Add unreachable replica state

Fix accumulated path evaluation in DFS, BFS, WeghtedShortestPath and AllShortestPath algorithm.

.clang-tidy

@@ -6,6 +6,7 @@ Checks: '*,
          -altera-unroll-loops,
          -android-*,
          -cert-err58-cpp,
+         -cppcoreguidelines-avoid-do-while,
          -cppcoreguidelines-avoid-c-arrays,
          -cppcoreguidelines-avoid-goto,
          -cppcoreguidelines-avoid-magic-numbers,
@@ -60,6 +61,7 @@ Checks: '*,
          -readability-implicit-bool-conversion,
          -readability-magic-numbers,
          -readability-named-parameter,
+         -readability-identifier-length,
          -misc-no-recursion,
          -concurrency-mt-unsafe,
          -bugprone-easily-swappable-parameters'

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.

CMakeLists.txt

@@ -275,7 +275,7 @@ option(MG_EXPERIMENTAL_HIGH_AVAILABILITY "Feature flag for experimental high ava
 
 if (NOT MG_ENTERPRISE AND MG_EXPERIMENTAL_HIGH_AVAILABILITY)
     set(MG_EXPERIMENTAL_HIGH_AVAILABILITY OFF)
-    message(FATAL_ERROR "MG_EXPERIMENTAL_HIGH_AVAILABILITY must be used with enterpise version of the code.")
+    message(FATAL_ERROR "MG_EXPERIMENTAL_HIGH_AVAILABILITY can only be used with enterpise version of the code.")
 endif ()
 
 if (MG_EXPERIMENTAL_HIGH_AVAILABILITY)

libs/setup.sh

@@ -123,10 +123,10 @@ declare -A primary_urls=(
   ["pulsar"]="http://$local_cache_host/git/pulsar.git"
   ["librdtsc"]="http://$local_cache_host/git/librdtsc.git"
   ["ctre"]="http://$local_cache_host/file/hanickadot/compile-time-regular-expressions/v3.7.2/single-header/ctre.hpp"
-  ["absl"]="https://$local_cache_host/git/abseil-cpp.git"
-  ["jemalloc"]="https://$local_cache_host/git/jemalloc.git"
-  ["range-v3"]="https://$local_cache_host/git/ericniebler/range-v3.git"
-  ["nuraft"]="https://$local_cache_host/git/eBay/NuRaft.git"
+  ["absl"]="http://$local_cache_host/git/abseil-cpp.git"
+  ["jemalloc"]="http://$local_cache_host/git/jemalloc.git"
+  ["range-v3"]="http://$local_cache_host/git/range-v3.git"
+  ["nuraft"]="http://$local_cache_host/git/NuRaft.git"
 )
 
 # The goal of secondary urls is to have links to the "source of truth" of
@@ -270,6 +270,8 @@ pushd jemalloc
 MALLOC_CONF="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000" \
 ./configure \
   --disable-cxx \
+  --with-lg-page=12 \
+  --with-lg-hugepage=21 \
   --enable-shared=no --prefix=$working_dir \
   --with-malloc-conf="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000"
 

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 +44,89 @@ 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, const utils::UUID &main_uuid,
+                     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, main_uuid, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(),
+          *user_);
+    }
+    if (role_) {
+      return client.SteamAndFinalizeDelta<replication::UpdateAuthDataRpc>(
+          check_response, main_uuid, 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, const utils::UUID &main_uuid,
+                     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, main_uuid, 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 +228,12 @@ 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 +326,7 @@ 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 +338,12 @@ 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 +376,8 @@ 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 +387,23 @@ 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 +420,19 @@ 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 +450,30 @@ 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 +486,12 @@ 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 +509,18 @@ 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 +540,48 @@ 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 +1,190 @@
+// 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 {
+
+void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
+                  std::string_view rpc_req) {
+  spdlog::error(fmt::format("Received {} with main_id: {} != current_main_uuid: {}", rpc_req, std::string(main_req_id),
+                            current_main_uuid.has_value() ? std::string(current_main_uuid.value()) : ""));
+}
+
+#ifdef MG_ENTERPRISE
+void UpdateAuthDataHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                           const std::optional<utils::UUID> &current_main_uuid, 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);
+
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, replication::UpdateAuthDataReq::kType.name);
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  // 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,
+                         const std::optional<utils::UUID> &current_main_uuid, 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);
+
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, replication::DropAuthDataRes::kType.name);
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  // 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>(
+      [&data, system_state_access, &auth](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received UpdateAuthDataRpc");
+        UpdateAuthDataHandler(system_state_access, data.uuid_, auth, req_reader, res_builder);
+      });
+  data.server->rpc_server_.Register<replication::DropAuthDataRpc>(
+      [&data, system_state_access, &auth](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received DropAuthDataRpc");
+        DropAuthDataHandler(system_state_access, data.uuid_, auth, req_reader, res_builder);
+      });
+}
+#endif
+
+}  // namespace memgraph::auth

src/auth/replication_handlers.hpp

@@ -0,0 +1,37 @@
+// 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 {
+
+void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
+                  std::string_view rpc_req);
+
+#ifdef MG_ENTERPRISE
+void UpdateAuthDataHandler(system::ReplicaHandlerAccessToState &system_state_access,
+                           const std::optional<utils::UUID> &current_main_uuid, auth::SynchedAuth &auth,
+                           slk::Reader *req_reader, slk::Builder *res_builder);
+void DropAuthDataHandler(system::ReplicaHandlerAccessToState &system_state_access,
+                         const std::optional<utils::UUID> &current_main_uuid, 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 +1,182 @@
+// 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.main_uuid, 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->main_uuid, 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.main_uuid, 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->main_uuid, 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 +1,127 @@
+// 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(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_ts, uint64_t new_ts,
+                    auth::User user)
+      : main_uuid(main_uuid),
+        epoch_id{std::move(epoch_id)},
+        expected_group_timestamp{expected_ts},
+        new_group_timestamp{new_ts},
+        user{std::move(user)} {}
+  UpdateAuthDataReq(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_ts, uint64_t new_ts,
+                    auth::Role role)
+      : main_uuid(main_uuid),
+        epoch_id{std::move(epoch_id)},
+        expected_group_timestamp{expected_ts},
+        new_group_timestamp{new_ts},
+        role{std::move(role)} {}
+
+  utils::UUID main_uuid;
+  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(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_ts, uint64_t new_ts,
+                  DataType type, std::string_view name)
+      : main_uuid(main_uuid),
+        epoch_id{std::move(epoch_id)},
+        expected_group_timestamp{expected_ts},
+        new_group_timestamp{new_ts},
+        type{type},
+        name{name} {}
+
+  utils::UUID main_uuid;
+  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

@@ -8,22 +8,35 @@ target_sources(mg-coordination
         include/coordination/coordinator_server.hpp
         include/coordination/coordinator_config.hpp
         include/coordination/coordinator_exceptions.hpp
-        include/coordination/coordinator_instance.hpp
         include/coordination/coordinator_slk.hpp
-        include/coordination/coordinator_data.hpp
-        include/coordination/constants.hpp
-        include/coordination/failover_status.hpp
+        include/coordination/coordinator_instance.hpp
         include/coordination/coordinator_cluster_config.hpp
+        include/coordination/coordinator_handlers.hpp
+        include/coordination/constants.hpp
+        include/coordination/instance_status.hpp
+        include/coordination/replication_instance.hpp
+        include/coordination/raft_instance.hpp
+
+        include/nuraft/coordinator_log_store.hpp
+        include/nuraft/coordinator_state_machine.hpp
+        include/nuraft/coordinator_state_manager.hpp
 
         PRIVATE
         coordinator_client.cpp
         coordinator_state.cpp
         coordinator_rpc.cpp
         coordinator_server.cpp
-        coordinator_data.cpp
+        coordinator_handlers.cpp
+        coordinator_instance.cpp
+        replication_instance.cpp
+        raft_instance.cpp
+
+        coordinator_log_store.cpp
+        coordinator_state_machine.cpp
+        coordinator_state_manager.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

@@ -9,6 +9,7 @@
 // by the Apache License, Version 2.0, included in the file
 // licenses/APL.txt.
 
+#include "utils/uuid.hpp"
 #ifdef MG_ENTERPRISE
 
 #include "coordination/coordinator_client.hpp"
@@ -20,20 +21,20 @@
 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{};
 }
 }  // namespace
 
-CoordinatorClient::CoordinatorClient(CoordinatorData *coord_data, CoordinatorClientConfig config,
+CoordinatorClient::CoordinatorClient(CoordinatorInstance *coord_instance, CoordinatorClientConfig config,
                                      HealthCheckCallback succ_cb, HealthCheckCallback fail_cb)
     : rpc_context_{CreateClientContext(config)},
       rpc_client_{io::network::Endpoint(io::network::Endpoint::needs_resolving, config.ip_address, config.port),
                   &rpc_context_},
       config_{std::move(config)},
-      coord_data_{coord_data},
+      coord_instance_{coord_instance},
       succ_cb_{std::move(succ_cb)},
       fail_cb_{std::move(fail_cb)} {}
 
@@ -41,66 +42,81 @@ auto CoordinatorClient::InstanceName() const -> std::string { return config_.ins
 auto CoordinatorClient::SocketAddress() const -> std::string { return rpc_client_.Endpoint().SocketAddress(); }
 
 void CoordinatorClient::StartFrequentCheck() {
+  if (instance_checker_.IsRunning()) {
+    return;
+  }
+
   MG_ASSERT(config_.health_check_frequency_sec > std::chrono::seconds(0),
             "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();
-          succ_cb_(coord_data_, instance_name);
-        } catch (const rpc::RpcFailedException &) {
-          fail_cb_(coord_data_, instance_name);
+          {  // 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_instance_, instance_name);
+        } catch (rpc::RpcFailedException const &) {
+          fail_cb_(coord_instance_, 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(const utils::UUID &uuid,
+                                                    ReplicationClientsInfo replication_clients_info) const -> bool {
   try {
-    auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(std::move(replication_clients_info))};
+    auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(uuid, std::move(replication_clients_info))};
     if (!stream.AwaitResponse().success) {
-      spdlog::error("Failed to receive successful RPC failover response!");
+      spdlog::error("Failed to receive successful PromoteReplicaToMainRpc response!");
       return false;
     }
     return true;
-  } catch (const rpc::RpcFailedException &) {
-    spdlog::error("RPC error occurred while sending failover RPC!");
+  } catch (rpc::RpcFailedException const &) {
+    spdlog::error("RPC error occurred while sending PromoteReplicaToMainRpc!");
   }
   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 (rpc::RpcFailedException const &) {
+    spdlog::error("Failed to set instance {} to replica!", instance_name);
+  }
+  return false;
+}
+
+auto CoordinatorClient::SendSwapMainUUIDRpc(const utils::UUID &uuid) const -> bool {
+  try {
+    auto stream{rpc_client_.Stream<replication_coordination_glue::SwapMainUUIDRpc>(uuid)};
+    if (!stream.AwaitResponse().success) {
+      spdlog::error("Failed to receive successful RPC swapping of uuid response!");
+      return false;
+    }
+    return true;
   } catch (const rpc::RpcFailedException &) {
-    spdlog::error("Failed to send failover RPC from coordinator to new main!");
+    spdlog::error("RPC error occurred while sending swapping uuid RPC!");
   }
   return false;
 }

src/coordination/coordinator_data.cpp

@@ -1,221 +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 "coordination/coordinator_instance.hpp"
-#include "coordination/register_main_replica_coordinator_status.hpp"
-#ifdef MG_ENTERPRISE
-
-#include "coordination/coordinator_data.hpp"
-
-#include <range/v3/view.hpp>
-#include <shared_mutex>
-#include "libnuraft/nuraft.hxx"
-
-namespace memgraph::coordination {
-
-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; });
-
-    MG_ASSERT(instance != coord_data->registered_instances_.end(), "Instance {} not found during callback!",
-              instance_name);
-    return *instance;
-  };
-
-  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();
-  };
-
-  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();
-  };
-
-  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();
-  };
-
-  main_fail_cb_ = [this, 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;
-      }
-    }
-  };
-}
-
-auto CoordinatorData::DoFailover() -> DoFailoverStatus {
-  using ReplicationClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
-
-  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;
-  }
-
-  chosen_replica_instance->PrepareForFailover();
-
-  std::vector<ReplicationClientInfo> repl_clients_info;
-  repl_clients_info.reserve(std::ranges::distance(replica_instances));
-
-  auto const not_chosen_replica_instance = [&chosen_replica_instance](const CoordinatorInstance &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());
-  }
-
-  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;
-}
-
-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 "";
-    if (instance.IsMain()) return "main";
-    return "replica";
-  };
-
-  auto const instance_to_status =
-      [&stringify_repl_role](const CoordinatorInstance &instance) -> CoordinatorInstanceStatus {
-    return {.instance_name = instance.InstanceName(),
-            .socket_address = instance.SocketAddress(),
-            .replication_role = stringify_repl_role(instance),
-            .is_alive = instance.IsAlive()};
-  };
-
-  {
-    auto lock = std::shared_lock{coord_data_lock_};
-    std::ranges::transform(registered_instances_, std::back_inserter(instances_status), instance_to_status);
-  }
-
-  return instances_status;
-}
-
-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; });
-
-  // if replica not found...
-  if (registered_replica == registered_instances_.end()) {
-    spdlog::error("You didn't register instance with given name {}", instance_name);
-    return SetInstanceToMainCoordinatorStatus::NO_INSTANCE_WITH_NAME;
-  }
-
-  registered_replica->client_.PauseFrequentCheck();
-
-  std::vector<CoordinatorClientConfig::ReplicationClientInfo> 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();
-    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();
-
-  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) {
-        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());
-        return instance.SocketAddress() == config.SocketAddress();
-      })) {
-    return RegisterInstanceCoordinatorStatus::END_POINT_EXISTS;
-  }
-
-  CoordinatorClientConfig::ReplicationClientInfo replication_client_info_copy = config.replication_client_info;
-
-  // 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) {
-    return RegisterInstanceCoordinatorStatus::RPC_FAILED;
-  }
-
-  instance->client_.StartFrequentCheck();
-
-  return RegisterInstanceCoordinatorStatus::SUCCESS;
-}
-
-}  // namespace memgraph::coordination
-#endif

src/coordination/coordinator_handlers.cpp

@@ -0,0 +1,150 @@
+// 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"
+#include "replication/state.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);
+      });
+
+  server.Register<replication_coordination_glue::SwapMainUUIDRpc>(
+      [&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
+        spdlog::info("Received SwapMainUUIDRPC on coordinator server");
+        CoordinatorHandlers::SwapMainUUIDHandler(replication_handler, req_reader, res_builder);
+      });
+}
+
+void CoordinatorHandlers::SwapMainUUIDHandler(replication::ReplicationHandler &replication_handler,
+                                              slk::Reader *req_reader, slk::Builder *res_builder) {
+  if (!replication_handler.IsReplica()) {
+    spdlog::error("Setting main uuid must be performed on replica.");
+    slk::Save(replication_coordination_glue::SwapMainUUIDRes{false}, res_builder);
+    return;
+  }
+
+  replication_coordination_glue::SwapMainUUIDReq req;
+  slk::Load(&req, req_reader);
+  spdlog::info(fmt::format("Set replica data UUID  to main uuid {}", std::string(req.uuid)));
+  std::get<memgraph::replication::RoleReplicaData>(replication_handler.GetReplState().ReplicationData()).uuid_ =
+      req.uuid;
+
+  slk::Save(replication_coordination_glue::SwapMainUUIDRes{true}, 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, std::nullopt)) {
+    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;
+  }
+  coordination::PromoteReplicaToMainReq req;
+  slk::Load(&req, req_reader);
+
+  // This can fail because of disk. If it does, the cluster state could get inconsistent.
+  // We don't handle disk issues.
+  if (const bool success = replication_handler.DoReplicaToMainPromotion(req.main_uuid_); !success) {
+    spdlog::error("Promoting replica to main failed!");
+    slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+    return;
+  }
+
+  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, false);
+    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::ERROR_ACCEPTING_MAIN:
+          spdlog::error("Replica didn't accept change of main!");
+          slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
+          return;
+        case memgraph::query::RegisterReplicaError::CONNECTION_FAILED:
+          // Connection failure is not a fatal error
+          break;
+      }
+    }
+  }
+  spdlog::error(fmt::format("FICO : Promote replica to main was success {}", std::string(req.main_uuid_)));
+  slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
+}
+
+}  // namespace memgraph::dbms
+#endif

src/coordination/coordinator_instance.cpp

@@ -0,0 +1,289 @@
+// 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"
+
+#include "coordination/coordinator_exceptions.hpp"
+#include "nuraft/coordinator_state_machine.hpp"
+#include "nuraft/coordinator_state_manager.hpp"
+#include "utils/counter.hpp"
+
+#include <range/v3/view.hpp>
+#include <shared_mutex>
+
+namespace memgraph::coordination {
+
+using nuraft::ptr;
+using nuraft::srv_config;
+
+CoordinatorInstance::CoordinatorInstance()
+    : self_([this] { std::ranges::for_each(repl_instances_, &ReplicationInstance::StartFrequentCheck); },
+            [this] { std::ranges::for_each(repl_instances_, &ReplicationInstance::StopFrequentCheck); }) {
+  auto find_instance = [](CoordinatorInstance *coord_instance,
+                          std::string_view instance_name) -> ReplicationInstance & {
+    auto instance = std::ranges::find_if(
+        coord_instance->repl_instances_,
+        [instance_name](ReplicationInstance const &instance) { return instance.InstanceName() == instance_name; });
+
+    MG_ASSERT(instance != coord_instance->repl_instances_.end(), "Instance {} not found during callback!",
+              instance_name);
+    return *instance;
+  };
+
+  replica_succ_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
+    auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
+    spdlog::trace("Instance {} performing replica successful callback", instance_name);
+    find_instance(coord_instance, instance_name).OnSuccessPing();
+  };
+
+  replica_fail_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
+    auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
+    spdlog::trace("Instance {} performing replica failure callback", instance_name);
+    find_instance(coord_instance, instance_name).OnFailPing();
+  };
+
+  main_succ_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
+    auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
+    spdlog::trace("Instance {} performing main successful callback", instance_name);
+
+    auto &instance = find_instance(coord_instance, instance_name);
+
+    if (instance.IsAlive()) {
+      instance.OnSuccessPing();
+      return;
+    }
+
+    bool const is_latest_main = !coord_instance->ClusterHasAliveMain_();
+    if (is_latest_main) {
+      spdlog::info("Instance {} is the latest main", instance_name);
+      instance.OnSuccessPing();
+      return;
+    }
+
+    bool const demoted = instance.DemoteToReplica(coord_instance->replica_succ_cb_, coord_instance->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_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
+    auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
+    spdlog::trace("Instance {} performing main failure callback", instance_name);
+    find_instance(coord_instance, instance_name).OnFailPing();
+
+    if (!coord_instance->ClusterHasAliveMain_()) {
+      spdlog::info("Cluster without main instance, trying automatic failover");
+      coord_instance->TryFailover();
+    }
+  };
+}
+
+auto CoordinatorInstance::ClusterHasAliveMain_() const -> bool {
+  auto const alive_main = [](ReplicationInstance const &instance) { return instance.IsMain() && instance.IsAlive(); };
+  return std::ranges::any_of(repl_instances_, alive_main);
+}
+
+auto CoordinatorInstance::TryFailover() -> void {
+  auto alive_replicas = repl_instances_ | ranges::views::filter(&ReplicationInstance::IsReplica) |
+                        ranges::views::filter(&ReplicationInstance::IsAlive);
+
+  if (ranges::empty(alive_replicas)) {
+    spdlog::warn("Failover failed since all replicas are down!");
+    return;
+  }
+
+  // TODO: Smarter choice
+  auto chosen_replica_instance = ranges::begin(alive_replicas);
+
+  chosen_replica_instance->PauseFrequentCheck();
+  utils::OnScopeExit scope_exit{[&chosen_replica_instance] { chosen_replica_instance->ResumeFrequentCheck(); }};
+
+  auto const potential_new_main_uuid = utils::UUID{};
+
+  auto const is_not_chosen_replica_instance = [&chosen_replica_instance](ReplicationInstance &instance) {
+    return instance != *chosen_replica_instance;
+  };
+
+  // If for some replicas swap fails, for others on successful ping we will revert back on next change
+  // or we will do failover first again and then it will be consistent again
+  for (auto &other_replica_instance : alive_replicas | ranges::views::filter(is_not_chosen_replica_instance)) {
+    if (!other_replica_instance.SendSwapAndUpdateUUID(potential_new_main_uuid)) {
+      spdlog::error(fmt::format("Failed to swap uuid for instance {} which is alive, aborting failover",
+                                other_replica_instance.InstanceName()));
+      return;
+    }
+  }
+
+  std::vector<ReplClientInfo> repl_clients_info;
+  repl_clients_info.reserve(repl_instances_.size() - 1);
+  std::ranges::transform(repl_instances_ | ranges::views::filter(is_not_chosen_replica_instance),
+                         std::back_inserter(repl_clients_info), &ReplicationInstance::ReplicationClientInfo);
+
+  if (!chosen_replica_instance->PromoteToMain(potential_new_main_uuid, std::move(repl_clients_info), main_succ_cb_,
+                                              main_fail_cb_)) {
+    spdlog::warn("Failover failed since promoting replica to main failed!");
+    return;
+  }
+  chosen_replica_instance->SetNewMainUUID(potential_new_main_uuid);
+  main_uuid_ = potential_new_main_uuid;
+
+  spdlog::info("Failover successful! Instance {} promoted to main.", chosen_replica_instance->InstanceName());
+}
+
+auto CoordinatorInstance::ShowInstances() const -> std::vector<InstanceStatus> {
+  auto const coord_instances = self_.GetAllCoordinators();
+
+  std::vector<InstanceStatus> instances_status;
+  instances_status.reserve(repl_instances_.size() + coord_instances.size());
+
+  auto const stringify_repl_role = [](ReplicationInstance const &instance) -> std::string {
+    if (!instance.IsAlive()) return "unknown";
+    if (instance.IsMain()) return "main";
+    return "replica";
+  };
+
+  auto const repl_instance_to_status = [&stringify_repl_role](ReplicationInstance const &instance) -> InstanceStatus {
+    return {.instance_name = instance.InstanceName(),
+            .coord_socket_address = instance.SocketAddress(),
+            .cluster_role = stringify_repl_role(instance),
+            .is_alive = instance.IsAlive()};
+  };
+
+  auto const coord_instance_to_status = [](ptr<srv_config> const &instance) -> InstanceStatus {
+    return {.instance_name = "coordinator_" + std::to_string(instance->get_id()),
+            .raft_socket_address = instance->get_endpoint(),
+            .cluster_role = "coordinator",
+            .is_alive = true};  // TODO: (andi) Get this info from RAFT and test it or when we will move
+                                // CoordinatorState to every instance, we can be smarter about this using our RPC.
+  };
+
+  std::ranges::transform(coord_instances, std::back_inserter(instances_status), coord_instance_to_status);
+
+  {
+    auto lock = std::shared_lock{coord_instance_lock_};
+    std::ranges::transform(repl_instances_, std::back_inserter(instances_status), repl_instance_to_status);
+  }
+
+  return instances_status;
+}
+
+// TODO: (andi) Make sure you cannot put coordinator instance to the main
+auto CoordinatorInstance::SetReplicationInstanceToMain(std::string instance_name)
+    -> SetInstanceToMainCoordinatorStatus {
+  auto lock = std::lock_guard{coord_instance_lock_};
+
+  auto const is_new_main = [&instance_name](ReplicationInstance const &instance) {
+    return instance.InstanceName() == instance_name;
+  };
+  auto new_main = std::ranges::find_if(repl_instances_, is_new_main);
+
+  if (new_main == repl_instances_.end()) {
+    spdlog::error("Instance {} not registered. Please register it using REGISTER INSTANCE {}", instance_name,
+                  instance_name);
+    return SetInstanceToMainCoordinatorStatus::NO_INSTANCE_WITH_NAME;
+  }
+
+  new_main->PauseFrequentCheck();
+  utils::OnScopeExit scope_exit{[&new_main] { new_main->ResumeFrequentCheck(); }};
+
+  ReplicationClientsInfo repl_clients_info;
+  repl_clients_info.reserve(repl_instances_.size() - 1);
+
+  auto const is_not_new_main = [&instance_name](ReplicationInstance const &instance) {
+    return instance.InstanceName() != instance_name;
+  };
+
+  auto potential_new_main_uuid = utils::UUID{};
+  spdlog::trace("Generated potential new main uuid");
+
+  for (auto &other_instance : repl_instances_ | ranges::views::filter(is_not_new_main)) {
+    if (!other_instance.SendSwapAndUpdateUUID(potential_new_main_uuid)) {
+      spdlog::error(
+          fmt::format("Failed to swap uuid for instance {}, aborting failover", other_instance.InstanceName()));
+      return SetInstanceToMainCoordinatorStatus::SWAP_UUID_FAILED;
+    }
+  }
+
+  std::ranges::transform(repl_instances_ | ranges::views::filter(is_not_new_main),
+                         std::back_inserter(repl_clients_info),
+                         [](const ReplicationInstance &instance) { return instance.ReplicationClientInfo(); });
+
+  if (!new_main->PromoteToMain(potential_new_main_uuid, std::move(repl_clients_info), main_succ_cb_, main_fail_cb_)) {
+    return SetInstanceToMainCoordinatorStatus::COULD_NOT_PROMOTE_TO_MAIN;
+  }
+
+  new_main->SetNewMainUUID(potential_new_main_uuid);
+  main_uuid_ = potential_new_main_uuid;
+  spdlog::info("Instance {} promoted to main", instance_name);
+  return SetInstanceToMainCoordinatorStatus::SUCCESS;
+}
+
+auto CoordinatorInstance::RegisterReplicationInstance(CoordinatorClientConfig config)
+    -> RegisterInstanceCoordinatorStatus {
+  auto lock = std::lock_guard{coord_instance_lock_};
+
+  auto const name_matches = [&config](ReplicationInstance const &instance) {
+    return instance.InstanceName() == config.instance_name;
+  };
+
+  if (std::ranges::any_of(repl_instances_, name_matches)) {
+    return RegisterInstanceCoordinatorStatus::NAME_EXISTS;
+  }
+
+  auto const socket_address_matches = [&config](ReplicationInstance const &instance) {
+    return instance.SocketAddress() == config.SocketAddress();
+  };
+
+  if (std::ranges::any_of(repl_instances_, socket_address_matches)) {
+    return RegisterInstanceCoordinatorStatus::ENDPOINT_EXISTS;
+  }
+
+  if (!self_.RequestLeadership()) {
+    return RegisterInstanceCoordinatorStatus::NOT_LEADER;
+  }
+
+  auto const res = self_.AppendRegisterReplicationInstance(config.instance_name);
+  if (!res->get_accepted()) {
+    spdlog::error(
+        "Failed to accept request for registering instance {}. Most likely the reason is that the instance is not the "
+        "leader.",
+        config.instance_name);
+    return RegisterInstanceCoordinatorStatus::RAFT_COULD_NOT_ACCEPT;
+  }
+
+  spdlog::info("Request for registering instance {} accepted", config.instance_name);
+  try {
+    repl_instances_.emplace_back(this, std::move(config), replica_succ_cb_, replica_fail_cb_);
+  } catch (CoordinatorRegisterInstanceException const &) {
+    return RegisterInstanceCoordinatorStatus::RPC_FAILED;
+  }
+
+  if (res->get_result_code() != nuraft::cmd_result_code::OK) {
+    spdlog::error("Failed to register instance {} with error code {}", config.instance_name, res->get_result_code());
+    return RegisterInstanceCoordinatorStatus::RAFT_COULD_NOT_APPEND;
+  }
+
+  spdlog::info("Instance {} registered", config.instance_name);
+  return RegisterInstanceCoordinatorStatus::SUCCESS;
+}
+
+auto CoordinatorInstance::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
+    -> void {
+  self_.AddCoordinatorInstance(raft_server_id, raft_port, std::move(raft_address));
+}
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/coordinator_log_store.cpp

@@ -0,0 +1,201 @@
+// 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 "nuraft/coordinator_log_store.hpp"
+
+#include "coordination/coordinator_exceptions.hpp"
+
+namespace memgraph::coordination {
+
+using nuraft::cs_new;
+
+namespace {
+
+ptr<log_entry> MakeClone(const ptr<log_entry> &entry) {
+  return cs_new<log_entry>(entry->get_term(), buffer::clone(entry->get_buf()), entry->get_val_type(),
+                           entry->get_timestamp());
+}
+
+}  // namespace
+
+CoordinatorLogStore::CoordinatorLogStore() : start_idx_(1) {
+  ptr<buffer> buf = buffer::alloc(sizeof(uint64_t));
+  logs_[0] = cs_new<log_entry>(0, buf);
+}
+
+CoordinatorLogStore::~CoordinatorLogStore() {}
+
+auto CoordinatorLogStore::FindOrDefault_(uint64_t index) const -> ptr<log_entry> {
+  auto entry = logs_.find(index);
+  if (entry == logs_.end()) {
+    entry = logs_.find(0);
+  }
+  return entry->second;
+}
+
+uint64_t CoordinatorLogStore::next_slot() const {
+  auto lock = std::lock_guard{logs_lock_};
+  return start_idx_ + logs_.size() - 1;
+}
+
+uint64_t CoordinatorLogStore::start_index() const { return start_idx_; }
+
+ptr<log_entry> CoordinatorLogStore::last_entry() const {
+  auto lock = std::lock_guard{logs_lock_};
+
+  uint64_t const last_idx = start_idx_ + logs_.size() - 1;
+  auto const last_src = FindOrDefault_(last_idx - 1);
+
+  return MakeClone(last_src);
+}
+
+uint64_t CoordinatorLogStore::append(ptr<log_entry> &entry) {
+  ptr<log_entry> clone = MakeClone(entry);
+  uint64_t next_slot{0};
+  {
+    auto lock = std::lock_guard{logs_lock_};
+    next_slot = start_idx_ + logs_.size() - 1;
+    logs_[next_slot] = clone;
+  }
+  return next_slot;
+}
+
+void CoordinatorLogStore::write_at(uint64_t index, ptr<log_entry> &entry) {
+  ptr<log_entry> clone = MakeClone(entry);
+
+  // Discard all logs equal to or greater than `index.
+  {
+    auto lock = std::lock_guard{logs_lock_};
+    auto itr = logs_.lower_bound(index);
+    while (itr != logs_.end()) {
+      itr = logs_.erase(itr);
+    }
+    logs_[index] = clone;
+  }
+}
+
+ptr<std::vector<ptr<log_entry>>> CoordinatorLogStore::log_entries(uint64_t start, uint64_t end) {
+  auto ret = cs_new<std::vector<ptr<log_entry>>>();
+  ret->resize(end - start);
+
+  for (uint64_t i = start, curr_index = 0; i < end; ++i, ++curr_index) {
+    ptr<log_entry> src = nullptr;
+    {
+      auto lock = std::lock_guard{logs_lock_};
+      if (auto const entry = logs_.find(i); entry != logs_.end()) {
+        src = entry->second;
+      } else {
+        throw RaftCouldNotFindEntryException("Could not find entry at index {}", i);
+      }
+    }
+    (*ret)[curr_index] = MakeClone(src);
+  }
+  return ret;
+}
+
+ptr<log_entry> CoordinatorLogStore::entry_at(uint64_t index) {
+  ptr<log_entry> src = nullptr;
+  {
+    auto lock = std::lock_guard{logs_lock_};
+    src = FindOrDefault_(index);
+  }
+  return MakeClone(src);
+}
+
+uint64_t CoordinatorLogStore::term_at(uint64_t index) {
+  uint64_t term = 0;
+  {
+    auto lock = std::lock_guard{logs_lock_};
+    term = FindOrDefault_(index)->get_term();
+  }
+  return term;
+}
+
+ptr<buffer> CoordinatorLogStore::pack(uint64_t index, int32 cnt) {
+  std::vector<ptr<buffer>> logs;
+
+  size_t size_total = 0;
+  uint64_t const end_index = index + cnt;
+  for (uint64_t i = index; i < end_index; ++i) {
+    ptr<log_entry> le = nullptr;
+    {
+      auto lock = std::lock_guard{logs_lock_};
+      le = logs_[i];
+    }
+    assert(le.get());
+    auto buf = le->serialize();
+    size_total += buf->size();
+    logs.push_back(buf);
+  }
+
+  auto buf_out = buffer::alloc(sizeof(int32) + cnt * sizeof(int32) + size_total);
+  buf_out->pos(0);
+  buf_out->put((int32)cnt);
+
+  for (auto &entry : logs) {
+    auto &bb = entry;  // TODO: (andi) This smells like not needed
+    buf_out->put(static_cast<int32>(bb->size()));
+    buf_out->put(*bb);
+  }
+  return buf_out;
+}
+
+void CoordinatorLogStore::apply_pack(uint64_t index, buffer &pack) {
+  pack.pos(0);
+  int32 const num_logs = pack.get_int();
+
+  for (int32 i = 0; i < num_logs; ++i) {
+    uint64_t cur_idx = index + i;
+    int32 buf_size = pack.get_int();
+
+    ptr<buffer> buf_local = buffer::alloc(buf_size);
+    pack.get(buf_local);
+
+    ptr<log_entry> le = log_entry::deserialize(*buf_local);
+    {
+      auto lock = std::lock_guard{logs_lock_};
+      logs_[cur_idx] = le;
+    }
+  }
+
+  {
+    auto lock = std::lock_guard{logs_lock_};
+    auto const entry = logs_.upper_bound(0);
+    if (entry != logs_.end()) {
+      start_idx_ = entry->first;
+    } else {
+      start_idx_ = 1;
+    }
+  }
+}
+
+// NOTE: Remove all logs up to given 'last_log_index' (inclusive).
+bool CoordinatorLogStore::compact(uint64_t last_log_index) {
+  auto lock = std::lock_guard{logs_lock_};
+  for (uint64_t ii = start_idx_; ii <= last_log_index; ++ii) {
+    auto const entry = logs_.find(ii);
+    if (entry != logs_.end()) {
+      logs_.erase(entry);
+    }
+  }
+
+  if (start_idx_ <= last_log_index) {
+    start_idx_ = last_log_index + 1;
+  }
+  return true;
+}
+
+bool CoordinatorLogStore::flush() { return true; }
+
+}  // 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 {
 
@@ -77,26 +77,28 @@ void Load(memgraph::coordination::PromoteReplicaToMainRes *self, memgraph::slk::
 }
 
 void Save(const memgraph::coordination::PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder) {
+  memgraph::slk::Save(self.main_uuid_, builder);
   memgraph::slk::Save(self.replication_clients_info, builder);
 }
 
 void Load(memgraph::coordination::PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader) {
+  memgraph::slk::Load(&self->main_uuid_, reader);
   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_server.cpp

@@ -12,7 +12,7 @@
 #ifdef MG_ENTERPRISE
 
 #include "coordination/coordinator_server.hpp"
-#include "replication_coordination_glue/messages.hpp"
+#include "replication_coordination_glue/handler.hpp"
 
 namespace memgraph::coordination {
 

src/coordination/coordinator_state.cpp

@@ -25,7 +25,7 @@
 namespace memgraph::coordination {
 
 CoordinatorState::CoordinatorState() {
-  MG_ASSERT(!(FLAGS_coordinator && FLAGS_coordinator_server_port),
+  MG_ASSERT(!(FLAGS_raft_server_id && FLAGS_coordinator_server_port),
             "Instance cannot be a coordinator and have registered coordinator server.");
 
   spdlog::info("Executing coordinator constructor");
@@ -41,43 +41,39 @@ CoordinatorState::CoordinatorState() {
   }
 }
 
-auto CoordinatorState::RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus {
-  MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
+auto CoordinatorState::RegisterReplicationInstance(CoordinatorClientConfig config)
+    -> RegisterInstanceCoordinatorStatus {
+  MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
             "Coordinator cannot register replica since variant holds wrong alternative");
 
   return std::visit(
-      memgraph::utils::Overloaded{
-          [](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
-            return RegisterInstanceCoordinatorStatus::NOT_COORDINATOR;
-          },
-          [config](CoordinatorData &coordinator_data) { return coordinator_data.RegisterInstance(config); }},
+      memgraph::utils::Overloaded{[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
+                                    return RegisterInstanceCoordinatorStatus::NOT_COORDINATOR;
+                                  },
+                                  [config](CoordinatorInstance &coordinator_instance) {
+                                    return coordinator_instance.RegisterReplicationInstance(config);
+                                  }},
       data_);
 }
 
-auto CoordinatorState::SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus {
-  MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
+auto CoordinatorState::SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus {
+  MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
             "Coordinator cannot register replica since variant holds wrong alternative");
 
   return std::visit(
       memgraph::utils::Overloaded{[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
                                     return SetInstanceToMainCoordinatorStatus::NOT_COORDINATOR;
                                   },
-                                  [&instance_name](CoordinatorData &coordinator_data) {
-                                    return coordinator_data.SetInstanceToMain(instance_name);
+                                  [&instance_name](CoordinatorInstance &coordinator_instance) {
+                                    return coordinator_instance.SetReplicationInstanceToMain(instance_name);
                                   }},
       data_);
 }
 
-auto CoordinatorState::ShowInstances() const -> std::vector<CoordinatorInstanceStatus> {
-  MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
+auto CoordinatorState::ShowInstances() const -> std::vector<InstanceStatus> {
+  MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
             "Can't call show instances on data_, as variant holds wrong alternative");
-  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();
+  return std::get<CoordinatorInstance>(data_).ShowInstances();
 }
 
 auto CoordinatorState::GetCoordinatorServer() const -> CoordinatorServer & {
@@ -85,5 +81,13 @@ auto CoordinatorState::GetCoordinatorServer() const -> CoordinatorServer & {
             "Cannot get coordinator server since variant holds wrong alternative");
   return *std::get<CoordinatorMainReplicaData>(data_).coordinator_server_;
 }
+
+auto CoordinatorState::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
+    -> void {
+  MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
+            "Coordinator cannot register replica since variant holds wrong alternative");
+  return std::get<CoordinatorInstance>(data_).AddCoordinatorInstance(raft_server_id, raft_port, raft_address);
+}
+
 }  // namespace memgraph::coordination
 #endif

src/coordination/coordinator_state_machine.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.
+
+#ifdef MG_ENTERPRISE
+
+#include "nuraft/coordinator_state_machine.hpp"
+
+namespace memgraph::coordination {
+
+auto CoordinatorStateMachine::EncodeRegisterReplicationInstance(const std::string &name) -> ptr<buffer> {
+  std::string str_log = name + "_replica";
+  ptr<buffer> log = buffer::alloc(sizeof(uint32_t) + str_log.size());
+  buffer_serializer bs(log);
+  bs.put_str(str_log);
+  return log;
+}
+
+auto CoordinatorStateMachine::DecodeRegisterReplicationInstance(buffer &data) -> std::string {
+  buffer_serializer bs(data);
+  return bs.get_str();
+}
+
+auto CoordinatorStateMachine::pre_commit(ulong const log_idx, buffer &data) -> ptr<buffer> {
+  buffer_serializer bs(data);
+  std::string str = bs.get_str();
+
+  spdlog::info("pre_commit {} : {}", log_idx, str);
+  return nullptr;
+}
+
+auto CoordinatorStateMachine::commit(ulong const log_idx, buffer &data) -> ptr<buffer> {
+  buffer_serializer bs(data);
+  std::string str = bs.get_str();
+
+  spdlog::info("commit {} : {}", log_idx, str);
+
+  last_committed_idx_ = log_idx;
+  return nullptr;
+}
+
+auto CoordinatorStateMachine::commit_config(ulong const log_idx, ptr<cluster_config> & /*new_conf*/) -> void {
+  last_committed_idx_ = log_idx;
+}
+
+auto CoordinatorStateMachine::rollback(ulong const log_idx, buffer &data) -> void {
+  buffer_serializer bs(data);
+  std::string str = bs.get_str();
+
+  spdlog::info("rollback {} : {}", log_idx, str);
+}
+
+auto CoordinatorStateMachine::read_logical_snp_obj(snapshot & /*snapshot*/, void *& /*user_snp_ctx*/, ulong /*obj_id*/,
+                                                   ptr<buffer> &data_out, bool &is_last_obj) -> int {
+  // Put dummy data.
+  data_out = buffer::alloc(sizeof(int32));
+  buffer_serializer bs(data_out);
+  bs.put_i32(0);
+
+  is_last_obj = true;
+  return 0;
+}
+
+auto CoordinatorStateMachine::save_logical_snp_obj(snapshot &s, ulong &obj_id, buffer & /*data*/, bool /*is_first_obj*/,
+                                                   bool /*is_last_obj*/) -> void {
+  spdlog::info("save snapshot {} term {} object ID", s.get_last_log_idx(), s.get_last_log_term(), obj_id);
+  // Request next object.
+  obj_id++;
+}
+
+auto CoordinatorStateMachine::apply_snapshot(snapshot &s) -> bool {
+  spdlog::info("apply snapshot {} term {}", s.get_last_log_idx(), s.get_last_log_term());
+  {
+    auto lock = std::lock_guard{last_snapshot_lock_};
+    ptr<buffer> snp_buf = s.serialize();
+    last_snapshot_ = snapshot::deserialize(*snp_buf);
+  }
+  return true;
+}
+
+auto CoordinatorStateMachine::free_user_snp_ctx(void *&user_snp_ctx) -> void {}
+
+auto CoordinatorStateMachine::last_snapshot() -> ptr<snapshot> {
+  auto lock = std::lock_guard{last_snapshot_lock_};
+  return last_snapshot_;
+}
+
+auto CoordinatorStateMachine::last_commit_index() -> ulong { return last_committed_idx_; }
+
+auto CoordinatorStateMachine::create_snapshot(snapshot &s, async_result<bool>::handler_type &when_done) -> void {
+  spdlog::info("create snapshot {} term {}", s.get_last_log_idx(), s.get_last_log_term());
+  // Clone snapshot from `s`.
+  {
+    auto lock = std::lock_guard{last_snapshot_lock_};
+    ptr<buffer> snp_buf = s.serialize();
+    last_snapshot_ = snapshot::deserialize(*snp_buf);
+  }
+  ptr<std::exception> except(nullptr);
+  bool ret = true;
+  when_done(ret, except);
+}
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/coordinator_state_manager.cpp

@@ -0,0 +1,68 @@
+// 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 "nuraft/coordinator_state_manager.hpp"
+
+namespace memgraph::coordination {
+
+using nuraft::cluster_config;
+using nuraft::cs_new;
+using nuraft::srv_config;
+using nuraft::srv_state;
+using nuraft::state_mgr;
+
+CoordinatorStateManager::CoordinatorStateManager(int srv_id, std::string const &endpoint)
+    : my_id_(srv_id), my_endpoint_(endpoint), cur_log_store_(cs_new<CoordinatorLogStore>()) {
+  my_srv_config_ = cs_new<srv_config>(srv_id, endpoint);
+
+  // Initial cluster config: contains only one server (myself).
+  cluster_config_ = cs_new<cluster_config>();
+  cluster_config_->get_servers().push_back(my_srv_config_);
+}
+
+auto CoordinatorStateManager::load_config() -> ptr<cluster_config> {
+  // Just return in-memory data in this example.
+  // May require reading from disk here, if it has been written to disk.
+  return cluster_config_;
+}
+
+auto CoordinatorStateManager::save_config(cluster_config const &config) -> void {
+  // Just keep in memory in this example.
+  // Need to write to disk here, if want to make it durable.
+  ptr<buffer> buf = config.serialize();
+  cluster_config_ = cluster_config::deserialize(*buf);
+}
+
+auto CoordinatorStateManager::save_state(srv_state const &state) -> void {
+  // Just keep in memory in this example.
+  // Need to write to disk here, if want to make it durable.
+  ptr<buffer> buf = state.serialize();
+  saved_state_ = srv_state::deserialize(*buf);
+}
+
+auto CoordinatorStateManager::read_state() -> ptr<srv_state> {
+  // Just return in-memory data in this example.
+  // May require reading from disk here, if it has been written to disk.
+  return saved_state_;
+}
+
+auto CoordinatorStateManager::load_log_store() -> ptr<log_store> { return cur_log_store_; }
+
+auto CoordinatorStateManager::server_id() -> int32 { return my_id_; }
+
+auto CoordinatorStateManager::system_exit(int const exit_code) -> void {}
+
+auto CoordinatorStateManager::GetSrvConfig() const -> ptr<srv_config> { return my_srv_config_; }
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/include/coordination/coordinator_client.hpp

@@ -11,6 +11,7 @@
 
 #pragma once
 
+#include "utils/uuid.hpp"
 #ifdef MG_ENTERPRISE
 
 #include "coordination/coordinator_config.hpp"
@@ -19,16 +20,14 @@
 
 namespace memgraph::coordination {
 
-class CoordinatorData;
-using HealthCheckCallback = std::function<void(CoordinatorData *, std::string_view)>;
+class CoordinatorInstance;
+using HealthCheckCallback = std::function<void(CoordinatorInstance *, 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);
+  explicit CoordinatorClient(CoordinatorInstance *coord_instance, CoordinatorClientConfig config,
+                             HealthCheckCallback succ_cb, HealthCheckCallback fail_cb);
 
   ~CoordinatorClient() = default;
 
@@ -46,15 +45,17 @@ class CoordinatorClient {
   auto InstanceName() const -> std::string;
   auto SocketAddress() const -> std::string;
 
-  auto SendPromoteReplicaToMainRpc(ReplicationClientsInfo replication_clients_info) const -> bool;
+  [[nodiscard]] auto DemoteToReplica() const -> bool;
+  auto SendPromoteReplicaToMainRpc(const utils::UUID &uuid, ReplicationClientsInfo replication_clients_info) const
+      -> bool;
 
-  auto ReplicationClientInfo() const -> const ReplClientInfo &;
-  auto ResetReplicationClientInfo() -> void;
+  auto SendSwapMainUUIDRpc(const utils::UUID &uuid) const -> bool;
 
-  auto SendSetToReplicaRpc(ReplClientInfo replication_client_info) const -> bool;
+  auto ReplicationClientInfo() const -> ReplClientInfo;
 
-  auto SetSuccCallback(HealthCheckCallback succ_cb) -> void;
-  auto SetFailCallback(HealthCheckCallback fail_cb) -> void;
+  auto SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void;
+
+  auto RpcClient() -> rpc::Client & { return rpc_client_; }
 
   friend bool operator==(CoordinatorClient const &first, CoordinatorClient const &second) {
     return first.config_ == second.config_;
@@ -68,7 +69,7 @@ class CoordinatorClient {
   mutable rpc::Client rpc_client_;
 
   CoordinatorClientConfig config_;
-  CoordinatorData *coord_data_;
+  CoordinatorInstance *coord_instance_;
   HealthCheckCallback succ_cb_;
   HealthCheckCallback fail_cb_;
 };

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

@@ -1,49 +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
-
-#ifdef MG_ENTERPRISE
-
-#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 <list>
-
-namespace memgraph::coordination {
-class CoordinatorData {
- public:
-  CoordinatorData();
-
-  [[nodiscard]] auto DoFailover() -> DoFailoverStatus;
-
-  [[nodiscard]] auto RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
-  [[nodiscard]] auto SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
-
-  auto ShowInstances() const -> std::vector<CoordinatorInstanceStatus>;
-
- private:
-  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
-  std::list<CoordinatorInstance> registered_instances_;
-};
-
-struct CoordinatorMainReplicaData {
-  std::unique_ptr<CoordinatorServer> coordinator_server_;
-};
-
-}  // namespace memgraph::coordination
-#endif

src/coordination/include/coordination/coordinator_exceptions.hpp

@@ -16,16 +16,60 @@
 #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)
+};
+
+class RaftServerStartException final : public utils::BasicException {
+ public:
+  explicit RaftServerStartException(std::string_view what) noexcept : BasicException(what) {}
+
+  template <class... Args>
+  explicit RaftServerStartException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
+      : RaftServerStartException(fmt::format(fmt, std::forward<Args>(args)...)) {}
+
+  SPECIALIZE_GET_EXCEPTION_NAME(RaftServerStartException)
+};
+
+class RaftAddServerException final : public utils::BasicException {
+ public:
+  explicit RaftAddServerException(std::string_view what) noexcept : BasicException(what) {}
+
+  template <class... Args>
+  explicit RaftAddServerException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
+      : RaftAddServerException(fmt::format(fmt, std::forward<Args>(args)...)) {}
+
+  SPECIALIZE_GET_EXCEPTION_NAME(RaftAddServerException)
+};
+
+class RaftBecomeLeaderException final : public utils::BasicException {
+ public:
+  explicit RaftBecomeLeaderException(std::string_view what) noexcept : BasicException(what) {}
+
+  template <class... Args>
+  explicit RaftBecomeLeaderException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
+      : RaftBecomeLeaderException(fmt::format(fmt, std::forward<Args>(args)...)) {}
+
+  SPECIALIZE_GET_EXCEPTION_NAME(RaftBecomeLeaderException)
+};
+
+class RaftCouldNotFindEntryException final : public utils::BasicException {
+ public:
+  explicit RaftCouldNotFindEntryException(std::string_view what) noexcept : BasicException(what) {}
+
+  template <class... Args>
+  explicit RaftCouldNotFindEntryException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
+      : RaftCouldNotFindEntryException(fmt::format(fmt, std::forward<Args>(args)...)) {}
+
+  SPECIALIZE_GET_EXCEPTION_NAME(RaftCouldNotFindEntryException)
 };
 
 }  // 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 +23,16 @@ 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);
+  static void SwapMainUUIDHandler(replication::ReplicationHandler &replication_handler, slk::Reader *req_reader,
+                                  slk::Builder *res_builder);
 };
 
 }  // namespace memgraph::dbms

src/coordination/include/coordination/coordinator_instance.hpp

@@ -13,64 +13,44 @@
 
 #ifdef MG_ENTERPRISE
 
-#include "coordination/coordinator_client.hpp"
-#include "coordination/coordinator_cluster_config.hpp"
-#include "replication_coordination_glue/role.hpp"
+#include "coordination/coordinator_server.hpp"
+#include "coordination/instance_status.hpp"
+#include "coordination/raft_instance.hpp"
+#include "coordination/register_main_replica_coordinator_status.hpp"
+#include "coordination/replication_instance.hpp"
+#include "utils/rw_lock.hpp"
+#include "utils/thread_pool.hpp"
+
+#include <list>
 
 namespace memgraph::coordination {
 
-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) {}
+  CoordinatorInstance();
 
-  CoordinatorInstance(CoordinatorInstance const &other) = delete;
-  CoordinatorInstance &operator=(CoordinatorInstance const &other) = delete;
-  CoordinatorInstance(CoordinatorInstance &&other) noexcept = delete;
-  CoordinatorInstance &operator=(CoordinatorInstance &&other) noexcept = delete;
-  ~CoordinatorInstance() = default;
+  [[nodiscard]] auto RegisterReplicationInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
 
-  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(); }
+  [[nodiscard]] auto SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
 
-  auto InstanceName() const -> std::string { return client_.InstanceName(); }
-  auto SocketAddress() const -> std::string { return client_.SocketAddress(); }
-  auto IsAlive() const -> bool { return is_alive_; }
+  auto ShowInstances() const -> std::vector<InstanceStatus>;
 
-  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 TryFailover() -> void;
 
-  auto PrepareForFailover() -> void { client_.PauseFrequentCheck(); }
-  auto RestoreAfterFailedFailover() -> void { client_.ResumeFrequentCheck(); }
+  auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
 
-  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();
-  }
+ private:
+  auto ClusterHasAliveMain_() const -> bool;
 
-  CoordinatorClient client_;
-  replication_coordination_glue::ReplicationRole replication_role_;
-  std::chrono::system_clock::time_point last_response_time_{};
-  bool is_alive_{false};
+  HealthCheckCallback main_succ_cb_, main_fail_cb_, replica_succ_cb_, replica_fail_cb_;
 
-  friend bool operator==(CoordinatorInstance const &first, CoordinatorInstance const &second) {
-    return first.client_ == second.client_ && first.replication_role_ == second.replication_role_;
-  }
+  // NOTE: Must be std::list because we rely on pointer stability
+  std::list<ReplicationInstance> repl_instances_;
+  mutable utils::RWLock coord_instance_lock_{utils::RWLock::Priority::READ};
+
+  utils::UUID main_uuid_;
+
+  RaftInstance self_;
 };
 
 }  // namespace memgraph::coordination

src/coordination/include/coordination/coordinator_rpc.hpp

@@ -11,6 +11,7 @@
 
 #pragma once
 
+#include "utils/uuid.hpp"
 #ifdef MG_ENTERPRISE
 
 #include "coordination/coordinator_config.hpp"
@@ -26,10 +27,13 @@ struct PromoteReplicaToMainReq {
   static void Load(PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader);
   static void Save(const PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder);
 
-  explicit PromoteReplicaToMainReq(std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info)
-      : replication_clients_info(std::move(replication_clients_info)) {}
+  explicit PromoteReplicaToMainReq(const utils::UUID &uuid,
+                                   std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info)
+      : main_uuid_(uuid), replication_clients_info(std::move(replication_clients_info)) {}
   PromoteReplicaToMainReq() = default;
 
+  // get uuid here
+  utils::UUID main_uuid_;
   std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info;
 };
 
@@ -48,56 +52,53 @@ 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
 
 // SLK serialization declarations
 namespace memgraph::slk {
 
+// PromoteReplicaToMainRpc
 void Save(const memgraph::coordination::PromoteReplicaToMainRes &self, memgraph::slk::Builder *builder);
-
 void Load(memgraph::coordination::PromoteReplicaToMainRes *self, memgraph::slk::Reader *reader);
-
 void Save(const memgraph::coordination::PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder);
-
 void Load(memgraph::coordination::PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader);
 
-void Save(const memgraph::coordination::SetMainToReplicaRes &self, memgraph::slk::Builder *builder);
+// DemoteMainToReplicaRpc
+void Save(const memgraph::coordination::DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder);
+void Load(memgraph::coordination::DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader);
+void Save(const memgraph::coordination::DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder);
+void Load(memgraph::coordination::DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader);
 
-void Load(memgraph::coordination::SetMainToReplicaRes *self, memgraph::slk::Reader *reader);
-
-void Save(const memgraph::coordination::SetMainToReplicaReq &self, memgraph::slk::Builder *builder);
-
-void Load(memgraph::coordination::SetMainToReplicaReq *self, memgraph::slk::Reader *reader);
 
 }  // namespace memgraph::slk
 

src/coordination/include/coordination/coordinator_state.hpp

@@ -13,10 +13,9 @@
 
 #ifdef MG_ENTERPRISE
 
-#include "coordination/coordinator_data.hpp"
-#include "coordination/coordinator_instance_status.hpp"
+#include "coordination/coordinator_instance.hpp"
 #include "coordination/coordinator_server.hpp"
-#include "coordination/failover_status.hpp"
+#include "coordination/instance_status.hpp"
 #include "coordination/register_main_replica_coordinator_status.hpp"
 
 #include <variant>
@@ -28,25 +27,29 @@ 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;
 
-  [[nodiscard]] auto RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
+  [[nodiscard]] auto RegisterReplicationInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
 
-  [[nodiscard]] auto SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
+  [[nodiscard]] auto SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
 
-  auto ShowInstances() const -> std::vector<CoordinatorInstanceStatus>;
+  auto ShowInstances() const -> std::vector<InstanceStatus>;
 
-  // The client code must check that the server exists before calling this method.
+  auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
+
+  // NOTE: 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_;
+  struct CoordinatorMainReplicaData {
+    std::unique_ptr<CoordinatorServer> coordinator_server_;
+  };
+
+  std::variant<CoordinatorInstance, CoordinatorMainReplicaData> data_;
 };
 
 }  // namespace memgraph::coordination

src/coordination/include/coordination/instance_status.hpp

@@ -19,10 +19,13 @@
 
 namespace memgraph::coordination {
 
-struct CoordinatorInstanceStatus {
+// TODO: (andi) For phase IV. Some instances won't have raft_socket_address, coord_socket_address, replication_role and
+// cluster role... At the end, all instances will have everything.
+struct InstanceStatus {
   std::string instance_name;
-  std::string socket_address;
-  std::string replication_role;
+  std::string raft_socket_address;
+  std::string coord_socket_address;
+  std::string cluster_role;
   bool is_alive;
 };
 

src/coordination/include/coordination/raft_instance.hpp

@@ -0,0 +1,73 @@
+// 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
+
+#ifdef MG_ENTERPRISE
+
+#include <flags/replication.hpp>
+
+#include <libnuraft/nuraft.hxx>
+
+namespace memgraph::coordination {
+
+using BecomeLeaderCb = std::function<void()>;
+using BecomeFollowerCb = std::function<void()>;
+
+using nuraft::buffer;
+using nuraft::logger;
+using nuraft::ptr;
+using nuraft::raft_launcher;
+using nuraft::raft_server;
+using nuraft::srv_config;
+using nuraft::state_machine;
+using nuraft::state_mgr;
+using raft_result = nuraft::cmd_result<ptr<buffer>>;
+
+class RaftInstance {
+ public:
+  RaftInstance(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb);
+
+  RaftInstance(RaftInstance const &other) = delete;
+  RaftInstance &operator=(RaftInstance const &other) = delete;
+  RaftInstance(RaftInstance &&other) noexcept = delete;
+  RaftInstance &operator=(RaftInstance &&other) noexcept = delete;
+  ~RaftInstance();
+
+  auto InstanceName() const -> std::string;
+  auto RaftSocketAddress() const -> std::string;
+
+  auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
+  auto GetAllCoordinators() const -> std::vector<ptr<srv_config>>;
+
+  auto RequestLeadership() -> bool;
+  auto IsLeader() const -> bool;
+
+  auto AppendRegisterReplicationInstance(std::string const &instance) -> ptr<raft_result>;
+
+ private:
+  ptr<state_machine> state_machine_;
+  ptr<state_mgr> state_manager_;
+  ptr<raft_server> raft_server_;
+  ptr<logger> logger_;
+  raft_launcher launcher_;
+
+  // TODO: (andi) I think variables below can be abstracted
+  uint32_t raft_server_id_;
+  uint32_t raft_port_;
+  std::string raft_address_;
+
+  BecomeLeaderCb become_leader_cb_;
+  BecomeFollowerCb become_follower_cb_;
+};
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/include/coordination/register_main_replica_coordinator_status.hpp

@@ -19,10 +19,12 @@ namespace memgraph::coordination {
 
 enum class RegisterInstanceCoordinatorStatus : uint8_t {
   NAME_EXISTS,
-  END_POINT_EXISTS,
-  COULD_NOT_BE_PERSISTED,
+  ENDPOINT_EXISTS,
   NOT_COORDINATOR,
   RPC_FAILED,
+  NOT_LEADER,
+  RAFT_COULD_NOT_ACCEPT,
+  RAFT_COULD_NOT_APPEND,
   SUCCESS
 };
 
@@ -31,6 +33,7 @@ enum class SetInstanceToMainCoordinatorStatus : uint8_t {
   NOT_COORDINATOR,
   SUCCESS,
   COULD_NOT_PROMOTE_TO_MAIN,
+  SWAP_UUID_FAILED
 };
 
 }  // namespace memgraph::coordination

src/coordination/include/coordination/replication_instance.hpp

@@ -0,0 +1,86 @@
+// 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
+
+#ifdef MG_ENTERPRISE
+
+#include "coordination/coordinator_client.hpp"
+#include "coordination/coordinator_cluster_config.hpp"
+#include "coordination/coordinator_exceptions.hpp"
+#include "replication_coordination_glue/role.hpp"
+
+#include <libnuraft/nuraft.hxx>
+#include "utils/uuid.hpp"
+
+namespace memgraph::coordination {
+
+class CoordinatorInstance;
+
+class ReplicationInstance {
+ public:
+  ReplicationInstance(CoordinatorInstance *peer, CoordinatorClientConfig config, HealthCheckCallback succ_cb,
+                      HealthCheckCallback fail_cb);
+
+  ReplicationInstance(ReplicationInstance const &other) = delete;
+  ReplicationInstance &operator=(ReplicationInstance const &other) = delete;
+  ReplicationInstance(ReplicationInstance &&other) noexcept = delete;
+  ReplicationInstance &operator=(ReplicationInstance &&other) noexcept = delete;
+  ~ReplicationInstance() = default;
+
+  auto OnSuccessPing() -> void;
+  auto OnFailPing() -> bool;
+
+  auto IsAlive() const -> bool;
+
+  auto InstanceName() const -> std::string;
+  auto SocketAddress() const -> std::string;
+
+  auto IsReplica() const -> bool;
+  auto IsMain() const -> bool;
+
+  auto PromoteToMain(utils::UUID uuid, 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 StartFrequentCheck() -> void;
+  auto StopFrequentCheck() -> void;
+  auto PauseFrequentCheck() -> void;
+  auto ResumeFrequentCheck() -> void;
+
+  auto ReplicationClientInfo() const -> ReplClientInfo;
+
+  auto SendSwapAndUpdateUUID(const utils::UUID &main_uuid) -> bool;
+  auto GetClient() -> CoordinatorClient &;
+
+  void SetNewMainUUID(const std::optional<utils::UUID> &main_uuid = std::nullopt);
+  auto GetMainUUID() -> const std::optional<utils::UUID> &;
+
+ private:
+  CoordinatorClient client_;
+  replication_coordination_glue::ReplicationRole replication_role_;
+  std::chrono::system_clock::time_point last_response_time_{};
+  bool is_alive_{false};
+
+  // for replica this is main uuid of current main
+  // for "main" main this same as in CoordinatorData
+  // it is set to nullopt when replica is down
+  // TLDR; when replica is down and comes back up we reset uuid of main replica is listening to
+  // so we need to send swap uuid again
+  std::optional<utils::UUID> main_uuid_;
+
+  friend bool operator==(ReplicationInstance const &first, ReplicationInstance const &second) {
+    return first.client_ == second.client_ && first.replication_role_ == second.replication_role_;
+  }
+};
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/include/nuraft/coordinator_log_store.hpp

@@ -0,0 +1,70 @@
+// 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
+
+#ifdef MG_ENTERPRISE
+
+#include <libnuraft/nuraft.hxx>
+
+namespace memgraph::coordination {
+
+using nuraft::buffer;
+using nuraft::int32;
+using nuraft::int64;
+using nuraft::log_entry;
+using nuraft::log_store;
+using nuraft::ptr;
+using nuraft::raft_server;
+
+class CoordinatorLogStore : public log_store {
+ public:
+  CoordinatorLogStore();
+  CoordinatorLogStore(CoordinatorLogStore const &) = delete;
+  CoordinatorLogStore &operator=(CoordinatorLogStore const &) = delete;
+  CoordinatorLogStore(CoordinatorLogStore &&) = delete;
+  CoordinatorLogStore &operator=(CoordinatorLogStore &&) = delete;
+  ~CoordinatorLogStore() override;
+
+  ulong next_slot() const override;
+
+  ulong start_index() const override;
+
+  ptr<log_entry> last_entry() const override;
+
+  ulong append(ptr<log_entry> &entry) override;
+
+  void write_at(ulong index, ptr<log_entry> &entry) override;
+
+  ptr<std::vector<ptr<log_entry>>> log_entries(ulong start, ulong end) override;
+
+  ptr<log_entry> entry_at(ulong index) override;
+
+  ulong term_at(ulong index) override;
+
+  ptr<buffer> pack(ulong index, int32 cnt) override;
+
+  void apply_pack(ulong index, buffer &pack) override;
+
+  bool compact(ulong last_log_index) override;
+
+  bool flush() override;
+
+ private:
+  auto FindOrDefault_(ulong index) const -> ptr<log_entry>;
+
+  std::map<ulong, ptr<log_entry>> logs_;
+  mutable std::mutex logs_lock_;
+  std::atomic<ulong> start_idx_;
+};
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/include/nuraft/coordinator_state_machine.hpp

@@ -0,0 +1,76 @@
+// 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
+
+#ifdef MG_ENTERPRISE
+
+#include <spdlog/spdlog.h>
+#include <libnuraft/nuraft.hxx>
+
+namespace memgraph::coordination {
+
+using nuraft::async_result;
+using nuraft::buffer;
+using nuraft::buffer_serializer;
+using nuraft::cluster_config;
+using nuraft::int32;
+using nuraft::ptr;
+using nuraft::snapshot;
+using nuraft::state_machine;
+
+class CoordinatorStateMachine : public state_machine {
+ public:
+  CoordinatorStateMachine() = default;
+  CoordinatorStateMachine(CoordinatorStateMachine const &) = delete;
+  CoordinatorStateMachine &operator=(CoordinatorStateMachine const &) = delete;
+  CoordinatorStateMachine(CoordinatorStateMachine &&) = delete;
+  CoordinatorStateMachine &operator=(CoordinatorStateMachine &&) = delete;
+  ~CoordinatorStateMachine() override {}
+
+  static auto EncodeRegisterReplicationInstance(const std::string &name) -> ptr<buffer>;
+
+  static auto DecodeRegisterReplicationInstance(buffer &data) -> std::string;
+
+  auto pre_commit(ulong log_idx, buffer &data) -> ptr<buffer> override;
+
+  auto commit(ulong log_idx, buffer &data) -> ptr<buffer> override;
+
+  auto commit_config(ulong log_idx, ptr<cluster_config> & /*new_conf*/) -> void override;
+
+  auto rollback(ulong log_idx, buffer &data) -> void override;
+
+  auto read_logical_snp_obj(snapshot & /*snapshot*/, void *& /*user_snp_ctx*/, ulong /*obj_id*/, ptr<buffer> &data_out,
+                            bool &is_last_obj) -> int override;
+
+  auto save_logical_snp_obj(snapshot &s, ulong &obj_id, buffer & /*data*/, bool /*is_first_obj*/, bool /*is_last_obj*/)
+      -> void override;
+
+  auto apply_snapshot(snapshot &s) -> bool override;
+
+  auto free_user_snp_ctx(void *&user_snp_ctx) -> void override;
+
+  auto last_snapshot() -> ptr<snapshot> override;
+
+  auto last_commit_index() -> ulong override;
+
+  auto create_snapshot(snapshot &s, async_result<bool>::handler_type &when_done) -> void override;
+
+ private:
+  std::atomic<uint64_t> last_committed_idx_{0};
+
+  ptr<snapshot> last_snapshot_;
+
+  std::mutex last_snapshot_lock_;
+};
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/include/nuraft/coordinator_state_manager.hpp

@@ -0,0 +1,66 @@
+// 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
+
+#ifdef MG_ENTERPRISE
+
+#include "nuraft/coordinator_log_store.hpp"
+
+#include <spdlog/spdlog.h>
+#include <libnuraft/nuraft.hxx>
+
+namespace memgraph::coordination {
+
+using nuraft::cluster_config;
+using nuraft::cs_new;
+using nuraft::srv_config;
+using nuraft::srv_state;
+using nuraft::state_mgr;
+
+class CoordinatorStateManager : public state_mgr {
+ public:
+  explicit CoordinatorStateManager(int srv_id, std::string const &endpoint);
+
+  CoordinatorStateManager(CoordinatorStateManager const &) = delete;
+  CoordinatorStateManager &operator=(CoordinatorStateManager const &) = delete;
+  CoordinatorStateManager(CoordinatorStateManager &&) = delete;
+  CoordinatorStateManager &operator=(CoordinatorStateManager &&) = delete;
+
+  ~CoordinatorStateManager() override = default;
+
+  auto load_config() -> ptr<cluster_config> override;
+
+  auto save_config(cluster_config const &config) -> void override;
+
+  auto save_state(srv_state const &state) -> void override;
+
+  auto read_state() -> ptr<srv_state> override;
+
+  auto load_log_store() -> ptr<log_store> override;
+
+  auto server_id() -> int32 override;
+
+  auto system_exit(int exit_code) -> void override;
+
+  auto GetSrvConfig() const -> ptr<srv_config>;
+
+ private:
+  int my_id_;
+  std::string my_endpoint_;
+  ptr<CoordinatorLogStore> cur_log_store_;
+  ptr<srv_config> my_srv_config_;
+  ptr<cluster_config> cluster_config_;
+  ptr<srv_state> saved_state_;
+};
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/raft_instance.cpp

@@ -0,0 +1,126 @@
+// 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/raft_instance.hpp"
+
+#include "coordination/coordinator_exceptions.hpp"
+#include "nuraft/coordinator_state_machine.hpp"
+#include "nuraft/coordinator_state_manager.hpp"
+#include "utils/counter.hpp"
+
+namespace memgraph::coordination {
+
+using nuraft::asio_service;
+using nuraft::cb_func;
+using nuraft::CbReturnCode;
+using nuraft::cmd_result;
+using nuraft::cs_new;
+using nuraft::ptr;
+using nuraft::raft_params;
+using nuraft::raft_server;
+using nuraft::srv_config;
+using raft_result = cmd_result<ptr<buffer>>;
+
+RaftInstance::RaftInstance(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb)
+    : raft_server_id_(FLAGS_raft_server_id),
+      raft_port_(FLAGS_raft_server_port),
+      raft_address_("127.0.0.1"),
+      become_leader_cb_(std::move(become_leader_cb)),
+      become_follower_cb_(std::move(become_follower_cb)) {
+  auto raft_endpoint = raft_address_ + ":" + std::to_string(raft_port_);
+  state_manager_ = cs_new<CoordinatorStateManager>(raft_server_id_, raft_endpoint);
+  state_machine_ = cs_new<CoordinatorStateMachine>();
+  logger_ = nullptr;
+
+  // TODO: (andi) Maybe params file
+
+  asio_service::options asio_opts;
+  asio_opts.thread_pool_size_ = 1;  // TODO: (andi) Improve this
+
+  raft_params params;
+  params.heart_beat_interval_ = 100;
+  params.election_timeout_lower_bound_ = 200;
+  params.election_timeout_upper_bound_ = 400;
+  // 5 logs are preserved before the last snapshot
+  params.reserved_log_items_ = 5;
+  // Create snapshot for every 5 log appends
+  params.snapshot_distance_ = 5;
+  params.client_req_timeout_ = 3000;
+  params.return_method_ = raft_params::blocking;
+
+  raft_server::init_options init_opts;
+  init_opts.raft_callback_ = [this](cb_func::Type event_type, cb_func::Param *param) -> nuraft::CbReturnCode {
+    if (event_type == cb_func::BecomeLeader) {
+      spdlog::info("Node {} became leader", param->leaderId);
+      become_leader_cb_();
+    } else if (event_type == cb_func::BecomeFollower) {
+      spdlog::info("Node {} became follower", param->myId);
+      become_follower_cb_();
+    }
+    return CbReturnCode::Ok;
+  };
+
+  raft_server_ = launcher_.init(state_machine_, state_manager_, logger_, static_cast<int>(raft_port_), asio_opts,
+                                params, init_opts);
+
+  if (!raft_server_) {
+    throw RaftServerStartException("Failed to launch raft server on {}", raft_endpoint);
+  }
+
+  auto maybe_stop = utils::ResettableCounter<20>();
+  while (!raft_server_->is_initialized() && !maybe_stop()) {
+    std::this_thread::sleep_for(std::chrono::milliseconds(250));
+  }
+
+  if (!raft_server_->is_initialized()) {
+    throw RaftServerStartException("Failed to initialize raft server on {}", raft_endpoint);
+  }
+
+  spdlog::info("Raft server started on {}", raft_endpoint);
+}
+
+RaftInstance::~RaftInstance() { launcher_.shutdown(); }
+
+auto RaftInstance::InstanceName() const -> std::string { return "coordinator_" + std::to_string(raft_server_id_); }
+
+auto RaftInstance::RaftSocketAddress() const -> std::string { return raft_address_ + ":" + std::to_string(raft_port_); }
+
+auto RaftInstance::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
+    -> void {
+  auto const endpoint = raft_address + ":" + std::to_string(raft_port);
+  srv_config const srv_config_to_add(static_cast<int>(raft_server_id), endpoint);
+  if (!raft_server_->add_srv(srv_config_to_add)->get_accepted()) {
+    throw RaftAddServerException("Failed to add server {} to the cluster", endpoint);
+  }
+  spdlog::info("Request to add server {} to the cluster accepted", endpoint);
+}
+
+auto RaftInstance::GetAllCoordinators() const -> std::vector<ptr<srv_config>> {
+  std::vector<ptr<srv_config>> all_srv_configs;
+  raft_server_->get_srv_config_all(all_srv_configs);
+  return all_srv_configs;
+}
+
+auto RaftInstance::IsLeader() const -> bool { return raft_server_->is_leader(); }
+
+auto RaftInstance::RequestLeadership() -> bool {
+  return raft_server_->is_leader() || raft_server_->request_leadership();
+}
+
+auto RaftInstance::AppendRegisterReplicationInstance(std::string const &instance) -> ptr<raft_result> {
+  auto new_log = CoordinatorStateMachine::EncodeRegisterReplicationInstance(instance);
+  return raft_server_->append_entries({new_log});
+}
+
+}  // namespace memgraph::coordination
+#endif

src/coordination/replication_instance.cpp

@@ -0,0 +1,100 @@
+// 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/replication_instance.hpp"
+
+#include "replication_coordination_glue/handler.hpp"
+
+namespace memgraph::coordination {
+
+ReplicationInstance::ReplicationInstance(CoordinatorInstance *peer, CoordinatorClientConfig config,
+                                         HealthCheckCallback succ_cb, HealthCheckCallback fail_cb)
+    : client_(peer, std::move(config), std::move(succ_cb), std::move(fail_cb)),
+      replication_role_(replication_coordination_glue::ReplicationRole::REPLICA) {
+  if (!client_.DemoteToReplica()) {
+    throw CoordinatorRegisterInstanceException("Failed to demote instance {} to replica", client_.InstanceName());
+  }
+
+  client_.StartFrequentCheck();
+}
+
+auto ReplicationInstance::OnSuccessPing() -> void {
+  last_response_time_ = std::chrono::system_clock::now();
+  is_alive_ = true;
+}
+
+auto ReplicationInstance::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 ReplicationInstance::InstanceName() const -> std::string { return client_.InstanceName(); }
+auto ReplicationInstance::SocketAddress() const -> std::string { return client_.SocketAddress(); }
+auto ReplicationInstance::IsAlive() const -> bool { return is_alive_; }
+
+auto ReplicationInstance::IsReplica() const -> bool {
+  return replication_role_ == replication_coordination_glue::ReplicationRole::REPLICA;
+}
+auto ReplicationInstance::IsMain() const -> bool {
+  return replication_role_ == replication_coordination_glue::ReplicationRole::MAIN;
+}
+
+auto ReplicationInstance::PromoteToMain(utils::UUID uuid, ReplicationClientsInfo repl_clients_info,
+                                        HealthCheckCallback main_succ_cb, HealthCheckCallback main_fail_cb) -> bool {
+  if (!client_.SendPromoteReplicaToMainRpc(uuid, 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 ReplicationInstance::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 ReplicationInstance::StartFrequentCheck() -> void { client_.StartFrequentCheck(); }
+auto ReplicationInstance::StopFrequentCheck() -> void { client_.StopFrequentCheck(); }
+auto ReplicationInstance::PauseFrequentCheck() -> void { client_.PauseFrequentCheck(); }
+auto ReplicationInstance::ResumeFrequentCheck() -> void { client_.ResumeFrequentCheck(); }
+
+auto ReplicationInstance::ReplicationClientInfo() const -> CoordinatorClientConfig::ReplicationClientInfo {
+  return client_.ReplicationClientInfo();
+}
+
+auto ReplicationInstance::GetClient() -> CoordinatorClient & { return client_; }
+void ReplicationInstance::SetNewMainUUID(const std::optional<utils::UUID> &main_uuid) { main_uuid_ = main_uuid; }
+auto ReplicationInstance::GetMainUUID() -> const std::optional<utils::UUID> & { return main_uuid_; }
+
+auto ReplicationInstance::SendSwapAndUpdateUUID(const utils::UUID &main_uuid) -> bool {
+  if (!replication_coordination_glue::SendSwapMainUUIDRpc(client_.RpcClient(), main_uuid)) {
+    return false;
+  }
+  SetNewMainUUID(main_uuid_);
+  return true;
+}
+
+}  // namespace memgraph::coordination
+#endif

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

@@ -9,30 +9,36 @@
 // by the Apache License, Version 2.0, included in the file
 // licenses/APL.txt.
 
-#include "coordination/register_main_replica_coordinator_status.hpp"
 #ifdef MG_ENTERPRISE
 
 #include "dbms/coordinator_handler.hpp"
 
-#include "dbms/dbms_handler.hpp"
+#include "coordination/register_main_replica_coordinator_status.hpp"
 
 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)
+auto CoordinatorHandler::RegisterReplicationInstance(memgraph::coordination::CoordinatorClientConfig config)
     -> coordination::RegisterInstanceCoordinatorStatus {
-  return dbms_handler_.CoordinatorState().RegisterInstance(config);
+  return coordinator_state_.RegisterReplicationInstance(config);
 }
 
-auto CoordinatorHandler::SetInstanceToMain(std::string instance_name)
+auto CoordinatorHandler::SetReplicationInstanceToMain(std::string instance_name)
     -> coordination::SetInstanceToMainCoordinatorStatus {
-  return dbms_handler_.CoordinatorState().SetInstanceToMain(std::move(instance_name));
+  return coordinator_state_.SetReplicationInstanceToMain(std::move(instance_name));
 }
 
-auto CoordinatorHandler::ShowInstances() const -> std::vector<coordination::CoordinatorInstanceStatus> {
-  return dbms_handler_.CoordinatorState().ShowInstances();
+auto CoordinatorHandler::ShowInstances() const -> std::vector<coordination::InstanceStatus> {
+  return coordinator_state_.ShowInstances();
 }
+
+auto CoordinatorHandler::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
+    -> void {
+  coordinator_state_.AddCoordinatorInstance(raft_server_id, raft_port, std::move(raft_address));
+}
+
 }  // namespace memgraph::dbms
 
 #endif

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/instance_status.hpp"
 #include "coordination/register_main_replica_coordinator_status.hpp"
 
-#include <cstdint>
-#include <optional>
 #include <vector>
 
 namespace memgraph::dbms {
@@ -30,17 +26,19 @@ class DbmsHandler;
 
 class CoordinatorHandler {
  public:
-  explicit CoordinatorHandler(DbmsHandler &dbms_handler);
+  explicit CoordinatorHandler(coordination::CoordinatorState &coordinator_state);
 
-  auto RegisterInstance(coordination::CoordinatorClientConfig config)
+  auto RegisterReplicationInstance(coordination::CoordinatorClientConfig config)
       -> coordination::RegisterInstanceCoordinatorStatus;
 
-  auto SetInstanceToMain(std::string instance_name) -> coordination::SetInstanceToMainCoordinatorStatus;
+  auto SetReplicationInstanceToMain(std::string instance_name) -> coordination::SetInstanceToMainCoordinatorStatus;
 
-  auto ShowInstances() const -> std::vector<coordination::CoordinatorInstanceStatus>;
+  auto ShowInstances() const -> std::vector<coordination::InstanceStatus>;
+
+  auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
 
  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 +11,75 @@
 
 #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";
+    case ERROR_ACCEPTING_MAIN:
+      return "ERROR_ACCEPTING_MAIN";
+  }
+}
+
+// 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, mainData.uuid_);
+          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 +158,9 @@ 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 +194,13 @@ 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 +210,6 @@ DbmsHandler::DbmsHandler(
       locked_auth->DeleteDatabase(name);
       durability_->Delete(key);
     }
-    // Delete the last timestamp
-    durability_->Delete(kLastCommitedSystemTsKey);
   }
 
   /*
@@ -198,45 +234,31 @@ 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, const utils::UUID &main_uuid,
+                     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, main_uuid, std::string(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 +295,10 @@ 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 +319,50 @@ 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, const utils::UUID &main_uuid,
+                     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, main_uuid, std::string(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 +416,16 @@ 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 +25,25 @@
 #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 "replication_coordination_glue/handler.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 +52,6 @@
 
 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 +107,8 @@ 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 +116,14 @@ 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 +133,10 @@ 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 +229,7 @@ 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.
@@ -274,7 +269,7 @@ class DbmsHandler {
   bool IsReplica() const { return repl_state_.IsReplica(); }
 
 #ifdef MG_ENTERPRISE
-  coordination::CoordinatorState &CoordinatorState() { return coordinator_state_; }
+  // coordination::CoordinatorState &CoordinatorState() { return coordinator_state_; }
 #endif
 
   /**
@@ -282,8 +277,7 @@ class DbmsHandler {
    *
    * @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 +313,7 @@ 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 +400,17 @@ 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 +436,8 @@ 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 +448,7 @@ 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 +457,11 @@ 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 +470,7 @@ 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 +485,8 @@ 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 +573,24 @@ 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/replication_handlers.cpp

@@ -76,47 +76,84 @@ std::optional<DatabaseAccess> GetDatabaseAccessor(dbms::DbmsHandler *dbms_handle
     return std::nullopt;
   }
 }
+
+void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
+                  std::string_view rpc_req) {
+  spdlog::error("Received {} with main_id: {} != current_main_uuid: {}", rpc_req, std::string(main_req_id),
+                current_main_uuid.has_value() ? std::string(current_main_uuid.value()) : "");
+}
 }  // namespace
 
-void InMemoryReplicationHandlers::Register(dbms::DbmsHandler *dbms_handler, replication::ReplicationServer &server) {
-  server.rpc_server_.Register<storage::replication::HeartbeatRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
-    spdlog::debug("Received HeartbeatRpc");
-    InMemoryReplicationHandlers::HeartbeatHandler(dbms_handler, req_reader, res_builder);
-  });
-  server.rpc_server_.Register<storage::replication::AppendDeltasRpc>(
-      [dbms_handler](auto *req_reader, auto *res_builder) {
-        spdlog::debug("Received AppendDeltasRpc");
-        InMemoryReplicationHandlers::AppendDeltasHandler(dbms_handler, req_reader, res_builder);
+void InMemoryReplicationHandlers::Register(dbms::DbmsHandler *dbms_handler, replication::RoleReplicaData &data) {
+  auto &server = *data.server;
+  server.rpc_server_.Register<storage::replication::HeartbeatRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received HeartbeatRpc");
+        InMemoryReplicationHandlers::HeartbeatHandler(dbms_handler, data.uuid_, req_reader, res_builder);
+      });
+  server.rpc_server_.Register<storage::replication::AppendDeltasRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received AppendDeltasRpc");
+        InMemoryReplicationHandlers::AppendDeltasHandler(dbms_handler, data.uuid_, req_reader, res_builder);
+      });
+  server.rpc_server_.Register<storage::replication::SnapshotRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received SnapshotRpc");
+        InMemoryReplicationHandlers::SnapshotHandler(dbms_handler, data.uuid_, req_reader, res_builder);
+      });
+  server.rpc_server_.Register<storage::replication::WalFilesRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received WalFilesRpc");
+        InMemoryReplicationHandlers::WalFilesHandler(dbms_handler, data.uuid_, req_reader, res_builder);
+      });
+  server.rpc_server_.Register<storage::replication::CurrentWalRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received CurrentWalRpc");
+        InMemoryReplicationHandlers::CurrentWalHandler(dbms_handler, data.uuid_, req_reader, res_builder);
+      });
+  server.rpc_server_.Register<storage::replication::TimestampRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received TimestampRpc");
+        InMemoryReplicationHandlers::TimestampHandler(dbms_handler, data.uuid_, req_reader, res_builder);
+      });
+  server.rpc_server_.Register<replication_coordination_glue::SwapMainUUIDRpc>(
+      [&data, dbms_handler](auto *req_reader, auto *res_builder) {
+        spdlog::debug("Received SwapMainUUIDHandler");
+        InMemoryReplicationHandlers::SwapMainUUIDHandler(dbms_handler, data, req_reader, res_builder);
       });
-  server.rpc_server_.Register<storage::replication::SnapshotRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
-    spdlog::debug("Received SnapshotRpc");
-    InMemoryReplicationHandlers::SnapshotHandler(dbms_handler, req_reader, res_builder);
-  });
-  server.rpc_server_.Register<storage::replication::WalFilesRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
-    spdlog::debug("Received WalFilesRpc");
-    InMemoryReplicationHandlers::WalFilesHandler(dbms_handler, req_reader, res_builder);
-  });
-  server.rpc_server_.Register<storage::replication::CurrentWalRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
-    spdlog::debug("Received CurrentWalRpc");
-    InMemoryReplicationHandlers::CurrentWalHandler(dbms_handler, req_reader, res_builder);
-  });
-  server.rpc_server_.Register<storage::replication::TimestampRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
-    spdlog::debug("Received TimestampRpc");
-    InMemoryReplicationHandlers::TimestampHandler(dbms_handler, req_reader, res_builder);
-  });
 }
 
-void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
-                                                   slk::Builder *res_builder) {
+void InMemoryReplicationHandlers::SwapMainUUIDHandler(dbms::DbmsHandler *dbms_handler,
+                                                      replication::RoleReplicaData &role_replica_data,
+                                                      slk::Reader *req_reader, slk::Builder *res_builder) {
+  if (!dbms_handler->IsReplica()) {
+    spdlog::error("Setting main uuid must be performed on replica.");
+    slk::Save(replication_coordination_glue::SwapMainUUIDRes{false}, res_builder);
+    return;
+  }
+
+  replication_coordination_glue::SwapMainUUIDReq req;
+  slk::Load(&req, req_reader);
+  spdlog::info(fmt::format("Set replica data UUID  to main uuid {}", std::string(req.uuid)));
+  dbms_handler->ReplicationState().TryPersistRoleReplica(role_replica_data.config, req.uuid);
+  role_replica_data.uuid_ = req.uuid;
+
+  slk::Save(replication_coordination_glue::SwapMainUUIDRes{true}, res_builder);
+}
+
+void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handler,
+                                                   const std::optional<utils::UUID> &current_main_uuid,
+                                                   slk::Reader *req_reader, slk::Builder *res_builder) {
   storage::replication::HeartbeatReq req;
   slk::Load(&req, req_reader);
   auto const db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
-  if (!db_acc) {
+
+  if (!current_main_uuid.has_value() || req.main_uuid != *current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::HeartbeatReq::kType.name);
     storage::replication::HeartbeatRes res{false, 0, ""};
     slk::Save(res, res_builder);
     return;
   }
-
   // TODO: this handler is agnostic of InMemory, move to be reused by on-disk
   auto const *storage = db_acc->get()->storage();
   storage::replication::HeartbeatRes res{true, storage->repl_storage_state_.last_commit_timestamp_.load(),
@@ -124,10 +161,19 @@ void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handl
   slk::Save(res, res_builder);
 }
 
-void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
-                                                      slk::Builder *res_builder) {
+void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_handler,
+                                                      const std::optional<utils::UUID> &current_main_uuid,
+                                                      slk::Reader *req_reader, slk::Builder *res_builder) {
   storage::replication::AppendDeltasReq req;
   slk::Load(&req, req_reader);
+
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::AppendDeltasReq::kType.name);
+    storage::replication::AppendDeltasRes res{false, 0};
+    slk::Save(res, res_builder);
+    return;
+  }
+
   auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
   if (!db_acc) {
     storage::replication::AppendDeltasRes res{false, 0};
@@ -187,8 +233,9 @@ void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_ha
   spdlog::debug("Replication recovery from append deltas finished, replica is now up to date!");
 }
 
-void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
-                                                  slk::Builder *res_builder) {
+void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handler,
+                                                  const std::optional<utils::UUID> &current_main_uuid,
+                                                  slk::Reader *req_reader, slk::Builder *res_builder) {
   storage::replication::SnapshotReq req;
   slk::Load(&req, req_reader);
   auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -197,6 +244,12 @@ void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handle
     slk::Save(res, res_builder);
     return;
   }
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::SnapshotReq::kType.name);
+    storage::replication::SnapshotRes res{false, 0};
+    slk::Save(res, res_builder);
+    return;
+  }
 
   storage::replication::Decoder decoder(req_reader);
 
@@ -270,8 +323,9 @@ void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handle
   spdlog::debug("Replication recovery from snapshot finished!");
 }
 
-void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
-                                                  slk::Builder *res_builder) {
+void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handler,
+                                                  const std::optional<utils::UUID> &current_main_uuid,
+                                                  slk::Reader *req_reader, slk::Builder *res_builder) {
   storage::replication::WalFilesReq req;
   slk::Load(&req, req_reader);
   auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -280,6 +334,12 @@ void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handle
     slk::Save(res, res_builder);
     return;
   }
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::WalFilesReq::kType.name);
+    storage::replication::WalFilesRes res{false, 0};
+    slk::Save(res, res_builder);
+    return;
+  }
 
   const auto wal_file_number = req.file_number;
   spdlog::debug("Received WAL files: {}", wal_file_number);
@@ -298,8 +358,9 @@ void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handle
   spdlog::debug("Replication recovery from WAL files ended successfully, replica is now up to date!");
 }
 
-void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
-                                                    slk::Builder *res_builder) {
+void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_handler,
+                                                    const std::optional<utils::UUID> &current_main_uuid,
+                                                    slk::Reader *req_reader, slk::Builder *res_builder) {
   storage::replication::CurrentWalReq req;
   slk::Load(&req, req_reader);
   auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -309,6 +370,13 @@ void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_hand
     return;
   }
 
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::CurrentWalReq::kType.name);
+    storage::replication::CurrentWalRes res{false, 0};
+    slk::Save(res, res_builder);
+    return;
+  }
+
   storage::replication::Decoder decoder(req_reader);
 
   auto *storage = static_cast<storage::InMemoryStorage *>(db_acc->get()->storage());
@@ -370,8 +438,9 @@ void InMemoryReplicationHandlers::LoadWal(storage::InMemoryStorage *storage, sto
   }
 }
 
-void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
-                                                   slk::Builder *res_builder) {
+void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handler,
+                                                   const std::optional<utils::UUID> &current_main_uuid,
+                                                   slk::Reader *req_reader, slk::Builder *res_builder) {
   storage::replication::TimestampReq req;
   slk::Load(&req, req_reader);
   auto const db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -381,12 +450,20 @@ void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handl
     return;
   }
 
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::TimestampReq::kType.name);
+    storage::replication::CurrentWalRes res{false, 0};
+    slk::Save(res, res_builder);
+    return;
+  }
+
   // TODO: this handler is agnostic of InMemory, move to be reused by on-disk
   auto const *storage = db_acc->get()->storage();
   storage::replication::TimestampRes res{true, storage->repl_storage_state_.last_commit_timestamp_.load()};
   slk::Save(res, res_builder);
 }
 
+/////// AF how does this work, does it get all deltas at once or what?
 uint64_t InMemoryReplicationHandlers::ReadAndApplyDelta(storage::InMemoryStorage *storage,
                                                         storage::durability::BaseDecoder *decoder,
                                                         const uint64_t version) {

src/dbms/inmemory/replication_handlers.hpp

@@ -12,6 +12,7 @@
 #pragma once
 
 #include "replication/replication_server.hpp"
+#include "replication/state.hpp"
 #include "storage/v2/replication/serialization.hpp"
 
 namespace memgraph::storage {
@@ -23,21 +24,30 @@ class DbmsHandler;
 
 class InMemoryReplicationHandlers {
  public:
-  static void Register(dbms::DbmsHandler *dbms_handler, replication::ReplicationServer &server);
+  static void Register(dbms::DbmsHandler *dbms_handler, replication::RoleReplicaData &data);
 
  private:
   // RPC handlers
-  static void HeartbeatHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void HeartbeatHandler(dbms::DbmsHandler *dbms_handler, const std::optional<utils::UUID> &current_main_uuid,
+                               slk::Reader *req_reader, slk::Builder *res_builder);
 
-  static void AppendDeltasHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void AppendDeltasHandler(dbms::DbmsHandler *dbms_handler, const std::optional<utils::UUID> &current_main_uuid,
+                                  slk::Reader *req_reader, slk::Builder *res_builder);
 
-  static void SnapshotHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void SnapshotHandler(dbms::DbmsHandler *dbms_handler, const std::optional<utils::UUID> &current_main_uuid,
+                              slk::Reader *req_reader, slk::Builder *res_builder);
 
-  static void WalFilesHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void WalFilesHandler(dbms::DbmsHandler *dbms_handler, const std::optional<utils::UUID> &current_main_uuid,
+                              slk::Reader *req_reader, slk::Builder *res_builder);
 
-  static void CurrentWalHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void CurrentWalHandler(dbms::DbmsHandler *dbms_handler, const std::optional<utils::UUID> &current_main_uuid,
+                                slk::Reader *req_reader, slk::Builder *res_builder);
 
-  static void TimestampHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
+  static void TimestampHandler(dbms::DbmsHandler *dbms_handler, const std::optional<utils::UUID> &current_main_uuid,
+                               slk::Reader *req_reader, slk::Builder *res_builder);
+
+  static void SwapMainUUIDHandler(dbms::DbmsHandler *dbms_handler, replication::RoleReplicaData &role_replica_data,
+                                  slk::Reader *req_reader, slk::Builder *res_builder);
 
   static void LoadWal(storage::InMemoryStorage *storage, storage::replication::Decoder *decoder);
 

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 +1,205 @@
+// 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,
+                           const std::optional<utils::UUID> &current_main_uuid, 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);
+
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, memgraph::storage::replication::CreateDatabaseReq::kType.name);
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  // 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,
+                         const std::optional<utils::UUID> &current_main_uuid, 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);
+
+  if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
+    LogWrongMain(current_main_uuid, req.main_uuid, memgraph::storage::replication::DropDatabaseReq::kType.name);
+    memgraph::slk::Save(res, res_builder);
+    return;
+  }
+
+  // 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>(
+      [&data, system_state_access, &dbms_handler](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received CreateDatabaseRpc");
+        CreateDatabaseHandler(system_state_access, data.uuid_, dbms_handler, req_reader, res_builder);
+      });
+  data.server->rpc_server_.Register<storage::replication::DropDatabaseRpc>(
+      [&data, system_state_access, &dbms_handler](auto *req_reader, auto *res_builder) mutable {
+        spdlog::debug("Received DropDatabaseRpc");
+        DropDatabaseHandler(system_state_access, data.uuid_, dbms_handler, req_reader, res_builder);
+      });
+}
+#endif
+}  // namespace memgraph::dbms

src/dbms/replication_handlers.hpp

@@ -0,0 +1,42 @@
+// 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
+
+inline void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
+                         std::string_view rpc_req) {
+  spdlog::error("Received {} with main_id: {} != current_main_uuid: {}", rpc_req, std::string(main_req_id),
+                current_main_uuid.has_value() ? std::string(current_main_uuid.value()) : "");
+}
+
+// RPC handlers
+void CreateDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                           const std::optional<utils::UUID> &current_main_uuid, DbmsHandler &dbms_handler,
+                           slk::Reader *req_reader, slk::Builder *res_builder);
+void DropDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
+                         const std::optional<utils::UUID> &current_main_uuid, 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 +1,122 @@
+// 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(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_group_timestamp,
+                    uint64_t new_group_timestamp, storage::SalientConfig config)
+      : main_uuid(main_uuid),
+        epoch_id(std::move(epoch_id)),
+        expected_group_timestamp{expected_group_timestamp},
+        new_group_timestamp(new_group_timestamp),
+        config(std::move(config)) {}
+
+  utils::UUID main_uuid;
+  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(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_group_timestamp,
+                  uint64_t new_group_timestamp, const utils::UUID &uuid)
+      : main_uuid(main_uuid),
+        epoch_id(std::move(epoch_id)),
+        expected_group_timestamp{expected_group_timestamp},
+        new_group_timestamp(new_group_timestamp),
+        uuid(uuid) {}
+
+  utils::UUID main_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/flags/replication.cpp

@@ -13,9 +13,11 @@
 
 #ifdef MG_ENTERPRISE
 // NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
-DEFINE_bool(coordinator, false, "Controls whether the instance is a replication coordinator.");
-// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
 DEFINE_uint32(coordinator_server_port, 0, "Port on which coordinator servers will be started.");
+// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
+DEFINE_uint32(raft_server_port, 0, "Port on which raft servers will be started.");
+// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
+DEFINE_uint32(raft_server_id, 0, "Unique ID of the raft server.");
 #endif
 
 // NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)

src/flags/replication.hpp

@@ -15,9 +15,11 @@
 
 #ifdef MG_ENTERPRISE
 // NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
-DECLARE_bool(coordinator);
-// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
 DECLARE_uint32(coordinator_server_port);
+// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
+DECLARE_uint32(raft_server_port);
+// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
+DECLARE_uint32(raft_server_id);
 #endif
 
 // NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)

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/io/network/endpoint.cpp

@@ -39,7 +39,7 @@ Endpoint::IpFamily Endpoint::GetIpFamily(const std::string &address) {
 }
 
 std::optional<std::pair<std::string, uint16_t>> Endpoint::ParseSocketOrIpAddress(
-    const std::string &address, const std::optional<uint16_t> default_port = {}) {
+    const std::string &address, const std::optional<uint16_t> default_port) {
   /// expected address format:
   ///   - "ip_address:port_number"
   ///   - "ip_address"

src/io/network/endpoint.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
@@ -61,8 +61,8 @@ struct Endpoint {
    * it into an ip address and a port number; even if a default port is given,
    * it won't be used, as we expect that it is given in the address string.
    */
-  static std::optional<std::pair<std::string, uint16_t>> ParseSocketOrIpAddress(const std::string &address,
-                                                                                std::optional<uint16_t> default_port);
+  static std::optional<std::pair<std::string, uint16_t>> ParseSocketOrIpAddress(
+      const std::string &address, std::optional<uint16_t> default_port = {});
 
   /**
    * Tries to parse given string as either socket address or hostname.

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/ast/ast.hpp

@@ -3034,7 +3034,7 @@ class ReplicationQuery : public memgraph::query::Query {
 
   enum class SyncMode { SYNC, ASYNC };
 
-  enum class ReplicaState { READY, REPLICATING, RECOVERY, MAYBE_BEHIND };
+  enum class ReplicaState { READY, REPLICATING, RECOVERY, MAYBE_BEHIND, UNREACHABLE };
 
   ReplicationQuery() = default;
 
@@ -3071,11 +3071,7 @@ class CoordinatorQuery : public memgraph::query::Query {
   static const utils::TypeInfo kType;
   const utils::TypeInfo &GetTypeInfo() const override { return kType; }
 
-  enum class Action {
-    REGISTER_INSTANCE,
-    SET_INSTANCE_TO_MAIN,
-    SHOW_REPLICATION_CLUSTER,
-  };
+  enum class Action { REGISTER_INSTANCE, SET_INSTANCE_TO_MAIN, SHOW_INSTANCES, ADD_COORDINATOR_INSTANCE };
 
   enum class SyncMode { SYNC, ASYNC };
 
@@ -3087,6 +3083,8 @@ class CoordinatorQuery : public memgraph::query::Query {
   std::string instance_name_;
   memgraph::query::Expression *replication_socket_address_{nullptr};
   memgraph::query::Expression *coordinator_socket_address_{nullptr};
+  memgraph::query::Expression *raft_socket_address_{nullptr};
+  memgraph::query::Expression *raft_server_id_{nullptr};
   memgraph::query::CoordinatorQuery::SyncMode sync_mode_;
 
   CoordinatorQuery *Clone(AstStorage *storage) const override {
@@ -3098,6 +3096,8 @@ class CoordinatorQuery : public memgraph::query::Query {
     object->sync_mode_ = sync_mode_;
     object->coordinator_socket_address_ =
         coordinator_socket_address_ ? coordinator_socket_address_->Clone(storage) : nullptr;
+    object->raft_socket_address_ = raft_socket_address_ ? raft_socket_address_->Clone(storage) : nullptr;
+    object->raft_server_id_ = raft_server_id_ ? raft_server_id_->Clone(storage) : nullptr;
 
     return object;
   }

src/query/frontend/ast/cypher_main_visitor.cpp

@@ -374,7 +374,6 @@ antlrcpp::Any CypherMainVisitor::visitRegisterReplica(MemgraphCypher::RegisterRe
   return replication_query;
 }
 
-// License check is done in the interpreter.
 antlrcpp::Any CypherMainVisitor::visitRegisterInstanceOnCoordinator(
     MemgraphCypher::RegisterInstanceOnCoordinatorContext *ctx) {
   auto *coordinator_query = storage_->Create<CoordinatorQuery>();
@@ -400,10 +399,28 @@ antlrcpp::Any CypherMainVisitor::visitRegisterInstanceOnCoordinator(
   return coordinator_query;
 }
 
-// License check is done in the interpreter
-antlrcpp::Any CypherMainVisitor::visitShowReplicationCluster(MemgraphCypher::ShowReplicationClusterContext * /*ctx*/) {
+antlrcpp::Any CypherMainVisitor::visitAddCoordinatorInstance(MemgraphCypher::AddCoordinatorInstanceContext *ctx) {
   auto *coordinator_query = storage_->Create<CoordinatorQuery>();
-  coordinator_query->action_ = CoordinatorQuery::Action::SHOW_REPLICATION_CLUSTER;
+
+  if (!ctx->raftSocketAddress()->literal()->StringLiteral()) {
+    throw SemanticException("Raft socket address should be a string literal!");
+  }
+
+  if (!ctx->raftServerId()->literal()->numberLiteral()) {
+    throw SemanticException("Raft server id should be a number literal!");
+  }
+
+  coordinator_query->action_ = CoordinatorQuery::Action::ADD_COORDINATOR_INSTANCE;
+  coordinator_query->raft_socket_address_ = std::any_cast<Expression *>(ctx->raftSocketAddress()->accept(this));
+  coordinator_query->raft_server_id_ = std::any_cast<Expression *>(ctx->raftServerId()->accept(this));
+
+  return coordinator_query;
+}
+
+// License check is done in the interpreter
+antlrcpp::Any CypherMainVisitor::visitShowInstances(MemgraphCypher::ShowInstancesContext * /*ctx*/) {
+  auto *coordinator_query = storage_->Create<CoordinatorQuery>();
+  coordinator_query->action_ = CoordinatorQuery::Action::SHOW_INSTANCES;
   return coordinator_query;
 }
 

src/query/frontend/ast/cypher_main_visitor.hpp

@@ -251,7 +251,12 @@ class CypherMainVisitor : public antlropencypher::MemgraphCypherBaseVisitor {
   /**
    * @return CoordinatorQuery*
    */
-  antlrcpp::Any visitShowReplicationCluster(MemgraphCypher::ShowReplicationClusterContext *ctx) override;
+  antlrcpp::Any visitAddCoordinatorInstance(MemgraphCypher::AddCoordinatorInstanceContext *ctx) override;
+
+  /**
+   * @return CoordinatorQuery*
+   */
+  antlrcpp::Any visitShowInstances(MemgraphCypher::ShowInstancesContext *ctx) override;
 
   /**
    * @return LockPathQuery*

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

@@ -20,6 +20,7 @@ options { tokenVocab=MemgraphCypherLexer; }
 import Cypher ;
 
 memgraphCypherKeyword : cypherKeyword
+                      | ADD
                       | ACTIVE
                       | AFTER
                       | ALTER
@@ -64,6 +65,7 @@ memgraphCypherKeyword : cypherKeyword
                       | HEADER
                       | IDENTIFIED
                       | INSTANCE
+                      | INSTANCES
                       | NODE_LABELS
                       | NULLIF
                       | IMPORT
@@ -189,7 +191,8 @@ replicationQuery : setReplicationRole
 
 coordinatorQuery : registerInstanceOnCoordinator
                  | setInstanceToMain
-                 | showReplicationCluster
+                 | showInstances
+                 | addCoordinatorInstance
                  ;
 
 triggerQuery : createTrigger
@@ -374,7 +377,7 @@ setReplicationRole : SET REPLICATION ROLE TO ( MAIN | REPLICA )
 
 showReplicationRole : SHOW REPLICATION ROLE ;
 
-showReplicationCluster : SHOW REPLICATION CLUSTER ;
+showInstances : SHOW INSTANCES ;
 
 instanceName : symbolicName ;
 
@@ -382,14 +385,19 @@ socketAddress : literal ;
 
 coordinatorSocketAddress : literal ;
 replicationSocketAddress : literal ;
+raftSocketAddress : 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 ;
 
+raftServerId : literal ;
+
+addCoordinatorInstance : ADD COORDINATOR raftServerId ON raftSocketAddress ;
+
 dropReplica : DROP REPLICA instanceName ;
 
 showReplicas : SHOW REPLICAS ;

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

@@ -23,6 +23,7 @@ lexer grammar MemgraphCypherLexer ;
 
 import CypherLexer ;
 
+ADD                     : A D D ;
 ACTIVE                  : A C T I V E ;
 AFTER                   : A F T E R ;
 ALTER                   : A L T E R ;
@@ -39,7 +40,6 @@ BOOTSTRAP_SERVERS       : B O O T S T R A P UNDERSCORE S E R V E R S ;
 CALL                    : C A L L ;
 CHECK                   : C H E C K ;
 CLEAR                   : C L E A R ;
-CLUSTER                 : C L U S T E R ;
 COMMIT                  : C O M M I T ;
 COMMITTED               : C O M M I T T E D ;
 CONFIG                  : C O N F I G ;
@@ -80,6 +80,7 @@ INACTIVE                : I N A C T I V E ;
 IN_MEMORY_ANALYTICAL    : I N UNDERSCORE M E M O R Y UNDERSCORE A N A L Y T I C A L ;
 IN_MEMORY_TRANSACTIONAL : I N UNDERSCORE M E M O R Y UNDERSCORE T R A N S A C T I O N A L ;
 INSTANCE                : I N S T A N C E ;
+INSTANCES               : I N S T A N C E S ;
 ISOLATION               : I S O L A T I O N ;
 KAFKA                   : K A F K A ;
 LABELS                  : L A B E L S ;

src/query/frontend/stripped_lexer_constants.hpp

@@ -219,7 +219,8 @@ const trie::Trie kKeywords = {"union",
                               "lock",
                               "unlock",
                               "build",
-                              "instance"};
+                              "instance",
+                              "coordinator"};
 
 // Unicode codepoints that are allowed at the start of the unescaped name.
 const std::bitset<kBitsetSize> kUnescapedNameAllowedStarts(

src/query/interpreter.hpp

@@ -53,7 +53,7 @@
 #include "utils/tsc.hpp"
 
 #ifdef MG_ENTERPRISE
-#include "coordination/coordinator_instance_status.hpp"
+#include "coordination/instance_status.hpp"
 #endif
 
 namespace memgraph::metrics {
@@ -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;
@@ -103,23 +103,24 @@ 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,
-                                const std::chrono::seconds instance_check_frequency, const std::string &instance_name,
-                                CoordinatorQuery::SyncMode sync_mode) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual void SetInstanceToMain(const std::string &instance_name) = 0;
+  virtual void RegisterReplicationInstance(const std::string &coordinator_socket_address,
+                                           const std::string &replication_socket_address,
+                                           const std::chrono::seconds instance_check_frequency,
+                                           const std::string &instance_name, CoordinatorQuery::SyncMode sync_mode) = 0;
 
   /// @throw QueryRuntimeException if an error ocurred.
-  virtual std::vector<coordination::CoordinatorInstanceStatus> ShowInstances() const = 0;
+  virtual void SetReplicationInstanceToMain(const std::string &instance_name) = 0;
 
-#endif
+  /// @throw QueryRuntimeException if an error ocurred.
+  virtual std::vector<coordination::InstanceStatus> ShowInstances() const = 0;
+
+  /// @throw QueryRuntimeException if an error ocurred.
+  virtual auto AddCoordinatorInstance(uint32_t raft_server_id, std::string const &coordinator_socket_address)
+      -> void = 0;
 };
+#endif
 
 class AnalyzeGraphQueryHandler {
  public:
@@ -296,32 +297,12 @@ 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 +367,6 @@ 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/metadata.cpp

@@ -69,6 +69,8 @@ constexpr std::string_view GetCodeString(const NotificationCode code) {
 #ifdef MG_ENTERPRISE
     case NotificationCode::REGISTER_COORDINATOR_SERVER:
       return "RegisterCoordinatorServer"sv;
+    case NotificationCode::ADD_COORDINATOR_INSTANCE:
+      return "AddCoordinatorInstance"sv;
 #endif
     case NotificationCode::REPLICA_PORT_WARNING:
       return "ReplicaPortWarning"sv;

src/query/metadata.hpp

@@ -44,6 +44,7 @@ enum class NotificationCode : uint8_t {
   REGISTER_REPLICA,
 #ifdef MG_ENTERPRISE
   REGISTER_COORDINATOR_SERVER,
+  ADD_COORDINATOR_INSTANCE,
 #endif
   SET_REPLICA,
   START_STREAM,

src/query/path.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
@@ -114,12 +114,17 @@ class Path {
   /** Expands the path with the given vertex. */
   void Expand(const VertexAccessor &vertex) {
     DMG_ASSERT(vertices_.size() == edges_.size(), "Illegal path construction order");
+    DMG_ASSERT(edges_.empty() || (!edges_.empty() && (edges_.back().To().Gid() == vertex.Gid() ||
+                                                      edges_.back().From().Gid() == vertex.Gid())),
+               "Illegal path construction order");
     vertices_.emplace_back(vertex);
   }
 
   /** Expands the path with the given edge. */
   void Expand(const EdgeAccessor &edge) {
     DMG_ASSERT(vertices_.size() - 1 == edges_.size(), "Illegal path construction order");
+    DMG_ASSERT(vertices_.back().Gid() == edge.From().Gid() || vertices_.back().Gid() == edge.To().Gid(),
+               "Illegal path construction order");
     edges_.emplace_back(edge);
   }
 
@@ -130,6 +135,14 @@ class Path {
     Expand(others...);
   }
 
+  void Shrink() {
+    DMG_ASSERT(!vertices_.empty(), "Vertices should not be empty in the path before shrink.");
+    vertices_.pop_back();
+    if (!edges_.empty()) {
+      edges_.pop_back();
+    }
+  }
+
   /** Returns the number of expansions (edges) in this path. */
   auto size() const { return edges_.size(); }
 

src/query/plan/operator.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
@@ -1057,7 +1057,8 @@ class ExpandVariableCursor : public Cursor {
           if (!self_.common_.existing_node) {
             frame[self_.common_.node_symbol] = start_vertex;
             return true;
-          } else if (CheckExistingNode(start_vertex, self_.common_.node_symbol, frame)) {
+          }
+          if (CheckExistingNode(start_vertex, self_.common_.node_symbol, frame)) {
             return true;
           }
         }
@@ -1243,6 +1244,10 @@ class ExpandVariableCursor : public Cursor {
         MG_ASSERT(frame[self_.filter_lambda_.accumulated_path_symbol.value()].IsPath(),
                   "Accumulated path must be path");
         Path &accumulated_path = frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath();
+        // Shrink the accumulated path including current level if necessary
+        while (accumulated_path.size() >= edges_on_frame.size()) {
+          accumulated_path.Shrink();
+        }
         accumulated_path.Expand(current_edge.first);
         accumulated_path.Expand(current_vertex);
       }
@@ -1260,10 +1265,9 @@ class ExpandVariableCursor : public Cursor {
       if (self_.common_.existing_node && !CheckExistingNode(current_vertex, self_.common_.node_symbol, frame)) continue;
 
       // We only yield true if we satisfy the lower bound.
-      if (static_cast<int64_t>(edges_on_frame.size()) >= lower_bound_)
+      if (static_cast<int64_t>(edges_on_frame.size()) >= lower_bound_) {
         return true;
-      else
-        continue;
+      }
     }
   }
 };
@@ -1527,8 +1531,8 @@ class SingleSourceShortestPathCursor : public query::plan::Cursor {
       : self_(self),
         input_cursor_(self_.input()->MakeCursor(mem)),
         processed_(mem),
-        to_visit_current_(mem),
-        to_visit_next_(mem) {
+        to_visit_next_(mem),
+        to_visit_current_(mem) {
     MG_ASSERT(!self_.common_.existing_node,
               "Single source shortest path algorithm "
               "should not be used when `existing_node` "
@@ -1558,12 +1562,14 @@ class SingleSourceShortestPathCursor : public query::plan::Cursor {
 #endif
       frame[self_.filter_lambda_.inner_edge_symbol] = edge;
       frame[self_.filter_lambda_.inner_node_symbol] = vertex;
+      std::optional<Path> curr_acc_path = std::nullopt;
       if (self_.filter_lambda_.accumulated_path_symbol) {
         MG_ASSERT(frame[self_.filter_lambda_.accumulated_path_symbol.value()].IsPath(),
                   "Accumulated path must have Path type");
         Path &accumulated_path = frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath();
         accumulated_path.Expand(edge);
         accumulated_path.Expand(vertex);
+        curr_acc_path = accumulated_path;
       }
 
       if (self_.filter_lambda_.expression) {
@@ -1578,21 +1584,33 @@ class SingleSourceShortestPathCursor : public query::plan::Cursor {
             throw QueryRuntimeException("Expansion condition must evaluate to boolean or null.");
         }
       }
-      to_visit_next_.emplace_back(edge, vertex);
+      to_visit_next_.emplace_back(edge, vertex, std::move(curr_acc_path));
       processed_.emplace(vertex, edge);
     };
 
+    auto restore_frame_state_after_expansion = [this, &frame]() {
+      if (self_.filter_lambda_.accumulated_path_symbol) {
+        frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath().Shrink();
+      }
+    };
+
     // populates the to_visit_next_ structure with expansions
     // from the given vertex. skips expansions that don't satisfy
     // the "where" condition.
-    auto expand_from_vertex = [this, &expand_pair](const auto &vertex) {
+    auto expand_from_vertex = [this, &expand_pair, &restore_frame_state_after_expansion](const auto &vertex) {
       if (self_.common_.direction != EdgeAtom::Direction::IN) {
         auto out_edges = UnwrapEdgesResult(vertex.OutEdges(storage::View::OLD, self_.common_.edge_types)).edges;
-        for (const auto &edge : out_edges) expand_pair(edge, edge.To());
+        for (const auto &edge : out_edges) {
+          expand_pair(edge, edge.To());
+          restore_frame_state_after_expansion();
+        }
       }
       if (self_.common_.direction != EdgeAtom::Direction::OUT) {
         auto in_edges = UnwrapEdgesResult(vertex.InEdges(storage::View::OLD, self_.common_.edge_types)).edges;
-        for (const auto &edge : in_edges) expand_pair(edge, edge.From());
+        for (const auto &edge : in_edges) {
+          expand_pair(edge, edge.From());
+          restore_frame_state_after_expansion();
+        }
       }
     };
 
@@ -1638,14 +1656,14 @@ class SingleSourceShortestPathCursor : public query::plan::Cursor {
       }
 
       // take the next expansion from the queue
-      auto expansion = to_visit_current_.back();
+      auto [curr_edge, curr_vertex, curr_acc_path] = to_visit_current_.back();
       to_visit_current_.pop_back();
 
       // create the frame value for the edges
       auto *pull_memory = context.evaluation_context.memory;
       utils::pmr::vector<TypedValue> edge_list(pull_memory);
-      edge_list.emplace_back(expansion.first);
-      auto last_vertex = expansion.second;
+      edge_list.emplace_back(curr_edge);
+      auto last_vertex = curr_vertex;
       while (true) {
         const EdgeAccessor &last_edge = edge_list.back().ValueEdge();
         last_vertex = last_edge.From() == last_vertex ? last_edge.To() : last_edge.From();
@@ -1657,11 +1675,17 @@ class SingleSourceShortestPathCursor : public query::plan::Cursor {
       }
 
       // expand only if what we've just expanded is less then max depth
-      if (static_cast<int64_t>(edge_list.size()) < upper_bound_) expand_from_vertex(expansion.second);
+      if (static_cast<int64_t>(edge_list.size()) < upper_bound_) {
+        if (self_.filter_lambda_.accumulated_path_symbol) {
+          MG_ASSERT(curr_acc_path.has_value(), "Expected non-null accumulated path");
+          frame[self_.filter_lambda_.accumulated_path_symbol.value()] = std::move(curr_acc_path.value());
+        }
+        expand_from_vertex(curr_vertex);
+      }
 
       if (static_cast<int64_t>(edge_list.size()) < lower_bound_) continue;
 
-      frame[self_.common_.node_symbol] = expansion.second;
+      frame[self_.common_.node_symbol] = curr_vertex;
 
       // place edges on the frame in the correct order
       std::reverse(edge_list.begin(), edge_list.end());
@@ -1693,9 +1717,9 @@ class SingleSourceShortestPathCursor : public query::plan::Cursor {
   // edge because the root does not get expanded from anything.
   // contains visited vertices as well as those scheduled to be visited.
   utils::pmr::unordered_map<VertexAccessor, std::optional<EdgeAccessor>> processed_;
-  // edge/vertex pairs we have yet to visit, for current and next depth
-  utils::pmr::vector<std::pair<EdgeAccessor, VertexAccessor>> to_visit_current_;
-  utils::pmr::vector<std::pair<EdgeAccessor, VertexAccessor>> to_visit_next_;
+  // edge, vertex we have yet to visit, for current and next depth and their accumulated paths
+  utils::pmr::vector<std::tuple<EdgeAccessor, VertexAccessor, std::optional<Path>>> to_visit_next_;
+  utils::pmr::vector<std::tuple<EdgeAccessor, VertexAccessor, std::optional<Path>>> to_visit_current_;
 };
 
 namespace {
@@ -1768,6 +1792,7 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
 
     ExpressionEvaluator evaluator(&frame, context.symbol_table, context.evaluation_context, context.db_accessor,
                                   storage::View::OLD);
+
     auto create_state = [this](const VertexAccessor &vertex, int64_t depth) {
       return std::make_pair(vertex, upper_bound_set_ ? depth : 0);
     };
@@ -1791,6 +1816,7 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
       frame[self_.weight_lambda_->inner_node_symbol] = vertex;
       TypedValue next_weight = CalculateNextWeight(self_.weight_lambda_, total_weight, evaluator);
 
+      std::optional<Path> curr_acc_path = std::nullopt;
       if (self_.filter_lambda_.expression) {
         frame[self_.filter_lambda_.inner_edge_symbol] = edge;
         frame[self_.filter_lambda_.inner_node_symbol] = vertex;
@@ -1800,6 +1826,7 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
           Path &accumulated_path = frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath();
           accumulated_path.Expand(edge);
           accumulated_path.Expand(vertex);
+          curr_acc_path = accumulated_path;
 
           if (self_.filter_lambda_.accumulated_weight_symbol) {
             frame[self_.filter_lambda_.accumulated_weight_symbol.value()] = next_weight;
@@ -1815,24 +1842,32 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
       if (found_it != total_cost_.end() && (found_it->second.IsNull() || (found_it->second <= next_weight).ValueBool()))
         return;
 
-      pq_.emplace(next_weight, depth + 1, vertex, edge);
+      pq_.emplace(next_weight, depth + 1, vertex, edge, curr_acc_path);
+    };
+
+    auto restore_frame_state_after_expansion = [this, &frame]() {
+      if (self_.filter_lambda_.accumulated_path_symbol) {
+        frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath().Shrink();
+      }
     };
 
     // Populates the priority queue structure with expansions
     // from the given vertex. skips expansions that don't satisfy
     // the "where" condition.
-    auto expand_from_vertex = [this, &expand_pair](const VertexAccessor &vertex, const TypedValue &weight,
-                                                   int64_t depth) {
+    auto expand_from_vertex = [this, &expand_pair, &restore_frame_state_after_expansion](
+                                  const VertexAccessor &vertex, const TypedValue &weight, int64_t depth) {
       if (self_.common_.direction != EdgeAtom::Direction::IN) {
         auto out_edges = UnwrapEdgesResult(vertex.OutEdges(storage::View::OLD, self_.common_.edge_types)).edges;
         for (const auto &edge : out_edges) {
           expand_pair(edge, edge.To(), weight, depth);
+          restore_frame_state_after_expansion();
         }
       }
       if (self_.common_.direction != EdgeAtom::Direction::OUT) {
         auto in_edges = UnwrapEdgesResult(vertex.InEdges(storage::View::OLD, self_.common_.edge_types)).edges;
         for (const auto &edge : in_edges) {
           expand_pair(edge, edge.From(), weight, depth);
+          restore_frame_state_after_expansion();
         }
       }
     };
@@ -1850,9 +1885,12 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
           // Skip expansion for such nodes.
           if (node.IsNull()) continue;
         }
+
+        std::optional<Path> curr_acc_path;
         if (self_.filter_lambda_.accumulated_path_symbol) {
           // Add initial vertex of path to the accumulated path
-          frame[self_.filter_lambda_.accumulated_path_symbol.value()] = Path(vertex);
+          curr_acc_path = Path(vertex);
+          frame[self_.filter_lambda_.accumulated_path_symbol.value()] = curr_acc_path.value();
         }
         if (self_.upper_bound_) {
           upper_bound_ = EvaluateInt(&evaluator, self_.upper_bound_, "Max depth in weighted shortest path expansion");
@@ -1876,7 +1914,7 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
         total_cost_.clear();
         yielded_vertices_.clear();
 
-        pq_.emplace(current_weight, 0, vertex, std::nullopt);
+        pq_.emplace(current_weight, 0, vertex, std::nullopt, curr_acc_path);
         // We are adding the starting vertex to the set of yielded vertices
         // because we don't want to yield paths that end with the starting
         // vertex.
@@ -1885,7 +1923,7 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
 
       while (!pq_.empty()) {
         AbortCheck(context);
-        auto [current_weight, current_depth, current_vertex, current_edge] = pq_.top();
+        auto [current_weight, current_depth, current_vertex, current_edge, curr_acc_path] = pq_.top();
         pq_.pop();
 
         auto current_state = create_state(current_vertex, current_depth);
@@ -1898,7 +1936,12 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
         total_cost_.emplace(current_state, current_weight);
 
         // Expand only if what we've just expanded is less than max depth.
-        if (current_depth < upper_bound_) expand_from_vertex(current_vertex, current_weight, current_depth);
+        if (current_depth < upper_bound_) {
+          if (self_.filter_lambda_.accumulated_path_symbol) {
+            frame[self_.filter_lambda_.accumulated_path_symbol.value()] = std::move(curr_acc_path.value());
+          }
+          expand_from_vertex(current_vertex, current_weight, current_depth);
+        }
 
         // If we yielded a path for a vertex already, make the expansion but
         // don't return the path again.
@@ -1921,12 +1964,12 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
         // Place destination node on the frame, handle existence flag.
         if (self_.common_.existing_node) {
           const auto &node = frame[self_.common_.node_symbol];
-          if ((node != TypedValue(current_vertex, pull_memory)).ValueBool())
+          if ((node != TypedValue(current_vertex, pull_memory)).ValueBool()) {
             continue;
-          else
-            // Prevent expanding other paths, because we found the
-            // shortest to existing node.
-            ClearQueue();
+          }
+          // Prevent expanding other paths, because we found the
+          // shortest to existing node.
+          ClearQueue();
         } else {
           frame[self_.common_.node_symbol] = current_vertex;
         }
@@ -1979,8 +2022,9 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
   // Priority queue comparator. Keep lowest weight on top of the queue.
   class PriorityQueueComparator {
    public:
-    bool operator()(const std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>> &lhs,
-                    const std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>> &rhs) {
+    bool operator()(
+        const std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>, std::optional<Path>> &lhs,
+        const std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>, std::optional<Path>> &rhs) {
       const auto &lhs_weight = std::get<0>(lhs);
       const auto &rhs_weight = std::get<0>(rhs);
       // Null defines minimum value for all types
@@ -1997,8 +2041,9 @@ class ExpandWeightedShortestPathCursor : public query::plan::Cursor {
     }
   };
 
-  std::priority_queue<std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>>,
-                      utils::pmr::vector<std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>>>,
+  std::priority_queue<std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>, std::optional<Path>>,
+                      utils::pmr::vector<std::tuple<TypedValue, int64_t, VertexAccessor, std::optional<EdgeAccessor>,
+                                                    std::optional<Path>>>,
                       PriorityQueueComparator>
       pq_;
 
@@ -2024,6 +2069,9 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
 
     ExpressionEvaluator evaluator(&frame, context.symbol_table, context.evaluation_context, context.db_accessor,
                                   storage::View::OLD);
+
+    auto *memory = context.evaluation_context.memory;
+
     auto create_state = [this](const VertexAccessor &vertex, int64_t depth) {
       return std::make_pair(vertex, upper_bound_set_ ? depth : 0);
     };
@@ -2041,6 +2089,7 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
       TypedValue next_weight = CalculateNextWeight(self_.weight_lambda_, total_weight, evaluator);
 
       // If filter expression exists, evaluate filter
+      std::optional<Path> curr_acc_path = std::nullopt;
       if (self_.filter_lambda_.expression) {
         frame[self_.filter_lambda_.inner_edge_symbol] = edge;
         frame[self_.filter_lambda_.inner_node_symbol] = next_vertex;
@@ -2050,6 +2099,7 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
           Path &accumulated_path = frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath();
           accumulated_path.Expand(edge);
           accumulated_path.Expand(next_vertex);
+          curr_acc_path = accumulated_path;
 
           if (self_.filter_lambda_.accumulated_weight_symbol) {
             frame[self_.filter_lambda_.accumulated_weight_symbol.value()] = next_weight;
@@ -2076,14 +2126,20 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
       }
 
       DirectedEdge directed_edge = {edge, direction, next_weight};
-      pq_.emplace(next_weight, depth + 1, next_vertex, directed_edge);
+      pq_.emplace(next_weight, depth + 1, next_vertex, directed_edge, curr_acc_path);
+    };
+
+    auto restore_frame_state_after_expansion = [this, &frame]() {
+      if (self_.filter_lambda_.accumulated_path_symbol) {
+        frame[self_.filter_lambda_.accumulated_path_symbol.value()].ValuePath().Shrink();
+      }
     };
 
     // Populates the priority queue structure with expansions
     // from the given vertex. skips expansions that don't satisfy
     // the "where" condition.
-    auto expand_from_vertex = [this, &expand_vertex, &context](const VertexAccessor &vertex, const TypedValue &weight,
-                                                               int64_t depth) {
+    auto expand_from_vertex = [this, &expand_vertex, &context, &restore_frame_state_after_expansion](
+                                  const VertexAccessor &vertex, const TypedValue &weight, int64_t depth) {
       if (self_.common_.direction != EdgeAtom::Direction::IN) {
         auto out_edges = UnwrapEdgesResult(vertex.OutEdges(storage::View::OLD, self_.common_.edge_types)).edges;
         for (const auto &edge : out_edges) {
@@ -2096,6 +2152,7 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
           }
 #endif
           expand_vertex(edge, EdgeAtom::Direction::OUT, weight, depth);
+          restore_frame_state_after_expansion();
         }
       }
       if (self_.common_.direction != EdgeAtom::Direction::OUT) {
@@ -2110,12 +2167,12 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
           }
 #endif
           expand_vertex(edge, EdgeAtom::Direction::IN, weight, depth);
+          restore_frame_state_after_expansion();
         }
       }
     };
 
     std::optional<VertexAccessor> start_vertex;
-    auto *memory = context.evaluation_context.memory;
 
     auto create_path = [this, &frame, &memory]() {
       auto &current_level = traversal_stack_.back();
@@ -2167,11 +2224,11 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
       return true;
     };
 
-    auto create_DFS_traversal_tree = [this, &context, &memory, &create_state, &expand_from_vertex]() {
+    auto create_DFS_traversal_tree = [this, &context, &memory, &frame, &create_state, &expand_from_vertex]() {
       while (!pq_.empty()) {
         AbortCheck(context);
 
-        const auto [current_weight, current_depth, current_vertex, directed_edge] = pq_.top();
+        auto [current_weight, current_depth, current_vertex, directed_edge, acc_path] = pq_.top();
         pq_.pop();
 
         const auto &[current_edge, direction, weight] = directed_edge;
@@ -2183,6 +2240,10 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
         } else {
           total_cost_.emplace(current_state, current_weight);
           if (current_depth < upper_bound_) {
+            if (self_.filter_lambda_.accumulated_path_symbol) {
+              DMG_ASSERT(acc_path.has_value(), "Path must be already filled in AllShortestPath DFS traversals");
+              frame[self_.filter_lambda_.accumulated_path_symbol.value()] = std::move(acc_path.value());
+            }
             expand_from_vertex(current_vertex, current_weight, current_depth);
           }
         }
@@ -2315,8 +2376,8 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
   // Priority queue comparator. Keep lowest weight on top of the queue.
   class PriorityQueueComparator {
    public:
-    bool operator()(const std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge> &lhs,
-                    const std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge> &rhs) {
+    bool operator()(const std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge, std::optional<Path>> &lhs,
+                    const std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge, std::optional<Path>> &rhs) {
       const auto &lhs_weight = std::get<0>(lhs);
       const auto &rhs_weight = std::get<0>(rhs);
       // Null defines minimum value for all types
@@ -2335,9 +2396,10 @@ class ExpandAllShortestPathsCursor : public query::plan::Cursor {
 
   // Priority queue - core element of the algorithm.
   // Stores: {weight, depth, next vertex, edge and direction}
-  std::priority_queue<std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge>,
-                      utils::pmr::vector<std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge>>,
-                      PriorityQueueComparator>
+  std::priority_queue<
+      std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge, std::optional<Path>>,
+      utils::pmr::vector<std::tuple<TypedValue, int64_t, VertexAccessor, DirectedEdge, std::optional<Path>>>,
+      PriorityQueueComparator>
       pq_;
 
   void ClearQueue() {

src/query/replication_query_handler.hpp

@@ -0,0 +1,68 @@
+// 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"
+#include "utils/uuid.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,
+  ERROR_ACCEPTING_MAIN
+};
+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,
+                                         const std::optional<utils::UUID> &main_uuid) = 0;
+
+  // as MAIN, define and connect to REPLICAs
+  virtual auto TryRegisterReplica(const memgraph::replication::ReplicationClientConfig &config, bool send_swap_uuid)
+      -> utils::BasicResult<RegisterReplicaError> = 0;
+
+  virtual auto RegisterReplica(const memgraph::replication::ReplicationClientConfig &config, bool send_swap_uuid)
+      -> 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
 )
@@ -23,6 +21,6 @@ target_include_directories(mg-replication PUBLIC include)
 
 find_package(fmt REQUIRED)
 target_link_libraries(mg-replication
-    PUBLIC mg::utils mg::kvstore lib::json mg::rpc mg::slk mg::io mg::repl_coord_glue
+    PUBLIC mg::utils mg::kvstore lib::json mg::rpc mg::slk mg::io mg::repl_coord_glue mg-flags
     PRIVATE fmt::fmt
 )

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

@@ -41,30 +41,74 @@ 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 underlying 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_;
   std::chrono::seconds replica_check_frequency_;
+  // True only when we are migrating from V1 or V2 to V3 in replication durability
+  // and we want to set replica to listen to main
+  bool try_set_uuid{false};
 
   // TODO: Better, this was the easiest place to put this
   enum class State {

src/replication/include/replication/state.hpp

@@ -21,10 +21,12 @@
 #include "status.hpp"
 #include "utils/result.hpp"
 #include "utils/synchronized.hpp"
+#include "utils/uuid.hpp"
 
 #include <atomic>
 #include <cstdint>
 #include <list>
+#include <optional>
 #include <variant>
 #include <vector>
 
@@ -37,7 +39,11 @@ enum class RegisterReplicaError : uint8_t { NAME_EXISTS, ENDPOINT_EXISTS, COULD_
 
 struct RoleMainData {
   RoleMainData() = default;
-  explicit RoleMainData(ReplicationEpoch e) : epoch_(std::move(e)) {}
+  explicit RoleMainData(ReplicationEpoch e, std::optional<utils::UUID> uuid = std::nullopt) : epoch_(std::move(e)) {
+    if (uuid) {
+      uuid_ = *uuid;
+    }
+  }
   ~RoleMainData() = default;
 
   RoleMainData(RoleMainData const &) = delete;
@@ -47,11 +53,14 @@ struct RoleMainData {
 
   ReplicationEpoch epoch_;
   std::list<ReplicationClient> registered_replicas_{};  // TODO: data race issues
+  utils::UUID uuid_;
 };
 
 struct RoleReplicaData {
   ReplicationServerConfig config;
   std::unique_ptr<ReplicationServer> server;
+  // uuid of main replica is listening to
+  std::optional<utils::UUID> uuid_;
 };
 
 // Global (instance) level object
@@ -83,18 +92,19 @@ struct ReplicationState {
 
   bool HasDurability() const { return nullptr != durability_; }
 
-  bool TryPersistRoleMain(std::string new_epoch);
-  bool TryPersistRoleReplica(const ReplicationServerConfig &config);
+  bool TryPersistRoleMain(std::string new_epoch, utils::UUID main_uuid);
+  bool TryPersistRoleReplica(const ReplicationServerConfig &config, const std::optional<utils::UUID> &main_uuid);
   bool TryPersistUnregisterReplica(std::string_view name);
-  bool TryPersistRegisteredReplica(const ReplicationClientConfig &config);
+  bool TryPersistRegisteredReplica(const ReplicationClientConfig &config, utils::UUID main_uuid);
 
   // TODO: locked access
   auto ReplicationData() -> ReplicationData_t & { return replication_data_; }
   auto ReplicationData() const -> ReplicationData_t const & { return replication_data_; }
   utils::BasicResult<RegisterReplicaError, ReplicationClient *> RegisterReplica(const ReplicationClientConfig &config);
 
-  bool SetReplicationRoleMain();
-  bool SetReplicationRoleReplica(const ReplicationServerConfig &config);
+  bool SetReplicationRoleMain(const utils::UUID &main_uuid);
+  bool SetReplicationRoleReplica(const ReplicationServerConfig &config,
+                                 const std::optional<utils::UUID> &main_uuid = std::nullopt);
 
  private:
   bool HandleVersionMigration(durability::ReplicationRoleEntry &data) const;

src/replication/include/replication/status.hpp

@@ -31,25 +31,28 @@ constexpr auto *kReplicationReplicaPrefix{"__replication_replica:"};  // introdu
 
 enum class DurabilityVersion : uint8_t {
   V1,  // no distinct key for replicas
-  V2,  // this version, epoch, replica prefix introduced
+  V2,  // epoch, replica prefix introduced
+  V3,  // this version, main uuid introduced
 };
 
 // fragment of key: "__replication_role"
 struct MainRole {
   ReplicationEpoch epoch{};
+  std::optional<utils::UUID> main_uuid{};
   friend bool operator==(MainRole const &, MainRole const &) = default;
 };
 
 // fragment of key: "__replication_role"
 struct ReplicaRole {
   ReplicationServerConfig config{};
+  std::optional<utils::UUID> main_uuid{};
   friend bool operator==(ReplicaRole const &, ReplicaRole const &) = default;
 };
 
 // from key: "__replication_role"
 struct ReplicationRoleEntry {
   DurabilityVersion version =
-      DurabilityVersion::V2;  // if not latest then migration required for kReplicationReplicaPrefix
+      DurabilityVersion::V3;  // if not latest then migration required for kReplicationReplicaPrefix
   std::variant<MainRole, ReplicaRole> role;
 
   friend bool operator==(ReplicationRoleEntry const &, ReplicationRoleEntry const &) = default;

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/replication_server.cpp

@@ -10,7 +10,7 @@
 // licenses/APL.txt.
 
 #include "replication/replication_server.hpp"
-#include "replication_coordination_glue/messages.hpp"
+#include "replication_coordination_glue/handler.hpp"
 
 namespace memgraph::replication {
 namespace {

src/replication/state.cpp

@@ -10,12 +10,15 @@
 // licenses/APL.txt.
 
 #include "replication/state.hpp"
+#include <optional>
 
+#include "flags/replication.hpp"
 #include "replication/replication_client.hpp"
 #include "replication/replication_server.hpp"
 #include "replication/status.hpp"
 #include "utils/file.hpp"
 #include "utils/result.hpp"
+#include "utils/uuid.hpp"
 #include "utils/variant_helpers.hpp"
 
 constexpr auto kReplicationDirectory = std::string_view{"replication"};
@@ -36,9 +39,9 @@ ReplicationState::ReplicationState(std::optional<std::filesystem::path> durabili
   durability_ = std::make_unique<kvstore::KVStore>(std::move(repl_dir));
   spdlog::info("Replication configuration will be stored and will be automatically restored in case of a crash.");
 
-  auto replicationData = FetchReplicationData();
-  if (replicationData.HasError()) {
-    switch (replicationData.GetError()) {
+  auto fetched_replication_data = FetchReplicationData();
+  if (fetched_replication_data.HasError()) {
+    switch (fetched_replication_data.GetError()) {
       using enum ReplicationState::FetchReplicationError;
       case NOTHING_FETCHED: {
         spdlog::debug("Cannot find data needed for restore replication role in persisted metadata.");
@@ -51,15 +54,21 @@ ReplicationState::ReplicationState(std::optional<std::filesystem::path> durabili
       }
     }
   }
-  replication_data_ = std::move(replicationData).GetValue();
+  auto replication_data = std::move(fetched_replication_data).GetValue();
+#ifdef MG_ENTERPRISE
+  if (FLAGS_coordinator_server_port && std::holds_alternative<RoleReplicaData>(replication_data)) {
+    std::get<RoleReplicaData>(replication_data).uuid_.reset();
+  }
+#endif
+  replication_data_ = std::move(replication_data);
 }
 
-bool ReplicationState::TryPersistRoleReplica(const ReplicationServerConfig &config) {
+bool ReplicationState::TryPersistRoleReplica(const ReplicationServerConfig &config,
+                                             const std::optional<utils::UUID> &main_uuid) {
   if (!HasDurability()) return true;
 
-  auto data = durability::ReplicationRoleEntry{.role = durability::ReplicaRole{
-                                                   .config = config,
-                                               }};
+  auto data =
+      durability::ReplicationRoleEntry{.role = durability::ReplicaRole{.config = config, .main_uuid = main_uuid}};
 
   if (!durability_->Put(durability::kReplicationRoleName, nlohmann::json(data).dump())) {
     spdlog::error("Error when saving REPLICA replication role in settings.");
@@ -78,11 +87,11 @@ bool ReplicationState::TryPersistRoleReplica(const ReplicationServerConfig &conf
   return true;
 }
 
-bool ReplicationState::TryPersistRoleMain(std::string new_epoch) {
+bool ReplicationState::TryPersistRoleMain(std::string new_epoch, utils::UUID main_uuid) {
   if (!HasDurability()) return true;
 
-  auto data =
-      durability::ReplicationRoleEntry{.role = durability::MainRole{.epoch = ReplicationEpoch{std::move(new_epoch)}}};
+  auto data = durability::ReplicationRoleEntry{
+      .role = durability::MainRole{.epoch = ReplicationEpoch{std::move(new_epoch)}, .main_uuid = main_uuid}};
 
   if (durability_->Put(durability::kReplicationRoleName, nlohmann::json(data).dump())) {
     role_persisted = RolePersisted::YES;
@@ -128,7 +137,8 @@ auto ReplicationState::FetchReplicationData() -> FetchReplicationResult_t {
     return std::visit(
         utils::Overloaded{
             [&](durability::MainRole &&r) -> FetchReplicationResult_t {
-              auto res = RoleMainData{std::move(r.epoch)};
+              auto res =
+                  RoleMainData{std::move(r.epoch), r.main_uuid.has_value() ? r.main_uuid.value() : utils::UUID{}};
               auto b = durability_->begin(durability::kReplicationReplicaPrefix);
               auto e = durability_->end(durability::kReplicationReplicaPrefix);
               for (; b != e; ++b) {
@@ -143,6 +153,8 @@ auto ReplicationState::FetchReplicationData() -> FetchReplicationResult_t {
                   }
                   // Instance clients
                   res.registered_replicas_.emplace_back(data.config);
+                  // Bump for each replica uuid
+                  res.registered_replicas_.back().try_set_uuid = !r.main_uuid.has_value();
                 } catch (...) {
                   return FetchReplicationError::PARSE_ERROR;
                 }
@@ -150,7 +162,9 @@ auto ReplicationState::FetchReplicationData() -> FetchReplicationResult_t {
               return {std::move(res)};
             },
             [&](durability::ReplicaRole &&r) -> FetchReplicationResult_t {
-              return {RoleReplicaData{r.config, std::make_unique<ReplicationServer>(r.config)}};
+              // False positive report for the std::make_unique
+              // NOLINTNEXTLINE(clang-analyzer-cplusplus.NewDeleteLeaks)
+              return {RoleReplicaData{r.config, std::make_unique<ReplicationServer>(r.config), r.main_uuid}};
             },
         },
         std::move(data.role));
@@ -192,21 +206,29 @@ bool ReplicationState::HandleVersionMigration(durability::ReplicationRoleEntry &
       [[fallthrough]];
     }
     case durability::DurabilityVersion::V2: {
-      // do nothing - add code if V3 ever happens
+      if (std::holds_alternative<durability::MainRole>(data.role)) {
+        auto &main = std::get<durability::MainRole>(data.role);
+        main.main_uuid = utils::UUID{};
+      }
+      data.version = durability::DurabilityVersion::V3;
+      break;
+    }
+    case durability::DurabilityVersion::V3: {
+      // do nothing - add code if V4 ever happens
       break;
     }
   }
   return true;
 }
 
-bool ReplicationState::TryPersistRegisteredReplica(const ReplicationClientConfig &config) {
+bool ReplicationState::TryPersistRegisteredReplica(const ReplicationClientConfig &config, utils::UUID main_uuid) {
   if (!HasDurability()) return true;
 
   // If any replicas are persisted then Role must be persisted
   if (role_persisted != RolePersisted::YES) {
     DMG_ASSERT(IsMain(), "MAIN is expected");
     auto epoch_str = std::string(std::get<RoleMainData>(replication_data_).epoch_.id());
-    if (!TryPersistRoleMain(std::move(epoch_str))) return false;
+    if (!TryPersistRoleMain(std::move(epoch_str), main_uuid)) return false;
   }
 
   auto data = durability::ReplicationReplicaEntry{.config = config};
@@ -217,22 +239,28 @@ bool ReplicationState::TryPersistRegisteredReplica(const ReplicationClientConfig
   return false;
 }
 
-bool ReplicationState::SetReplicationRoleMain() {
+bool ReplicationState::SetReplicationRoleMain(const utils::UUID &main_uuid) {
   auto new_epoch = utils::GenerateUUID();
 
-  if (!TryPersistRoleMain(new_epoch)) {
+  if (!TryPersistRoleMain(new_epoch, main_uuid)) {
     return false;
   }
-  replication_data_ = RoleMainData{ReplicationEpoch{new_epoch}};
+
+  replication_data_ = RoleMainData{ReplicationEpoch{new_epoch}, main_uuid};
 
   return true;
 }
 
-bool ReplicationState::SetReplicationRoleReplica(const ReplicationServerConfig &config) {
-  if (!TryPersistRoleReplica(config)) {
+bool ReplicationState::SetReplicationRoleReplica(const ReplicationServerConfig &config,
+                                                 const std::optional<utils::UUID> &main_uuid) {
+  // False positive report for the std::make_unique
+  // NOLINTNEXTLINE(clang-analyzer-cplusplus.NewDeleteLeaks)
+  if (!TryPersistRoleReplica(config, main_uuid)) {
     return false;
   }
-  replication_data_ = RoleReplicaData{config, std::make_unique<ReplicationServer>(config)};
+  // False positive report for the std::make_unique
+  // NOLINTNEXTLINE(clang-analyzer-cplusplus.NewDeleteLeaks)
+  replication_data_ = RoleReplicaData{config, std::make_unique<ReplicationServer>(config), std::nullopt};
   return true;
 }
 
@@ -264,7 +292,7 @@ utils::BasicResult<RegisterReplicaError, ReplicationClient *> ReplicationState::
     }
 
     // Durability
-    if (!TryPersistRegisteredReplica(config)) {
+    if (!TryPersistRegisteredReplica(config, mainData.uuid_)) {
       return RegisterReplicaError::COULD_NOT_BE_PERSISTED;
     }