auto get() const -> const T * {
+      if (owner_ == nullptr) return nullptr;
+      return std::addressof(*owner_->value_);
+    }
+    T *operator->() {
+      if (owner_ == nullptr) return nullptr;
+      return std::addressof(*owner_->value_);
+    }
+    const T *operator->() const {
+      if (owner_ == nullptr) return nullptr;
+      return std::addressof(*owner_->value_);
+    }
 
     template <typename Func>
     [[nodiscard]] auto try_exclusively(Func &&func) -> EvalResult<std::invoke_result_t<Func, T &>> {

src/utils/on_scope_exit.hpp

@@ -32,7 +32,7 @@ namespace memgraph::utils {
  * void long_function() {
  *     resource.enable();
  *     OnScopeExit on_exit([&resource] { resource.disable(); });
- *     // long block of code, might trow an exception
+ *     // long block of code, might throw an exception
  * }
  */
 template <typename Callable>

src/utils/typeinfo.hpp

@@ -93,6 +93,10 @@ enum class TypeId : uint64_t {
   REP_SYSTEM_HEARTBEAT_RES,
   REP_SYSTEM_RECOVERY_REQ,
   REP_SYSTEM_RECOVERY_RES,
+  REP_UPDATE_AUTH_DATA_REQ,
+  REP_UPDATE_AUTH_DATA_RES,
+  REP_DROP_AUTH_DATA_REQ,
+  REP_DROP_AUTH_DATA_RES,
 
   // Coordinator
   COORD_FAILOVER_REQ,

tests/benchmark/expansion.cpp

@@ -1,4 +1,4 @@
-// Copyright 2023 Memgraph Ltd.
+// Copyright 2024 Memgraph Ltd.
 //
 // Use of this software is governed by the Business Source License
 // included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -26,6 +26,7 @@ std::filesystem::path data_directory{std::filesystem::temp_directory_path() / "e
 
 class ExpansionBenchFixture : public benchmark::Fixture {
  protected:
+  std::optional<memgraph::system::System> system;
   std::optional<memgraph::query::InterpreterContext> interpreter_context;
   std::optional<memgraph::query::Interpreter> interpreter;
   std::optional<memgraph::utils::Gatekeeper<memgraph::dbms::Database>> db_gk;
@@ -40,7 +41,14 @@ class ExpansionBenchFixture : public benchmark::Fixture {
     auto db_acc_opt = db_gk->access();
     MG_ASSERT(db_acc_opt, "Failed to access db");
     auto &db_acc = *db_acc_opt;
-    interpreter_context.emplace(memgraph::query::InterpreterConfig{}, nullptr, &repl_state.value());
+
+    system.emplace();
+    interpreter_context.emplace(memgraph::query::InterpreterConfig{}, nullptr, &repl_state.value(), *system
+#ifdef MG_ENTERPRISE
+                                ,
+                                nullptr
+#endif
+    );
 
     auto label = db_acc->storage()->NameToLabel("Starting");
 
@@ -70,6 +78,7 @@ class ExpansionBenchFixture : public benchmark::Fixture {
   void TearDown(const benchmark::State &) override {
     interpreter = std::nullopt;
     interpreter_context = std::nullopt;
+    system.reset();
     db_gk.reset();
     std::filesystem::remove_all(data_directory);
   }

tests/e2e/high_availability_experimental/CMakeLists.txt

@@ -4,7 +4,6 @@ copy_e2e_python_files(ha_experimental coordinator.py)
 copy_e2e_python_files(ha_experimental automatic_failover.py)
 copy_e2e_python_files(ha_experimental manual_setting_replicas.py)
 copy_e2e_python_files(ha_experimental common.py)
-copy_e2e_python_files(ha_experimental conftest.py)
 copy_e2e_python_files(ha_experimental workloads.yaml)
 
 copy_e2e_python_files_from_parent_folder(ha_experimental ".." memgraph.py)

tests/e2e/high_availability_experimental/automatic_failover.py

@@ -1,4 +1,4 @@
-# Copyright 2024 Memgraph Ltd.
+# Copyright 2022 Memgraph Ltd.
 #
 # Use of this software is governed by the Business Source License
 # included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -10,11 +10,13 @@
 # licenses/APL.txt.
 
 import os
+import shutil
 import sys
+import tempfile
 
 import interactive_mg_runner
 import pytest
-from common import execute_and_fetch_all
+from common import connect, execute_and_fetch_all, safe_execute
 from mg_utils import mg_sleep_and_assert
 
 interactive_mg_runner.SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__))
@@ -24,25 +26,51 @@ interactive_mg_runner.PROJECT_DIR = os.path.normpath(
 interactive_mg_runner.BUILD_DIR = os.path.normpath(os.path.join(interactive_mg_runner.PROJECT_DIR, "build"))
 interactive_mg_runner.MEMGRAPH_BINARY = os.path.normpath(os.path.join(interactive_mg_runner.BUILD_DIR, "memgraph"))
 
+TEMP_DIR = tempfile.TemporaryDirectory().name
+
 MEMGRAPH_INSTANCES_DESCRIPTION = {
     "instance_1": {
-        "args": ["--bolt-port", "7688", "--log-level", "TRACE", "--coordinator-server-port", "10011"],
-        "log_file": "replica1.log",
+        "args": [
+            "--bolt-port",
+            "7688",
+            "--log-level",
+            "TRACE",
+            "--coordinator-server-port",
+            "10011",
+        ],
+        "log_file": "instance_1.log",
+        "data_directory": f"{TEMP_DIR}/instance_1",
         "setup_queries": [],
     },
     "instance_2": {
-        "args": ["--bolt-port", "7689", "--log-level", "TRACE", "--coordinator-server-port", "10012"],
-        "log_file": "replica2.log",
+        "args": [
+            "--bolt-port",
+            "7689",
+            "--log-level",
+            "TRACE",
+            "--coordinator-server-port",
+            "10012",
+        ],
+        "log_file": "instance_2.log",
+        "data_directory": f"{TEMP_DIR}/instance_2",
         "setup_queries": [],
     },
     "instance_3": {
-        "args": ["--bolt-port", "7687", "--log-level", "TRACE", "--coordinator-server-port", "10013"],
-        "log_file": "main.log",
+        "args": [
+            "--bolt-port",
+            "7687",
+            "--log-level",
+            "TRACE",
+            "--coordinator-server-port",
+            "10013",
+        ],
+        "log_file": "instance_3.log",
+        "data_directory": f"{TEMP_DIR}/instance_3",
         "setup_queries": [],
     },
     "coordinator": {
         "args": ["--bolt-port", "7690", "--log-level=TRACE", "--coordinator"],
-        "log_file": "replica3.log",
+        "log_file": "coordinator.log",
         "setup_queries": [
             "REGISTER INSTANCE instance_1 ON '127.0.0.1:10011' WITH '127.0.0.1:10001';",
             "REGISTER INSTANCE instance_2 ON '127.0.0.1:10012' WITH '127.0.0.1:10002';",
@@ -53,7 +81,7 @@ MEMGRAPH_INSTANCES_DESCRIPTION = {
 }
 
 
-def test_replication_works_on_failover(connection):
+def test_replication_works_on_failover():
     # Goal of this test is to check the replication works after failover command.
     # 1. We start all replicas, main and coordinator manually: we want to be able to kill them ourselves without relying on external tooling to kill processes.
     # 2. We check that main has correct state
@@ -61,12 +89,13 @@ def test_replication_works_on_failover(connection):
     # 4. We check that coordinator and new main have correct state
     # 5. We insert one vertex on new main
     # 6. We check that vertex appears on new replica
+    safe_execute(shutil.rmtree, TEMP_DIR)
 
     # 1
     interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
 
     # 2
-    main_cursor = connection(7687, "instance_3").cursor()
+    main_cursor = connect(host="localhost", port=7687).cursor()
     expected_data_on_main = [
         ("instance_1", "127.0.0.1:10001", "sync", 0, 0, "ready"),
         ("instance_2", "127.0.0.1:10002", "sync", 0, 0, "ready"),
@@ -78,7 +107,7 @@ def test_replication_works_on_failover(connection):
     interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
 
     # 4
-    coord_cursor = connection(7690, "coordinator").cursor()
+    coord_cursor = connect(host="localhost", port=7690).cursor()
 
     def retrieve_data_show_repl_cluster():
         return sorted(list(execute_and_fetch_all(coord_cursor, "SHOW REPLICATION CLUSTER;")))
@@ -86,17 +115,25 @@ def test_replication_works_on_failover(connection):
     expected_data_on_coord = [
         ("instance_1", "127.0.0.1:10011", True, "main"),
         ("instance_2", "127.0.0.1:10012", True, "replica"),
-        ("instance_3", "127.0.0.1:10013", False, ""),
+        ("instance_3", "127.0.0.1:10013", False, "unknown"),
     ]
     mg_sleep_and_assert(expected_data_on_coord, retrieve_data_show_repl_cluster)
 
-    new_main_cursor = connection(7688, "instance_1").cursor()
+    new_main_cursor = connect(host="localhost", port=7688).cursor()
 
     def retrieve_data_show_replicas():
         return sorted(list(execute_and_fetch_all(new_main_cursor, "SHOW REPLICAS;")))
 
     expected_data_on_new_main = [
         ("instance_2", "127.0.0.1:10002", "sync", 0, 0, "ready"),
+        ("instance_3", "127.0.0.1:10003", "sync", 0, 0, "invalid"),
+    ]
+    mg_sleep_and_assert(expected_data_on_new_main, retrieve_data_show_replicas)
+
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
+    expected_data_on_new_main = [
+        ("instance_2", "127.0.0.1:10002", "sync", 0, 0, "ready"),
+        ("instance_3", "127.0.0.1:10003", "sync", 0, 0, "ready"),
     ]
     mg_sleep_and_assert(expected_data_on_new_main, retrieve_data_show_replicas)
 
@@ -104,62 +141,68 @@ def test_replication_works_on_failover(connection):
     execute_and_fetch_all(new_main_cursor, "CREATE ();")
 
     # 6
-    alive_replica_cursror = connection(7689, "instance_2").cursor()
+    alive_replica_cursror = connect(host="localhost", port=7689).cursor()
     res = execute_and_fetch_all(alive_replica_cursror, "MATCH (n) RETURN count(n) as count;")[0][0]
     assert res == 1, "Vertex should be replicated"
     interactive_mg_runner.stop_all(MEMGRAPH_INSTANCES_DESCRIPTION)
 
 
-def test_show_replication_cluster(connection):
-    # Goal of this test is to check the SHOW REPLICATION CLUSTER command.
-    # 1. We start all replicas, main and coordinator manually: we want to be able to kill them ourselves without relying on external tooling to kill processes.
-    # 2. We check that all replicas and main have the correct state: they should all be alive.
-    # 3. We kill one replica. It should not appear anymore in the SHOW REPLICATION CLUSTER command.
-    # 4. We kill main. It should not appear anymore in the SHOW REPLICATION CLUSTER command.
-
-    # 1.
+def test_show_replication_cluster():
+    safe_execute(shutil.rmtree, TEMP_DIR)
     interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
 
-    cursor = connection(7690, "coordinator").cursor()
+    instance1_cursor = connect(host="localhost", port=7688).cursor()
+    instance2_cursor = connect(host="localhost", port=7689).cursor()
+    instance3_cursor = connect(host="localhost", port=7687).cursor()
+    coord_cursor = connect(host="localhost", port=7690).cursor()
 
-    # 2.
-
-    # We leave some time for the coordinator to realise the replicas are down.
-    def retrieve_data():
-        return sorted(list(execute_and_fetch_all(cursor, "SHOW REPLICATION CLUSTER;")))
+    def show_repl_cluster():
+        return sorted(list(execute_and_fetch_all(coord_cursor, "SHOW REPLICATION CLUSTER;")))
 
     expected_data = [
         ("instance_1", "127.0.0.1:10011", True, "replica"),
         ("instance_2", "127.0.0.1:10012", True, "replica"),
         ("instance_3", "127.0.0.1:10013", True, "main"),
     ]
-    mg_sleep_and_assert(expected_data, retrieve_data)
+    mg_sleep_and_assert(expected_data, show_repl_cluster)
+
+    def retrieve_data_show_repl_role_instance1():
+        return sorted(list(execute_and_fetch_all(instance1_cursor, "SHOW REPLICATION ROLE;")))
+
+    def retrieve_data_show_repl_role_instance2():
+        return sorted(list(execute_and_fetch_all(instance2_cursor, "SHOW REPLICATION ROLE;")))
+
+    def retrieve_data_show_repl_role_instance3():
+        return sorted(list(execute_and_fetch_all(instance3_cursor, "SHOW REPLICATION ROLE;")))
+
+    mg_sleep_and_assert([("replica",)], retrieve_data_show_repl_role_instance1)
+    mg_sleep_and_assert([("replica",)], retrieve_data_show_repl_role_instance2)
+    mg_sleep_and_assert([("main",)], retrieve_data_show_repl_role_instance3)
 
-    # 3.
     interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_1")
 
     expected_data = [
-        ("instance_1", "127.0.0.1:10011", False, ""),
+        ("instance_1", "127.0.0.1:10011", False, "unknown"),
         ("instance_2", "127.0.0.1:10012", True, "replica"),
         ("instance_3", "127.0.0.1:10013", True, "main"),
     ]
-    mg_sleep_and_assert(expected_data, retrieve_data)
+    mg_sleep_and_assert(expected_data, show_repl_cluster)
 
-    # 4.
     interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_2")
 
     expected_data = [
-        ("instance_1", "127.0.0.1:10011", False, ""),
-        ("instance_2", "127.0.0.1:10012", False, ""),
+        ("instance_1", "127.0.0.1:10011", False, "unknown"),
+        ("instance_2", "127.0.0.1:10012", False, "unknown"),
         ("instance_3", "127.0.0.1:10013", True, "main"),
     ]
-    mg_sleep_and_assert(expected_data, retrieve_data)
+    mg_sleep_and_assert(expected_data, show_repl_cluster)
 
 
-def test_simple_automatic_failover(connection):
+def test_simple_automatic_failover():
+    safe_execute(shutil.rmtree, TEMP_DIR)
     interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
 
-    main_cursor = connection(7687, "instance_3").cursor()
+    main_cursor = connect(host="localhost", port=7687).cursor()
     expected_data_on_main = [
         ("instance_1", "127.0.0.1:10001", "sync", 0, 0, "ready"),
         ("instance_2", "127.0.0.1:10002", "sync", 0, 0, "ready"),
@@ -169,7 +212,7 @@ def test_simple_automatic_failover(connection):
 
     interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
 
-    coord_cursor = connection(7690, "coordinator").cursor()
+    coord_cursor = connect(host="localhost", port=7690).cursor()
 
     def retrieve_data_show_repl_cluster():
         return sorted(list(execute_and_fetch_all(coord_cursor, "SHOW REPLICATION CLUSTER;")))
@@ -177,46 +220,189 @@ def test_simple_automatic_failover(connection):
     expected_data_on_coord = [
         ("instance_1", "127.0.0.1:10011", True, "main"),
         ("instance_2", "127.0.0.1:10012", True, "replica"),
-        ("instance_3", "127.0.0.1:10013", False, ""),
+        ("instance_3", "127.0.0.1:10013", False, "unknown"),
     ]
     mg_sleep_and_assert(expected_data_on_coord, retrieve_data_show_repl_cluster)
 
-    new_main_cursor = connection(7688, "instance_1").cursor()
+    new_main_cursor = connect(host="localhost", port=7688).cursor()
 
     def retrieve_data_show_replicas():
         return sorted(list(execute_and_fetch_all(new_main_cursor, "SHOW REPLICAS;")))
 
     expected_data_on_new_main = [
         ("instance_2", "127.0.0.1:10002", "sync", 0, 0, "ready"),
+        ("instance_3", "127.0.0.1:10003", "sync", 0, 0, "invalid"),
     ]
     mg_sleep_and_assert(expected_data_on_new_main, retrieve_data_show_replicas)
 
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
+    expected_data_on_new_main_old_alive = [
+        ("instance_2", "127.0.0.1:10002", "sync", 0, 0, "ready"),
+        ("instance_3", "127.0.0.1:10003", "sync", 0, 0, "ready"),
+    ]
 
-def test_registering_replica_fails_name_exists(connection):
+    mg_sleep_and_assert(expected_data_on_new_main_old_alive, retrieve_data_show_replicas)
+
+
+def test_registering_replica_fails_name_exists():
+    safe_execute(shutil.rmtree, TEMP_DIR)
     interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
 
-    coord_cursor = connection(7690, "coordinator").cursor()
+    coord_cursor = connect(host="localhost", port=7690).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(
             coord_cursor,
             "REGISTER INSTANCE instance_1 ON '127.0.0.1:10051' WITH '127.0.0.1:10111';",
         )
     assert str(e.value) == "Couldn't register replica instance since instance with such name already exists!"
+    shutil.rmtree(TEMP_DIR)
 
 
-def test_registering_replica_fails_endpoint_exists(connection):
+def test_registering_replica_fails_endpoint_exists():
+    safe_execute(shutil.rmtree, TEMP_DIR)
     interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
 
-    coord_cursor = connection(7690, "coordinator").cursor()
+    coord_cursor = connect(host="localhost", port=7690).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(
             coord_cursor,
-            "REGISTER INSTANCE instance_5 ON '127.0.0.1:10001' WITH '127.0.0.1:10013';",
+            "REGISTER INSTANCE instance_5 ON '127.0.0.1:10011' WITH '127.0.0.1:10005';",
         )
-    assert (
-        str(e.value)
-        == "Couldn't register replica because promotion on replica failed! Check logs on replica to find out more info!"
-    )
+    assert str(e.value) == "Couldn't register replica instance since instance with such endpoint already exists!"
+
+
+def test_replica_instance_restarts():
+    safe_execute(shutil.rmtree, TEMP_DIR)
+    interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
+
+    cursor = connect(host="localhost", port=7690).cursor()
+
+    def show_repl_cluster():
+        return sorted(list(execute_and_fetch_all(cursor, "SHOW REPLICATION CLUSTER;")))
+
+    expected_data_up = [
+        ("instance_1", "127.0.0.1:10011", True, "replica"),
+        ("instance_2", "127.0.0.1:10012", True, "replica"),
+        ("instance_3", "127.0.0.1:10013", True, "main"),
+    ]
+    mg_sleep_and_assert(expected_data_up, show_repl_cluster)
+
+    interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_1")
+
+    expected_data_down = [
+        ("instance_1", "127.0.0.1:10011", False, "unknown"),
+        ("instance_2", "127.0.0.1:10012", True, "replica"),
+        ("instance_3", "127.0.0.1:10013", True, "main"),
+    ]
+    mg_sleep_and_assert(expected_data_down, show_repl_cluster)
+
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_1")
+
+    mg_sleep_and_assert(expected_data_up, show_repl_cluster)
+
+    instance1_cursor = connect(host="localhost", port=7688).cursor()
+
+    def retrieve_data_show_repl_role_instance1():
+        return sorted(list(execute_and_fetch_all(instance1_cursor, "SHOW REPLICATION ROLE;")))
+
+    expected_data_replica = [("replica",)]
+    mg_sleep_and_assert(expected_data_replica, retrieve_data_show_repl_role_instance1)
+
+
+def test_automatic_failover_main_back_as_replica():
+    safe_execute(shutil.rmtree, TEMP_DIR)
+    interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
+
+    interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
+
+    coord_cursor = connect(host="localhost", port=7690).cursor()
+
+    def retrieve_data_show_repl_cluster():
+        return sorted(list(execute_and_fetch_all(coord_cursor, "SHOW REPLICATION CLUSTER;")))
+
+    expected_data_after_failover = [
+        ("instance_1", "127.0.0.1:10011", True, "main"),
+        ("instance_2", "127.0.0.1:10012", True, "replica"),
+        ("instance_3", "127.0.0.1:10013", False, "unknown"),
+    ]
+    mg_sleep_and_assert(expected_data_after_failover, retrieve_data_show_repl_cluster)
+
+    expected_data_after_main_coming_back = [
+        ("instance_1", "127.0.0.1:10011", True, "main"),
+        ("instance_2", "127.0.0.1:10012", True, "replica"),
+        ("instance_3", "127.0.0.1:10013", True, "replica"),
+    ]
+
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
+    mg_sleep_and_assert(expected_data_after_main_coming_back, retrieve_data_show_repl_cluster)
+
+    instance3_cursor = connect(host="localhost", port=7687).cursor()
+
+    def retrieve_data_show_repl_role_instance3():
+        return sorted(list(execute_and_fetch_all(instance3_cursor, "SHOW REPLICATION ROLE;")))
+
+    mg_sleep_and_assert([("replica",)], retrieve_data_show_repl_role_instance3)
+
+
+def test_automatic_failover_main_back_as_main():
+    safe_execute(shutil.rmtree, TEMP_DIR)
+    interactive_mg_runner.start_all(MEMGRAPH_INSTANCES_DESCRIPTION)
+
+    interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_1")
+    interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_2")
+    interactive_mg_runner.kill(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
+
+    coord_cursor = connect(host="localhost", port=7690).cursor()
+
+    def retrieve_data_show_repl_cluster():
+        return sorted(list(execute_and_fetch_all(coord_cursor, "SHOW REPLICATION CLUSTER;")))
+
+    expected_data_all_down = [
+        ("instance_1", "127.0.0.1:10011", False, "unknown"),
+        ("instance_2", "127.0.0.1:10012", False, "unknown"),
+        ("instance_3", "127.0.0.1:10013", False, "unknown"),
+    ]
+
+    mg_sleep_and_assert(expected_data_all_down, retrieve_data_show_repl_cluster)
+
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_3")
+    expected_data_main_back = [
+        ("instance_1", "127.0.0.1:10011", False, "unknown"),
+        ("instance_2", "127.0.0.1:10012", False, "unknown"),
+        ("instance_3", "127.0.0.1:10013", True, "main"),
+    ]
+    mg_sleep_and_assert(expected_data_main_back, retrieve_data_show_repl_cluster)
+
+    instance3_cursor = connect(host="localhost", port=7687).cursor()
+
+    def retrieve_data_show_repl_role_instance3():
+        return sorted(list(execute_and_fetch_all(instance3_cursor, "SHOW REPLICATION ROLE;")))
+
+    mg_sleep_and_assert([("main",)], retrieve_data_show_repl_role_instance3)
+
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_1")
+    interactive_mg_runner.start(MEMGRAPH_INSTANCES_DESCRIPTION, "instance_2")
+
+    expected_data_replicas_back = [
+        ("instance_1", "127.0.0.1:10011", True, "replica"),
+        ("instance_2", "127.0.0.1:10012", True, "replica"),
+        ("instance_3", "127.0.0.1:10013", True, "main"),
+    ]
+
+    mg_sleep_and_assert(expected_data_replicas_back, retrieve_data_show_repl_cluster)
+
+    instance1_cursor = connect(host="localhost", port=7688).cursor()
+    instance2_cursor = connect(host="localhost", port=7689).cursor()
+
+    def retrieve_data_show_repl_role_instance1():
+        return sorted(list(execute_and_fetch_all(instance1_cursor, "SHOW REPLICATION ROLE;")))
+
+    def retrieve_data_show_repl_role_instance2():
+        return sorted(list(execute_and_fetch_all(instance2_cursor, "SHOW REPLICATION ROLE;")))
+
+    mg_sleep_and_assert([("replica",)], retrieve_data_show_repl_role_instance1)
+    mg_sleep_and_assert([("replica",)], retrieve_data_show_repl_role_instance2)
+    mg_sleep_and_assert([("main",)], retrieve_data_show_repl_role_instance3)
 
 
 if __name__ == "__main__":

tests/e2e/high_availability_experimental/common.py

@@ -23,3 +23,10 @@ def connect(**kwargs) -> mgclient.Connection:
     connection = mgclient.connect(**kwargs)
     connection.autocommit = True
     return connection
+
+
+def safe_execute(function, *args):
+    try:
+        function(*args)
+    except:
+        pass

tests/e2e/high_availability_experimental/conftest.py

@@ -1,43 +0,0 @@
-# Copyright 2022 Memgraph Ltd.
-#
-# Use of this software is governed by the Business Source License
-# included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
-# License, and you may not use this file except in compliance with the Business Source License.
-#
-# As of the Change Date specified in that file, in accordance with
-# the Business Source License, use of this software will be governed
-# by the Apache License, Version 2.0, included in the file
-# licenses/APL.txt.
-
-import pytest
-from common import connect, execute_and_fetch_all
-
-
-# The fixture here is more complex because the connection has to be
-# parameterized based on the test parameters (info has to be available on both
-# sides).
-#
-# https://docs.pytest.org/en/latest/example/parametrize.html#indirect-parametrization
-# is not an elegant/feasible solution here.
-#
-# The solution was independently developed and then I stumbled upon the same
-# approach here https://stackoverflow.com/a/68286553/4888809 which I think is
-# optimal.
-@pytest.fixture(scope="function")
-def connection():
-    connection_holder = None
-    role_holder = None
-
-    def inner_connection(port, role):
-        nonlocal connection_holder, role_holder
-        connection_holder = connect(host="localhost", port=port)
-        role_holder = role
-        return connection_holder
-
-    yield inner_connection
-
-    # Only main instance can be cleaned up because replicas do NOT accept
-    # writes.
-    if role_holder == "main":
-        cursor = connection_holder.cursor()
-        execute_and_fetch_all(cursor, "MATCH (n) DETACH DELETE n;")

tests/e2e/high_availability_experimental/coordinator.py

@@ -12,33 +12,33 @@
 import sys
 
 import pytest
-from common import execute_and_fetch_all
+from common import connect, execute_and_fetch_all
 from mg_utils import mg_sleep_and_assert
 
 
-def test_disable_cypher_queries(connection):
-    cursor = connection(7690, "coordinator").cursor()
+def test_disable_cypher_queries():
+    cursor = connect(host="localhost", port=7690).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(cursor, "CREATE (n:TestNode {prop: 'test'})")
     assert str(e.value) == "Coordinator can run only coordinator queries!"
 
 
-def test_coordinator_cannot_be_replica_role(connection):
-    cursor = connection(7690, "coordinator").cursor()
+def test_coordinator_cannot_be_replica_role():
+    cursor = connect(host="localhost", port=7690).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(cursor, "SET REPLICATION ROLE TO REPLICA WITH PORT 10001;")
     assert str(e.value) == "Coordinator can run only coordinator queries!"
 
 
-def test_coordinator_cannot_run_show_repl_role(connection):
-    cursor = connection(7690, "coordinator").cursor()
+def test_coordinator_cannot_run_show_repl_role():
+    cursor = connect(host="localhost", port=7690).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(cursor, "SHOW REPLICATION ROLE;")
     assert str(e.value) == "Coordinator can run only coordinator queries!"
 
 
-def test_coordinator_show_replication_cluster(connection):
-    cursor = connection(7690, "coordinator").cursor()
+def test_coordinator_show_replication_cluster():
+    cursor = connect(host="localhost", port=7690).cursor()
 
     def retrieve_data():
         return sorted(list(execute_and_fetch_all(cursor, "SHOW REPLICATION CLUSTER;")))
@@ -51,30 +51,30 @@ def test_coordinator_show_replication_cluster(connection):
     mg_sleep_and_assert(expected_data, retrieve_data)
 
 
-def test_coordinator_cannot_call_show_replicas(connection):
-    cursor = connection(7690, "coordinator").cursor()
+def test_coordinator_cannot_call_show_replicas():
+    cursor = connect(host="localhost", port=7690).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(cursor, "SHOW REPLICAS;")
     assert str(e.value) == "Coordinator can run only coordinator queries!"
 
 
 @pytest.mark.parametrize(
-    "port, role",
-    [(7687, "main"), (7688, "replica"), (7689, "replica")],
+    "port",
+    [7687, 7688, 7689],
 )
-def test_main_and_replicas_cannot_call_show_repl_cluster(port, role, connection):
-    cursor = connection(port, role).cursor()
+def test_main_and_replicas_cannot_call_show_repl_cluster(port):
+    cursor = connect(host="localhost", port=port).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(cursor, "SHOW REPLICATION CLUSTER;")
     assert str(e.value) == "Only coordinator can run SHOW REPLICATION CLUSTER."
 
 
 @pytest.mark.parametrize(
-    "port, role",
-    [(7687, "main"), (7688, "replica"), (7689, "replica")],
+    "port",
+    [7687, 7688, 7689],
 )
-def test_main_and_replicas_cannot_register_coord_server(port, role, connection):
-    cursor = connection(port, role).cursor()
+def test_main_and_replicas_cannot_register_coord_server(port):
+    cursor = connect(host="localhost", port=port).cursor()
     with pytest.raises(Exception) as e:
         execute_and_fetch_all(
             cursor,

tests/e2e/interactive_mg_runner.py

@@ -105,7 +105,9 @@ def is_port_in_use(port: int) -> bool:
         return s.connect_ex(("localhost", port)) == 0
 
 
-def _start_instance(name, args, log_file, setup_queries, use_ssl, procdir, data_directory):
+def _start_instance(
+    name, args, log_file, setup_queries, use_ssl, procdir, data_directory, username=None, password=None
+):
     assert (
         name not in MEMGRAPH_INSTANCES.keys()
     ), "If this raises, you are trying to start an instance with the same name than one already running."
@@ -115,7 +117,9 @@ def _start_instance(name, args, log_file, setup_queries, use_ssl, procdir, data_
 
     log_file_path = os.path.join(BUILD_DIR, "logs", log_file)
     data_directory_path = os.path.join(BUILD_DIR, data_directory)
-    mg_instance = MemgraphInstanceRunner(MEMGRAPH_BINARY, use_ssl, {data_directory_path})
+    mg_instance = MemgraphInstanceRunner(
+        MEMGRAPH_BINARY, use_ssl, {data_directory_path}, username=username, password=password
+    )
     MEMGRAPH_INSTANCES[name] = mg_instance
     binary_args = args + ["--log-file", log_file_path] + ["--data-directory", data_directory_path]
 
@@ -185,8 +189,14 @@ def start_instance(context, name, procdir):
             data_directory = value["data_directory"]
         else:
             data_directory = tempfile.TemporaryDirectory().name
+        username = None
+        if "username" in value:
+            username = value["username"]
+        password = None
+        if "password" in value:
+            password = value["password"]
 
-        instance = _start_instance(name, args, log_file, queries, use_ssl, procdir, data_directory)
+        instance = _start_instance(name, args, log_file, queries, use_ssl, procdir, data_directory, username, password)
         mg_instances[name] = instance
 
     assert len(mg_instances) == 1

tests/e2e/memgraph.py

@@ -57,7 +57,7 @@ def replace_paths(path):
 
 
 class MemgraphInstanceRunner:
-    def __init__(self, binary_path=MEMGRAPH_BINARY, use_ssl=False, delete_on_stop=None):
+    def __init__(self, binary_path=MEMGRAPH_BINARY, use_ssl=False, delete_on_stop=None, username=None, password=None):
         self.host = "127.0.0.1"
         self.bolt_port = None
         self.binary_path = binary_path
@@ -65,12 +65,19 @@ class MemgraphInstanceRunner:
         self.proc_mg = None
         self.ssl = use_ssl
         self.delete_on_stop = delete_on_stop
+        self.username = username
+        self.password = password
 
     def execute_setup_queries(self, setup_queries):
         if setup_queries is None:
             return
-        # An assumption being database instance is fresh, no need for the auth.
-        conn = mgclient.connect(host=self.host, port=self.bolt_port, sslmode=self.ssl)
+        conn = mgclient.connect(
+            host=self.host,
+            port=self.bolt_port,
+            sslmode=self.ssl,
+            username=(self.username or ""),
+            password=(self.password or ""),
+        )
         conn.autocommit = True
         cursor = conn.cursor()
         for query_coll in setup_queries: