Compare commits

..

24 Commits

Author SHA1 Message Date
Marko Barišić
b6d630f4c1 Merge branch 'master' into add-mgbuilder-workflow 2024-01-31 12:24:48 +01:00
Deda
f63cf18703 Fix run.sh build call 2023-12-18 19:00:36 +01:00
Marko Barišić
798c73f5de Merge branch 'master' into add-mgbuilder-workflow 2023-12-18 18:57:54 +01:00
Marko Barišić
5e24f9b0b4 Merge branch 'master' into add-mgbuilder-workflow 2023-11-23 10:35:57 +01:00
Marko Barišić
81c63e70d8 Merge branch 'master' into add-mgbuilder-workflow 2023-11-21 11:41:39 +01:00
Deda
90e0f620c4 Fix typo 2023-11-20 18:25:41 +01:00
Deda
84749b2f05 Add architecture argument to run.sh build calls 2023-11-20 18:20:47 +01:00
Deda
9b2cb3d4d8 Add architecture argument to build command 2023-11-20 18:20:12 +01:00
Deda
e46e3832d8 Modify input descriptions 2023-11-20 17:45:08 +01:00
Marko Barišić
647d62a886 Merge branch 'master' into add-mgbuilder-workflow 2023-11-20 17:31:08 +01:00
Deda
6504f04898 Polish workflow 2023-11-20 16:34:26 +01:00
Deda
bf9b95de70 Fix check for existing image 2023-11-20 16:15:18 +01:00
Deda
9f01e71c26 Add check for existing image 2023-11-20 15:51:32 +01:00
Deda
300f4d6163 Add password to unlock keychain 2023-11-20 14:41:43 +01:00
Deda
b7d143ccbe Add step to unlock keychain on macos self-hosted runner 2023-11-20 14:33:43 +01:00
Deda
1bf265dccd Modify run.sh to use single line docker build command 2023-11-20 14:14:11 +01:00
Deda
88c95f2949 Modify run.sh to use docker buildx 2023-11-20 14:03:48 +01:00
Deda
3791fe3585 Convert docker multiline command into singe line 2023-11-20 14:01:00 +01:00
Deda
84ab085183 Fix docker context typo 2023-11-20 13:58:30 +01:00
Deda
68d4884706 Fix multiline command typo 2023-11-20 13:55:44 +01:00
Deda
533d627a73 Fix run.sh argument positions 2023-11-20 13:52:51 +01:00
Deda
bfe61f0abb Run workflow for debian 11 amd64 and arm 2023-11-20 13:48:49 +01:00
Deda
adf2508f88 Add support for publishing builder images 2023-11-20 13:31:13 +01:00
Deda
b612e8a723 Add workflow for building and publishing memgraph/memgraph-builder docker images 2023-11-20 13:27:07 +01:00
261 changed files with 4644 additions and 12097 deletions

View File

@@ -2,7 +2,7 @@
BasedOnStyle: Google
---
Language: Cpp
Standard: c++20
Standard: "c++20"
UseTab: Never
DerivePointerAlignment: false
PointerAlignment: Right

View File

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

View File

@@ -3,6 +3,7 @@ name: Bug report
about: Create a report to help us improve
title: ""
labels: bug
assignees: gitbuda
---
**Memgraph version**

View File

@@ -268,6 +268,7 @@ jobs:
ctest -R memgraph__unit --output-on-failure -j$THREADS
- name: Ensure Kafka and Pulsar are up
if: false
run: |
cd tests/e2e/streams/kafka
docker-compose up -d
@@ -275,6 +276,7 @@ jobs:
docker-compose up -d
- name: Run e2e tests
if: false
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
@@ -283,6 +285,7 @@ jobs:
./run.sh
- name: Ensure Kafka and Pulsar are down
if: false
run: |
cd tests/e2e/streams/kafka
docker-compose down
@@ -383,6 +386,71 @@ jobs:
# multiple paths could be defined
build/logs
experimental_build_mt:
name: "MultiTenancy replication build"
runs-on: [self-hosted, Linux, X64, Diff]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Build release binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Build MT replication experimental binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=Release -D MG_EXPERIMENTAL_REPLICATION_MULTITENANCY=ON ..
make -j$THREADS
- name: Run unit tests
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Run unit tests.
cd build
ctest -R memgraph__unit --output-on-failure -j$THREADS
- name: Run e2e tests
if: false
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
source ve3/bin/activate_e2e
cd e2e
# Just the replication based e2e tests
./run.sh "Replicate multitenancy"
./run.sh "Show"
./run.sh "Show while creating invalid state"
./run.sh "Delete edge replication"
./run.sh "Read-write benchmark"
./run.sh "Index replication"
./run.sh "Constraints"
- name: Save test data
uses: actions/upload-artifact@v4
if: always()
with:
name: "Test data(MultiTenancy replication build)"
path: |
# multiple paths could be defined
build/logs
release_jepsen_test:
name: "Release Jepsen Test"
runs-on: [self-hosted, Linux, X64, Debian10, JepsenControl]

View File

@@ -1,7 +1,4 @@
name: Release Debian 10
concurrency:
group: ${{ github.workflow }}-${{ github.ref_name }}
cancel-in-progress: true
on:
workflow_dispatch:
@@ -13,12 +10,7 @@ on:
options:
- Release
- RelWithDebInfo
push:
branches:
- "release/**"
tags:
- "v*.*.*-rc*"
- "v*.*-rc*"
schedule:
- cron: "0 22 * * *"
@@ -32,7 +24,7 @@ jobs:
community_build:
name: "Community build"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -71,7 +63,7 @@ jobs:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -125,7 +117,7 @@ jobs:
debug_build:
name: "Debug build"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -179,7 +171,7 @@ jobs:
debug_integration_test:
name: "Debug integration tests"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -209,7 +201,7 @@ jobs:
release_build:
name: "Release build"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -280,7 +272,7 @@ jobs:
release_benchmark_tests:
name: "Release Benchmark Tests"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -329,9 +321,10 @@ jobs:
--no-strict
release_e2e_test:
if: false
name: "Release End-to-end Test"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -379,7 +372,7 @@ jobs:
release_durability_stress_tests:
name: "Release durability and stress tests"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -427,7 +420,7 @@ jobs:
release_jepsen_test:
name: "Release Jepsen Test"
runs-on: [self-hosted, Linux, X64, Debian10, JepsenControl]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository

190
.github/workflows/release_mgbuilder.yml vendored Normal file
View File

@@ -0,0 +1,190 @@
name: Publish memgraph-builder docker images
on:
workflow_dispatch:
inputs:
toolchain_version:
type: choice
description: "Toolchain version for the builder"
options:
- v4
- v5
target_os:
type: choice
description: "Target OS for the builder"
default: 'debian-11'
options:
- all
- all-amd64
- all-arm
- amzn-2
- centos-7
- centos-9
- debian-10
- debian-11
- debian-11-arm
- fedora-36
- ubuntu-18_04
- ubuntu-20_04
- ubuntu-22_04
- ubuntu-22_04-arm
force_release:
type: boolean
description: "Overwrite existing image on dockerhub"
required: false
default: false
jobs:
amd_build:
if: ${{ !startsWith(github.event.inputs.target_os, 'all') && !endsWith(github.event.inputs.target_os, '-arm') }}
name: "Build and publish specific amd64 image"
runs-on: [self-hosted, DockerMgBuild, X64]
env:
DOCKER_ORG: "memgraph"
DOCKER_REPO: "memgraph-builder"
DOCKER_IMAGE_TAG: "${{ github.event.inputs.toolchain_version }}_${{ github.event.inputs.target_os }}"
ARCHITECTURE: "amd64"
steps:
- name: "Set up repository"
uses: actions/checkout@v3
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: Log in to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Check if specified image is already pushed
run: |
if [[ $FORCE_RELEASE = true ]]; then
EXISTS=1
else
EXISTS=$(docker manifest inspect $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG > /dev/null; echo $?)
fi
if [[ $EXISTS -eq 0 ]]; then
echo "Image $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG already exists, skipping next step ..."
fi
echo "EXISTS=$EXISTS" >> $GITHUB_ENV
- name: "Build and publish image"
if: env.EXISTS == 1
run: |
echo "Building and publishing $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG"
./release/package/run.sh build ${{ github.event.inputs.toolchain_version }} ${{ github.event.inputs.target_os }} $ARCHITECTURE --publish
arm_build:
if: ${{ !startsWith(github.event.inputs.target_os, 'all') && endsWith(github.event.inputs.target_os, '-arm') }}
name: "Build and publish specific arm image"
runs-on: [self-hosted, DockerMgBuild, ARM64, strange]
env:
DOCKER_ORG: "memgraph"
DOCKER_REPO: "memgraph-builder"
DOCKER_IMAGE_TAG: "${{ github.event.inputs.toolchain_version }}_${{ github.event.inputs.target_os }}"
ARCHITECTURE: "arm64"
steps:
- name: "Set up repository"
uses: actions/checkout@v3
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Unlock keychain for current session" # Required beacuse ARM builds run on macos
run: security -v unlock-keychain -p ${{ secrets.SELF_HOSTED_RUNNER_PASSWORD }} ~/Library/Keychains/login.keychain-db
- name: Log in to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Check if specified image is already pushed
run: |
if [[ $FORCE_RELEASE = true ]]; then
EXISTS=1
else
EXISTS=$(docker manifest inspect $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG > /dev/null; echo $?)
fi
if [[ $EXISTS -eq 0 ]]; then
echo "Image $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG already exists, skipping next step ..."
fi
echo "EXISTS=$EXISTS" >> $GITHUB_ENV
- name: "Build and publish image"
if: env.EXISTS == 1
run: |
echo "Building and publishing $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG"
./release/package/run.sh build ${{ github.event.inputs.toolchain_version }} ${{ github.event.inputs.target_os }} $ARCHITECTURE --publish
all_amd_build:
if: ${{ github.event.inputs.target_os == 'all-amd64' || github.event.inputs.target_os == 'all'}}
name: "Build and publish all amd64 images"
runs-on: [self-hosted, DockerMgBuild, X64]
strategy:
matrix:
target_os: [amzn-2, centos-7, centos-9, debian-10, debian-11, fedora-36, ubuntu-18_04, ubuntu-20_04, ubuntu-22_04]
env:
DOCKER_ORG: "memgraph"
DOCKER_REPO: "memgraph-builder"
DOCKER_IMAGE_TAG: "${{ github.event.inputs.toolchain_version }}_${{ github.event.inputs.target_os }}"
ARCHITECTURE: "amd64"
steps:
- name: "Set up repository"
uses: actions/checkout@v3
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: Log in to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Check if specified image is already pushed
run: |
if [[ $FORCE_RELEASE = true ]]; then
EXISTS=1
else
EXISTS=$(docker manifest inspect $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG > /dev/null; echo $?)
fi
if [[ $EXISTS -eq 0 ]]; then
echo "Image $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG already exists, skipping next step ..."
fi
echo "EXISTS=$EXISTS" >> $GITHUB_ENV
- name: "Build and publish image"
if: env.EXISTS == 1
run: |
echo "Building and publishing $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG"
./release/package/run.sh build ${{ github.event.inputs.toolchain_version }} ${{ github.event.inputs.target_os }} $ARCHITECTURE --publish
all_arm_build:
if: ${{ github.event.inputs.target_os == 'all-arm' || github.event.inputs.target_os == 'all'}}
name: "Build and publish all arm images"
runs-on: [self-hosted, DockerMgBuild, ARM64, strange]
strategy:
matrix:
target_os: [debian-11-arm, ubuntu-22_04-arm]
env:
DOCKER_ORG: "memgraph"
DOCKER_REPO: "memgraph-builder"
DOCKER_IMAGE_TAG: "${{ github.event.inputs.toolchain_version }}_${{ github.event.inputs.target_os }}"
ARCHITECTURE: "arm64"
steps:
- name: "Set up repository"
uses: actions/checkout@v3
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Unlock keychain for current session" # Required beacuse ARM builds run on macos
run: security -v unlock-keychain -p ${{ secrets.SELF_HOSTED_RUNNER_PASSWORD }} ~/Library/Keychains/login.keychain-db
- name: Log in to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKERHUB_USERNAME }}
password: ${{ secrets.DOCKERHUB_TOKEN }}
- name: Check if specified image is already pushed
run: |
if [[ $FORCE_RELEASE = true ]]; then
EXISTS=1
else
EXISTS=$(docker manifest inspect $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG > /dev/null; echo $?)
fi
if [[ $EXISTS -eq 0 ]]; then
echo "Image $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG already exists, skipping next step ..."
fi
echo "EXISTS=$EXISTS" >> $GITHUB_ENV
- name: "Build and publish image"
if: env.EXISTS == 1
run: |
echo "Building and publishing $DOCKER_ORG/$DOCKER_REPO:$DOCKER_IMAGE_TAG"
./release/package/run.sh build ${{ github.event.inputs.toolchain_version }} ${{ github.event.inputs.target_os }} $ARCHITECTURE --publish

View File

@@ -1,7 +1,4 @@
name: Release Ubuntu 20.04
concurrency:
group: ${{ github.workflow }}-${{ github.ref_name }}
cancel-in-progress: true
on:
workflow_dispatch:
@@ -13,12 +10,7 @@ on:
options:
- Release
- RelWithDebInfo
push:
branches:
- "release/**"
tags:
- "v*.*.*-rc*"
- "v*.*-rc*"
schedule:
- cron: "0 22 * * *"
@@ -32,7 +24,7 @@ jobs:
community_build:
name: "Community build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -67,7 +59,7 @@ jobs:
coverage_build:
name: "Coverage build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -121,7 +113,7 @@ jobs:
debug_build:
name: "Debug build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -175,7 +167,7 @@ jobs:
debug_integration_test:
name: "Debug integration tests"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -205,7 +197,7 @@ jobs:
release_build:
name: "Release build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -276,7 +268,7 @@ jobs:
release_benchmark_tests:
name: "Release Benchmark Tests"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -325,9 +317,10 @@ jobs:
--no-strict
release_e2e_test:
if: false
name: "Release End-to-end Test"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository
@@ -375,7 +368,7 @@ jobs:
release_durability_stress_tests:
name: "Release durability and stress tests"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
timeout-minutes: 90
steps:
- name: Set up repository

View File

@@ -1,7 +1,4 @@
name: Stress test large
concurrency:
group: ${{ github.workflow }}-${{ github.ref_name }}
cancel-in-progress: true
on:
workflow_dispatch:
@@ -13,10 +10,7 @@ on:
options:
- Release
- RelWithDebInfo
push:
tags:
- "v*.*.*-rc*"
- "v*.*-rc*"
schedule:
- cron: "0 22 * * *"
@@ -29,7 +23,7 @@ env:
jobs:
stress_test_large:
name: "Stress test large"
timeout-minutes: 720
timeout-minutes: 600
strategy:
matrix:
os: [Debian10, Ubuntu20.04]

View File

@@ -17,10 +17,9 @@ repos:
name: isort (python)
args: ["--profile", "black"]
- repo: https://github.com/pre-commit/mirrors-clang-format
rev: v17.0.6
rev: v13.0.0
hooks:
- id: clang-format
types_or: [c++, c]
# - repo: local
# hooks:
# - id: clang-tidy

View File

@@ -1,38 +0,0 @@
# 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.

View File

@@ -189,7 +189,7 @@ add_custom_target(clean_all
# is easier debugging of compilation and linker flags.
set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
set(CMAKE_CXX_STANDARD 23)
set(CMAKE_CXX_STANDARD 20)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
# c99-designator is disabled because of required mixture of designated and
# non-designated initializers in Python Query Module code (`py_module.cpp`).
@@ -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 can only be used with enterpise version of the code.")
message(FATAL_ERROR "MG_EXPERIMENTAL_HIGH_AVAILABILITY must be used with enterpise version of the code.")
endif ()
if (MG_EXPERIMENTAL_HIGH_AVAILABILITY)
@@ -291,6 +291,16 @@ option(TSAN "Build with Thread Sanitizer. To get a reasonable performance option
option(UBSAN "Build with Undefined Behaviour Sanitizer" OFF)
# Build feature flags
option(MG_EXPERIMENTAL_REPLICATION_MULTITENANCY "Feature flag for experimental replicaition of multitenacy" OFF)
if (NOT MG_ENTERPRISE AND MG_EXPERIMENTAL_REPLICATION_MULTITENANCY)
set(MG_EXPERIMENTAL_REPLICATION_MULTITENANCY OFF)
message(FATAL_ERROR "MG_EXPERIMENTAL_REPLICATION_MULTITENANCY with community edition build isn't possible")
endif ()
if (MG_EXPERIMENTAL_REPLICATION_MULTITENANCY)
add_compile_definitions(MG_EXPERIMENTAL_REPLICATION_MULTITENANCY)
endif ()
if (TEST_COVERAGE)
string(TOLOWER ${CMAKE_BUILD_TYPE} lower_build_type)

2
init
View File

@@ -120,7 +120,7 @@ fi
# develop on them -> pre-commit hook not required -> we can use latest
# packages.
if [ "${DISTRO}" != "centos-7" ] && [ "$DISTRO" != "debian-10" ] && [ "${DISTRO}" != "ubuntu-18.04" ] && [ "${DISTRO}" != "amzn-2" ]; then
python3 -m pip install pre-commit==3.5.*
python3 -m pip install pre-commit
python3 -m pre_commit install
# Install py format tools for usage during the development.
echo "Install black formatter"

View File

@@ -294,10 +294,3 @@ set_path_external_library(jemalloc STATIC
${CMAKE_CURRENT_SOURCE_DIR}/jemalloc/include/)
import_header_library(rangev3 ${CMAKE_CURRENT_SOURCE_DIR}/rangev3/include)
# Setup NuRaft
import_external_library(nuraft STATIC
${CMAKE_CURRENT_SOURCE_DIR}/nuraft/lib/libnuraft.a
${CMAKE_CURRENT_SOURCE_DIR}/nuraft/include/)
find_package(OpenSSL REQUIRED)
target_link_libraries(nuraft INTERFACE ${OPENSSL_LIBRARIES})

View File

@@ -123,10 +123,9 @@ declare -A primary_urls=(
["pulsar"]="http://$local_cache_host/git/pulsar.git"
["librdtsc"]="http://$local_cache_host/git/librdtsc.git"
["ctre"]="http://$local_cache_host/file/hanickadot/compile-time-regular-expressions/v3.7.2/single-header/ctre.hpp"
["absl"]="http://$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://$local_cache_host/git/NuRaft.git"
["absl"]="https://$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"
)
# The goal of secondary urls is to have links to the "source of truth" of
@@ -156,7 +155,6 @@ declare -A secondary_urls=(
["absl"]="https://github.com/abseil/abseil-cpp.git"
["jemalloc"]="https://github.com/jemalloc/jemalloc.git"
["range-v3"]="https://github.com/ericniebler/range-v3.git"
["nuraft"]="https://github.com/eBay/NuRaft.git"
)
# antlr
@@ -270,8 +268,6 @@ 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=$working_dir \
--with-malloc-conf="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000"
@@ -281,10 +277,3 @@ popd
#range-v3 release-0.12.0
range_v3_ref="release-0.12.0"
repo_clone_try_double "${primary_urls[range-v3]}" "${secondary_urls[range-v3]}" "rangev3" "$range_v3_ref"
# NuRaft
nuraft_tag="v2.1.0"
repo_clone_try_double "${primary_urls[nuraft]}" "${secondary_urls[nuraft]}" "nuraft" "$nuraft_tag" true
pushd nuraft
./prepare.sh
popd

View File

@@ -10,15 +10,21 @@ SUPPORTED_OS=(
fedora-36
amzn-2
)
SUPPORTED_TOOLCAHINS=(
v4
v5
)
SUPPORTED_BUILD_TYPES=(
Debug
Release
RelWithDebInfo
)
SUPPORTED_ARCHITECTURES=(
amd64
arm64
)
PROJECT_ROOT="$SCRIPT_DIR/../.."
TOOLCHAIN_VERSION="toolchain-v4"
TOOLCHAIN_VERSION="${TOOLCHAIN_VERSION:-toolchain-v4}"
ACTIVATE_TOOLCHAIN="source /opt/${TOOLCHAIN_VERSION}/activate"
HOST_OUTPUT_DIR="$PROJECT_ROOT/build/output"
@@ -30,6 +36,14 @@ print_help () {
echo " Build types: ${SUPPORTED_BUILD_TYPES[*]}"
exit 1
}
print_help_build () {
echo "$0 build {toolchain_version} {os} {architecture} [--publish]"
echo ""
echo " Toolchain versions: ${SUPPORTED_TOOLCAHINS[*]}"
echo " OSs: ${SUPPORTED_OS[*]}"
echo " Architectures: ${SUPPORTED_ARCHITECTURES[*]}"
exit 1
}
make_package () {
os="$1"
@@ -187,15 +201,26 @@ case "$1" in
build)
shift 1
if [[ "$#" -ne 2 ]]; then
print_help
if [[ "$#" -lt 3 || "$#" -gt 4 ]]; then
print_help_build
fi
# in the vX format, e.g. v5
toolchain_version="$1"
# a name of the os folder, e.g. ubuntu-22.04-arm
os="$2"
cd "$SCRIPT_DIR/$os"
docker build -f Dockerfile --build-arg TOOLCHAIN_VERSION="toolchain-$toolchain_version" -t "memgraph/memgraph-builder:${toolchain_version}_$os" .
# architecture for which the image will be built, e.g. linux/arm64
architecture="$3"
if [[ "$#" -eq 4 ]]; then
if [[ "$4" == "--publish" ]]; then
docker buildx build --file Dockerfile --build-arg TOOLCHAIN_VERSION="toolchain-$toolchain_version" --platform "linux/$architecture" --tag "memgraph/memgraph-builder:${toolchain_version}_$os" --push .
else
print_help_build
exit
fi
else
docker buildx build --file Dockerfile --build-arg TOOLCHAIN_VERSION="toolchain-$toolchain_version" --platform "linux/$architecture" --tag "memgraph/memgraph-builder:${toolchain_version}_$os" .
fi
;;
test)

View File

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

View File

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

View File

@@ -9,16 +9,13 @@
#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"
@@ -44,89 +41,12 @@ 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
@@ -228,12 +148,6 @@ 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) {
@@ -326,7 +240,7 @@ std::optional<User> Auth::GetUser(const std::string &username_orig) const {
return user;
}
void Auth::SaveUser(const User &user, system::Transaction *system_tx) {
void Auth::SaveUser(const User &user) {
bool success = false;
if (const auto *role = user.role(); role != nullptr) {
success = storage_.PutMultiple(
@@ -338,12 +252,6 @@ void Auth::SaveUser(const User &user, system::Transaction *system_tx) {
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) {
@@ -376,8 +284,7 @@ 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,
system::Transaction *system_tx) {
std::optional<User> Auth::AddUser(const std::string &username, const std::optional<std::string> &password) {
if (!NameRegexMatch(username)) {
throw AuthException("Invalid user name.");
}
@@ -387,23 +294,17 @@ 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, system_tx);
SaveUser(new_user);
return new_user;
}
bool Auth::RemoveUser(const std::string &username_orig, system::Transaction *system_tx) {
bool Auth::RemoveUser(const std::string &username_orig) {
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;
}
@@ -420,19 +321,6 @@ 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 {
@@ -450,30 +338,24 @@ std::optional<Role> Auth::GetRole(const std::string &rolename_orig) const {
return Role::Deserialize(data);
}
void Auth::SaveRole(const Role &role, system::Transaction *system_tx) {
void Auth::SaveRole(const Role &role) {
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, system::Transaction *system_tx) {
std::optional<Role> Auth::AddRole(const std::string &rolename) {
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, system_tx);
SaveRole(new_role);
return new_role;
}
bool Auth::RemoveRole(const std::string &rolename_orig, system::Transaction *system_tx) {
bool Auth::RemoveRole(const std::string &rolename_orig) {
auto rolename = utils::ToLowerCase(rolename_orig);
if (!storage_.Get(kRolePrefix + rolename)) return false;
std::vector<std::string> keys;
@@ -486,12 +368,6 @@ bool Auth::RemoveRole(const std::string &rolename_orig, system::Transaction *sys
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;
}
@@ -509,18 +385,6 @@ 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;
@@ -540,48 +404,48 @@ 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, system::Transaction *system_tx) {
bool Auth::GrantDatabaseToUser(const std::string &db, const std::string &name) {
if (auto user = GetUser(name)) {
if (db == kAllDatabases) {
user->db_access().GrantAll();
} else {
user->db_access().Add(db);
}
SaveUser(*user, system_tx);
SaveUser(*user);
return true;
}
return false;
}
bool Auth::RevokeDatabaseFromUser(const std::string &db, const std::string &name, system::Transaction *system_tx) {
bool Auth::RevokeDatabaseFromUser(const std::string &db, const std::string &name) {
if (auto user = GetUser(name)) {
if (db == kAllDatabases) {
user->db_access().DenyAll();
} else {
user->db_access().Remove(db);
}
SaveUser(*user, system_tx);
SaveUser(*user);
return true;
}
return false;
}
void Auth::DeleteDatabase(const std::string &db, system::Transaction *system_tx) {
void Auth::DeleteDatabase(const std::string &db) {
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, system_tx);
SaveUser(*user);
}
}
}
bool Auth::SetMainDatabase(std::string_view db, const std::string &name, system::Transaction *system_tx) {
bool Auth::SetMainDatabase(std::string_view db, const std::string &name) {
if (auto user = GetUser(name)) {
if (!user->db_access().SetDefault(db)) {
throw AuthException("Couldn't set default database '{}' for user '{}'!", db, name);
}
SaveUser(*user, system_tx);
SaveUser(*user);
return true;
}
return false;

View File

@@ -18,15 +18,10 @@
#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 = "*";
/**
@@ -73,13 +68,6 @@ class Auth final {
config_ = std::move(config);
}
/**
* @brief
*
* @return Config
*/
Config GetConfig() const { return config_; }
/**
* Authenticates a user using his username and password.
*
@@ -108,7 +96,7 @@ class Auth final {
*
* @throw AuthException if unable to save the user.
*/
void SaveUser(const User &user, system::Transaction *system_tx = nullptr);
void SaveUser(const User &user);
/**
* Creates a user if the user doesn't exist.
@@ -119,8 +107,7 @@ 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,
system::Transaction *system_tx = nullptr);
std::optional<User> AddUser(const std::string &username, const std::optional<std::string> &password = std::nullopt);
/**
* Removes a user from the storage.
@@ -131,7 +118,7 @@ class Auth final {
* doesn't exist
* @throw AuthException if unable to remove the user.
*/
bool RemoveUser(const std::string &username, system::Transaction *system_tx = nullptr);
bool RemoveUser(const std::string &username);
/**
* @brief
@@ -149,13 +136,6 @@ 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.
*
@@ -180,7 +160,7 @@ class Auth final {
*
* @throw AuthException if unable to save the role.
*/
void SaveRole(const Role &role, system::Transaction *system_tx = nullptr);
void SaveRole(const Role &role);
/**
* Creates a role if the role doesn't exist.
@@ -190,7 +170,7 @@ 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, system::Transaction *system_tx = nullptr);
std::optional<Role> AddRole(const std::string &rolename);
/**
* Removes a role from the storage.
@@ -201,7 +181,7 @@ class Auth final {
* doesn't exist
* @throw AuthException if unable to remove the role.
*/
bool RemoveRole(const std::string &rolename, system::Transaction *system_tx = nullptr);
bool RemoveRole(const std::string &rolename);
/**
* Gets all roles from the storage.
@@ -211,13 +191,6 @@ 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.
*
@@ -237,7 +210,7 @@ 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, system::Transaction *system_tx = nullptr);
bool RevokeDatabaseFromUser(const std::string &db, const std::string &name);
/**
* @brief Grant access to individual database for a user.
@@ -247,7 +220,7 @@ 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, system::Transaction *system_tx = nullptr);
bool GrantDatabaseToUser(const std::string &db, const std::string &name);
/**
* @brief Delete a database from all users.
@@ -255,7 +228,7 @@ class Auth final {
* @param db name of the database to delete
* @throw AuthException if unable to read data
*/
void DeleteDatabase(const std::string &db, system::Transaction *system_tx = nullptr);
void DeleteDatabase(const std::string &db);
/**
* @brief Set main database for an individual user.
@@ -265,7 +238,7 @@ 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, system::Transaction *system_tx = nullptr);
bool SetMainDatabase(std::string_view db, const std::string &name);
#endif
private:

View File

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

View File

@@ -207,8 +207,6 @@ 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
@@ -371,10 +369,6 @@ 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

View File

@@ -1,190 +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 "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

View File

@@ -1,37 +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 "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

View File

@@ -1,182 +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 "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

View File

@@ -1,127 +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 <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

View File

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

View File

@@ -8,35 +8,22 @@ target_sources(mg-coordination
include/coordination/coordinator_server.hpp
include/coordination/coordinator_config.hpp
include/coordination/coordinator_exceptions.hpp
include/coordination/coordinator_slk.hpp
include/coordination/coordinator_instance.hpp
include/coordination/coordinator_handlers.hpp
include/coordination/coordinator_slk.hpp
include/coordination/coordinator_data.hpp
include/coordination/constants.hpp
include/coordination/instance_status.hpp
include/coordination/replication_instance.hpp
include/coordination/raft_state.hpp
include/coordination/rpc_errors.hpp
include/nuraft/coordinator_log_store.hpp
include/nuraft/coordinator_state_machine.hpp
include/nuraft/coordinator_state_manager.hpp
include/coordination/failover_status.hpp
include/coordination/coordinator_cluster_config.hpp
PRIVATE
coordinator_client.cpp
coordinator_state.cpp
coordinator_rpc.cpp
coordinator_server.cpp
coordinator_handlers.cpp
coordinator_instance.cpp
replication_instance.cpp
raft_state.cpp
coordinator_log_store.cpp
coordinator_state_machine.cpp
coordinator_state_manager.cpp
coordinator_data.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 mg-replication_handler
PUBLIC mg::utils mg::rpc mg::slk mg::io mg::repl_coord_glue lib::rangev3
)

View File

@@ -9,7 +9,6 @@
// 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"
@@ -17,156 +16,91 @@
#include "coordination/coordinator_config.hpp"
#include "coordination/coordinator_rpc.hpp"
#include "replication_coordination_glue/messages.hpp"
#include "utils/result.hpp"
namespace memgraph::coordination {
namespace {
auto CreateClientContext(memgraph::coordination::CoordinatorClientConfig const &config)
auto CreateClientContext(const memgraph::coordination::CoordinatorClientConfig &config)
-> communication::ClientContext {
return (config.ssl) ? communication::ClientContext{config.ssl->key_file, config.ssl->cert_file}
: communication::ClientContext{};
}
} // namespace
CoordinatorClient::CoordinatorClient(CoordinatorInstance *coord_instance, CoordinatorClientConfig config,
CoordinatorClient::CoordinatorClient(CoordinatorData *coord_data, 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_instance_{coord_instance},
coord_data_{coord_data},
succ_cb_{std::move(succ_cb)},
fail_cb_{std::move(fail_cb)} {}
auto CoordinatorClient::InstanceName() const -> std::string { return config_.instance_name; }
auto CoordinatorClient::SocketAddress() const -> std::string { return rpc_client_.Endpoint().SocketAddress(); }
auto CoordinatorClient::InstanceDownTimeoutSec() const -> std::chrono::seconds {
return config_.instance_down_timeout_sec;
}
auto CoordinatorClient::InstanceGetUUIDFrequencySec() const -> std::chrono::seconds {
return config_.instance_get_uuid_frequency_sec;
}
void CoordinatorClient::StartFrequentCheck() {
if (instance_checker_.IsRunning()) {
return;
}
MG_ASSERT(config_.instance_health_check_frequency_sec > std::chrono::seconds(0),
MG_ASSERT(config_.health_check_frequency_sec > std::chrono::seconds(0),
"Health check frequency must be greater than 0");
instance_checker_.Run(
config_.instance_name, config_.instance_health_check_frequency_sec,
[this, instance_name = config_.instance_name] {
"Coord checker", 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());
{ // 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);
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);
}
});
}
void CoordinatorClient::StopFrequentCheck() { instance_checker_.Stop(); }
void CoordinatorClient::PauseFrequentCheck() { instance_checker_.Pause(); }
void CoordinatorClient::ResumeFrequentCheck() { instance_checker_.Resume(); }
auto CoordinatorClient::SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void {
succ_cb_ = std::move(succ_cb);
fail_cb_ = std::move(fail_cb);
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::ReplicationClientInfo() const -> ReplClientInfo { return config_.replication_client_info; }
auto CoordinatorClient::ResetReplicationClientInfo() -> void {
// TODO (antoniofilipovic) Sync with Andi on this one
// config_.replication_client_info.reset();
}
auto CoordinatorClient::SendPromoteReplicaToMainRpc(const utils::UUID &uuid,
ReplicationClientsInfo replication_clients_info) const -> bool {
auto CoordinatorClient::SendPromoteReplicaToMainRpc(
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info) const -> bool {
try {
auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(uuid, std::move(replication_clients_info))};
auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(std::move(replication_clients_info))};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful PromoteReplicaToMainRpc response!");
spdlog::error("Failed to receive successful RPC failover response!");
return false;
}
return true;
} catch (rpc::RpcFailedException const &) {
spdlog::error("RPC error occurred while sending PromoteReplicaToMainRpc!");
} catch (const rpc::RpcFailedException &) {
spdlog::error("RPC error occurred while sending failover RPC!");
}
return false;
}
auto CoordinatorClient::DemoteToReplica() const -> bool {
auto const &instance_name = config_.instance_name;
auto CoordinatorClient::SendSetToReplicaRpc(CoordinatorClient::ReplClientInfo replication_client_info) const -> bool {
try {
auto stream{rpc_client_.Stream<DemoteMainToReplicaRpc>(config_.replication_client_info)};
auto stream{rpc_client_.Stream<SetMainToReplicaRpc>(std::move(replication_client_info))};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful RPC response for setting instance {} to replica!", instance_name);
spdlog::error("Failed to set main to replica!");
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("RPC error occurred while sending swapping uuid RPC!");
}
return false;
}
auto CoordinatorClient::SendUnregisterReplicaRpc(std::string const &instance_name) const -> bool {
try {
auto stream{rpc_client_.Stream<UnregisterReplicaRpc>(instance_name)};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful RPC response for unregistering replica!");
return false;
}
return true;
} catch (rpc::RpcFailedException const &) {
spdlog::error("Failed to unregister replica!");
}
return false;
}
auto CoordinatorClient::SendGetInstanceUUIDRpc() const
-> utils::BasicResult<GetInstanceUUIDError, std::optional<utils::UUID>> {
try {
auto stream{rpc_client_.Stream<GetInstanceUUIDRpc>()};
auto res = stream.AwaitResponse();
return res.uuid;
} catch (const rpc::RpcFailedException &) {
spdlog::error("RPC error occured while sending GetInstance UUID RPC");
return GetInstanceUUIDError::RPC_EXCEPTION;
}
}
auto CoordinatorClient::SendEnableWritingOnMainRpc() const -> bool {
try {
auto stream{rpc_client_.Stream<EnableWritingOnMainRpc>()};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful RPC response for enabling writing on main!");
return false;
}
return true;
} catch (rpc::RpcFailedException const &) {
spdlog::error("Failed to enable writing on main!");
spdlog::error("Failed to send failover RPC from coordinator to new main!");
}
return false;
}

View File

@@ -0,0 +1,220 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#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>
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

View File

@@ -1,218 +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 "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);
});
server.Register<coordination::UnregisterReplicaRpc>(
[&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
spdlog::info("Received UnregisterReplicaRpc on coordinator server");
CoordinatorHandlers::UnregisterReplicaHandler(replication_handler, req_reader, res_builder);
});
server.Register<coordination::EnableWritingOnMainRpc>(
[&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
spdlog::info("Received EnableWritingOnMainRpc on coordinator server");
CoordinatorHandlers::EnableWritingOnMainHandler(replication_handler, req_reader, res_builder);
});
server.Register<coordination::GetInstanceUUIDRpc>(
[&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
spdlog::info("Received GetInstanceUUIDRpc on coordinator server");
CoordinatorHandlers::GetInstanceUUIDHandler(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");
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::DemoteMainToReplicaRes{false}, res_builder);
return;
}
slk::Save(coordination::DemoteMainToReplicaRes{true}, res_builder);
}
void CoordinatorHandlers::GetInstanceUUIDHandler(replication::ReplicationHandler &replication_handler,
slk::Reader * /*req_reader*/, slk::Builder *res_builder) {
spdlog::info("Executing GetInstanceUUIDHandler");
slk::Save(coordination::GetInstanceUUIDRes{replication_handler.GetReplicaUUID()}, 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::info("Promote replica to main was success {}", std::string(req.main_uuid_));
slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
}
void CoordinatorHandlers::UnregisterReplicaHandler(replication::ReplicationHandler &replication_handler,
slk::Reader *req_reader, slk::Builder *res_builder) {
if (!replication_handler.IsMain()) {
spdlog::error("Unregistering replica must be performed on main.");
slk::Save(coordination::UnregisterReplicaRes{false}, res_builder);
return;
}
coordination::UnregisterReplicaReq req;
slk::Load(&req, req_reader);
auto res = replication_handler.UnregisterReplica(req.instance_name);
switch (res) {
using enum memgraph::query::UnregisterReplicaResult;
case SUCCESS:
slk::Save(coordination::UnregisterReplicaRes{true}, res_builder);
break;
case NOT_MAIN:
spdlog::error("Unregistering replica must be performed on main.");
slk::Save(coordination::UnregisterReplicaRes{false}, res_builder);
break;
case CAN_NOT_UNREGISTER:
spdlog::error("Could not unregister replica.");
slk::Save(coordination::UnregisterReplicaRes{false}, res_builder);
break;
case COULD_NOT_BE_PERSISTED:
spdlog::error("Could not persist replica unregistration.");
slk::Save(coordination::UnregisterReplicaRes{false}, res_builder);
break;
}
}
void CoordinatorHandlers::EnableWritingOnMainHandler(replication::ReplicationHandler &replication_handler,
slk::Reader * /*req_reader*/, slk::Builder *res_builder) {
if (!replication_handler.IsMain()) {
spdlog::error("Enable writing on main must be performed on main!");
slk::Save(coordination::EnableWritingOnMainRes{false}, res_builder);
return;
}
if (!replication_handler.GetReplState().EnableWritingOnMain()) {
spdlog::error("Enabling writing on main failed!");
slk::Save(coordination::EnableWritingOnMainRes{false}, res_builder);
return;
}
slk::Save(coordination::EnableWritingOnMainRes{true}, res_builder);
}
} // namespace memgraph::dbms
#endif

View File

@@ -1,348 +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 "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 "utils/functional.hpp"
#include <range/v3/view.hpp>
#include <shared_mutex>
namespace memgraph::coordination {
using nuraft::ptr;
using nuraft::srv_config;
CoordinatorInstance::CoordinatorInstance()
: raft_state_(RaftState::MakeRaftState(
[this] { std::ranges::for_each(repl_instances_, &ReplicationInstance::StartFrequentCheck); },
[this] { std::ranges::for_each(repl_instances_, &ReplicationInstance::StopFrequentCheck); })) {
auto find_repl_instance = [](CoordinatorInstance *self,
std::string_view repl_instance_name) -> ReplicationInstance & {
auto repl_instance =
std::ranges::find_if(self->repl_instances_, [repl_instance_name](ReplicationInstance const &instance) {
return instance.InstanceName() == repl_instance_name;
});
MG_ASSERT(repl_instance != self->repl_instances_.end(), "Instance {} not found during callback!",
repl_instance_name);
return *repl_instance;
};
replica_succ_cb_ = [find_repl_instance](CoordinatorInstance *self, std::string_view repl_instance_name) -> void {
auto lock = std::lock_guard{self->coord_instance_lock_};
spdlog::trace("Instance {} performing replica successful callback", repl_instance_name);
auto &repl_instance = find_repl_instance(self, repl_instance_name);
// We need to get replicas UUID from time to time to ensure replica is listening to correct main
// and that it didn't go down for less time than we could notice
// We need to get id of main replica is listening to
// and swap if necessary
if (!repl_instance.EnsureReplicaHasCorrectMainUUID(self->GetMainUUID())) {
spdlog::error("Failed to swap uuid for replica instance {} which is alive", repl_instance.InstanceName());
return;
}
repl_instance.OnSuccessPing();
};
replica_fail_cb_ = [find_repl_instance](CoordinatorInstance *self, std::string_view repl_instance_name) -> void {
auto lock = std::lock_guard{self->coord_instance_lock_};
spdlog::trace("Instance {} performing replica failure callback", repl_instance_name);
auto &repl_instance = find_repl_instance(self, repl_instance_name);
repl_instance.OnFailPing();
};
main_succ_cb_ = [find_repl_instance](CoordinatorInstance *self, std::string_view repl_instance_name) -> void {
auto lock = std::lock_guard{self->coord_instance_lock_};
spdlog::trace("Instance {} performing main successful callback", repl_instance_name);
auto &repl_instance = find_repl_instance(self, repl_instance_name);
if (repl_instance.IsAlive()) {
repl_instance.OnSuccessPing();
return;
}
const auto &repl_instance_uuid = repl_instance.GetMainUUID();
MG_ASSERT(repl_instance_uuid.has_value(), "Instance must have uuid set.");
auto const curr_main_uuid = self->GetMainUUID();
if (curr_main_uuid == repl_instance_uuid.value()) {
if (!repl_instance.EnableWritingOnMain()) {
spdlog::error("Failed to enable writing on main instance {}", repl_instance_name);
return;
}
repl_instance.OnSuccessPing();
return;
}
// TODO(antoniof) make demoteToReplica idempotent since main can be demoted to replica but
// swapUUID can fail
if (repl_instance.DemoteToReplica(self->replica_succ_cb_, self->replica_fail_cb_)) {
repl_instance.OnSuccessPing();
spdlog::info("Instance {} demoted to replica", repl_instance_name);
} else {
spdlog::error("Instance {} failed to become replica", repl_instance_name);
return;
}
if (!repl_instance.SendSwapAndUpdateUUID(curr_main_uuid)) {
spdlog::error(fmt::format("Failed to swap uuid for demoted main instance {}", repl_instance.InstanceName()));
return;
}
};
main_fail_cb_ = [find_repl_instance](CoordinatorInstance *self, std::string_view repl_instance_name) -> void {
auto lock = std::lock_guard{self->coord_instance_lock_};
spdlog::trace("Instance {} performing main failure callback", repl_instance_name);
auto &repl_instance = find_repl_instance(self, repl_instance_name);
repl_instance.OnFailPing();
const auto &repl_instance_uuid = repl_instance.GetMainUUID();
MG_ASSERT(repl_instance_uuid.has_value(), "Instance must have uuid set");
if (!repl_instance.IsAlive() && self->GetMainUUID() == repl_instance_uuid.value()) {
spdlog::info("Cluster without main instance, trying automatic failover");
self->TryFailover(); // TODO: (andi) Initiate failover
}
};
}
auto CoordinatorInstance::ShowInstances() const -> std::vector<InstanceStatus> {
auto const coord_instances = raft_state_.GetAllCoordinators();
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.
};
auto instances_status = utils::fmap(coord_instance_to_status, coord_instances);
{
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;
}
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 new_main = ranges::begin(alive_replicas);
new_main->PauseFrequentCheck();
utils::OnScopeExit scope_exit{[&new_main] { new_main->ResumeFrequentCheck(); }};
auto const is_not_new_main = [&new_main](ReplicationInstance &instance) {
return instance.InstanceName() != new_main->InstanceName();
};
auto const new_main_uuid = utils::UUID{};
// 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_new_main)) {
if (!other_replica_instance.SendSwapAndUpdateUUID(new_main_uuid)) {
spdlog::error(fmt::format("Failed to swap uuid for instance {} which is alive, aborting failover",
other_replica_instance.InstanceName()));
return;
}
}
// TODO: (andi) fmap compliant
ReplicationClientsInfo repl_clients_info;
repl_clients_info.reserve(repl_instances_.size() - 1);
std::ranges::transform(repl_instances_ | ranges::views::filter(is_not_new_main),
std::back_inserter(repl_clients_info), &ReplicationInstance::ReplicationClientInfo);
if (!new_main->PromoteToMain(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;
}
// TODO: (andi) This should be replicated across all coordinator instances with Raft log
SetMainUUID(new_main_uuid);
spdlog::info("Failover successful! Instance {} promoted to main.", new_main->InstanceName());
}
// 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_};
if (std::ranges::any_of(repl_instances_, &ReplicationInstance::IsMain)) {
return SetInstanceToMainCoordinatorStatus::MAIN_ALREADY_EXISTS;
}
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(); }};
auto const is_not_new_main = [&instance_name](ReplicationInstance const &instance) {
return instance.InstanceName() != instance_name;
};
auto const new_main_uuid = utils::UUID{};
for (auto &other_instance : repl_instances_ | ranges::views::filter(is_not_new_main)) {
if (!other_instance.SendSwapAndUpdateUUID(new_main_uuid)) {
spdlog::error(
fmt::format("Failed to swap uuid for instance {}, aborting failover", other_instance.InstanceName()));
return SetInstanceToMainCoordinatorStatus::SWAP_UUID_FAILED;
}
}
ReplicationClientsInfo repl_clients_info;
repl_clients_info.reserve(repl_instances_.size() - 1);
std::ranges::transform(repl_instances_ | ranges::views::filter(is_not_new_main),
std::back_inserter(repl_clients_info), &ReplicationInstance::ReplicationClientInfo);
if (!new_main->PromoteToMain(new_main_uuid, std::move(repl_clients_info), main_succ_cb_, main_fail_cb_)) {
return SetInstanceToMainCoordinatorStatus::COULD_NOT_PROMOTE_TO_MAIN;
}
// TODO: (andi) This should be replicated across all coordinator instances with Raft log
SetMainUUID(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 instance_name = config.instance_name;
auto const name_matches = [&instance_name](ReplicationInstance const &instance) {
return instance.InstanceName() == 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 (!raft_state_.RequestLeadership()) {
return RegisterInstanceCoordinatorStatus::NOT_LEADER;
}
auto const res = raft_state_.AppendRegisterReplicationInstance(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", 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 {}", instance_name, res->get_result_code());
return RegisterInstanceCoordinatorStatus::RAFT_COULD_NOT_APPEND;
}
spdlog::info("Instance {} registered", instance_name);
return RegisterInstanceCoordinatorStatus::SUCCESS;
}
auto CoordinatorInstance::UnregisterReplicationInstance(std::string instance_name)
-> UnregisterInstanceCoordinatorStatus {
auto lock = std::lock_guard{coord_instance_lock_};
auto const name_matches = [&instance_name](ReplicationInstance const &instance) {
return instance.InstanceName() == instance_name;
};
auto inst_to_remove = std::ranges::find_if(repl_instances_, name_matches);
if (inst_to_remove == repl_instances_.end()) {
return UnregisterInstanceCoordinatorStatus::NO_INSTANCE_WITH_NAME;
}
if (inst_to_remove->IsMain() && inst_to_remove->IsAlive()) {
return UnregisterInstanceCoordinatorStatus::IS_MAIN;
}
inst_to_remove->StopFrequentCheck();
auto curr_main = std::ranges::find_if(repl_instances_, &ReplicationInstance::IsMain);
MG_ASSERT(curr_main != repl_instances_.end(), "There must be a main instance when unregistering a replica");
if (!curr_main->SendUnregisterReplicaRpc(instance_name)) {
inst_to_remove->StartFrequentCheck();
return UnregisterInstanceCoordinatorStatus::RPC_FAILED;
}
std::erase_if(repl_instances_, name_matches);
return UnregisterInstanceCoordinatorStatus::SUCCESS;
}
auto CoordinatorInstance::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
-> void {
raft_state_.AddCoordinatorInstance(raft_server_id, raft_port, std::move(raft_address));
}
auto CoordinatorInstance::GetMainUUID() const -> utils::UUID { return main_uuid_; }
// TODO: (andi) Add to the RAFT log.
auto CoordinatorInstance::SetMainUUID(utils::UUID new_uuid) -> void { main_uuid_ = new_uuid; }
} // namespace memgraph::coordination
#endif

View File

@@ -1,201 +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 "nuraft/coordinator_log_store.hpp"
#include "coordination/coordinator_exceptions.hpp"
#include "utils/logging.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];
}
MG_ASSERT(le.get(), "Could not find log entry at index {}", i);
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) {
buf_out->put(static_cast<int32>(entry->size()));
buf_out->put(*entry);
}
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

View File

@@ -36,64 +36,19 @@ void PromoteReplicaToMainRes::Load(PromoteReplicaToMainRes *self, memgraph::slk:
memgraph::slk::Load(self, reader);
}
void DemoteMainToReplicaReq::Save(const DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder) {
void SetMainToReplicaReq::Save(const SetMainToReplicaReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void DemoteMainToReplicaReq::Load(DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader) {
void SetMainToReplicaReq::Load(SetMainToReplicaReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void DemoteMainToReplicaRes::Save(const DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder) {
void SetMainToReplicaRes::Save(const SetMainToReplicaRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void DemoteMainToReplicaRes::Load(DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void UnregisterReplicaReq::Save(UnregisterReplicaReq const &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void UnregisterReplicaReq::Load(UnregisterReplicaReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void UnregisterReplicaRes::Save(UnregisterReplicaRes const &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void UnregisterReplicaRes::Load(UnregisterReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void EnableWritingOnMainRes::Save(EnableWritingOnMainRes const &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void EnableWritingOnMainRes::Load(EnableWritingOnMainRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void EnableWritingOnMainReq::Save(EnableWritingOnMainReq const &self, memgraph::slk::Builder *builder) {}
void EnableWritingOnMainReq::Load(EnableWritingOnMainReq *self, memgraph::slk::Reader *reader) {}
// GetInstanceUUID
void GetInstanceUUIDReq::Save(const GetInstanceUUIDReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void GetInstanceUUIDReq::Load(GetInstanceUUIDReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void GetInstanceUUIDRes::Save(const GetInstanceUUIDRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void GetInstanceUUIDRes::Load(GetInstanceUUIDRes *self, memgraph::slk::Reader *reader) {
void SetMainToReplicaRes::Load(SetMainToReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
@@ -105,35 +60,14 @@ 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::DemoteMainToReplicaReq::kType{utils::TypeId::COORD_SET_REPL_MAIN_REQ,
"CoordDemoteToReplicaReq", nullptr};
constexpr utils::TypeInfo coordination::SetMainToReplicaReq::kType{utils::TypeId::COORD_SET_REPL_MAIN_REQ,
"CoordSetReplMainReq", nullptr};
constexpr utils::TypeInfo coordination::DemoteMainToReplicaRes::kType{utils::TypeId::COORD_SET_REPL_MAIN_RES,
"CoordDemoteToReplicaRes", nullptr};
constexpr utils::TypeInfo coordination::UnregisterReplicaReq::kType{utils::TypeId::COORD_UNREGISTER_REPLICA_REQ,
"UnregisterReplicaReq", nullptr};
constexpr utils::TypeInfo coordination::UnregisterReplicaRes::kType{utils::TypeId::COORD_UNREGISTER_REPLICA_RES,
"UnregisterReplicaRes", nullptr};
constexpr utils::TypeInfo coordination::EnableWritingOnMainReq::kType{utils::TypeId::COORD_ENABLE_WRITING_ON_MAIN_REQ,
"CoordEnableWritingOnMainReq", nullptr};
constexpr utils::TypeInfo coordination::EnableWritingOnMainRes::kType{utils::TypeId::COORD_ENABLE_WRITING_ON_MAIN_RES,
"CoordEnableWritingOnMainRes", nullptr};
constexpr utils::TypeInfo coordination::GetInstanceUUIDReq::kType{utils::TypeId::COORD_GET_UUID_REQ, "CoordGetUUIDReq",
nullptr};
constexpr utils::TypeInfo coordination::GetInstanceUUIDRes::kType{utils::TypeId::COORD_GET_UUID_RES, "CoordGetUUIDRes",
nullptr};
constexpr utils::TypeInfo coordination::SetMainToReplicaRes::kType{utils::TypeId::COORD_SET_REPL_MAIN_RES,
"CoordSetReplMainRes", nullptr};
namespace slk {
// PromoteReplicaToMainRpc
void Save(const memgraph::coordination::PromoteReplicaToMainRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
@@ -143,76 +77,29 @@ 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);
}
// DemoteMainToReplicaRpc
void Save(const memgraph::coordination::DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder) {
void Save(const memgraph::coordination::SetMainToReplicaReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.replication_client_info, builder);
}
void Load(memgraph::coordination::DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader) {
void Load(memgraph::coordination::SetMainToReplicaReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->replication_client_info, reader);
}
void Save(const memgraph::coordination::DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder) {
void Save(const memgraph::coordination::SetMainToReplicaRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
void Load(memgraph::coordination::DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader) {
void Load(memgraph::coordination::SetMainToReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->success, reader);
}
// UnregisterReplicaRpc
void Save(memgraph::coordination::UnregisterReplicaReq const &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.instance_name, builder);
}
void Load(memgraph::coordination::UnregisterReplicaReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->instance_name, reader);
}
void Save(memgraph::coordination::UnregisterReplicaRes const &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
void Load(memgraph::coordination::UnregisterReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->success, reader);
}
void Save(memgraph::coordination::EnableWritingOnMainRes const &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
void Load(memgraph::coordination::EnableWritingOnMainRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->success, reader);
}
// GetInstanceUUIDRpc
void Save(const memgraph::coordination::GetInstanceUUIDReq & /*self*/, memgraph::slk::Builder * /*builder*/) {
/* nothing to serialize*/
}
void Load(memgraph::coordination::GetInstanceUUIDReq * /*self*/, memgraph::slk::Reader * /*reader*/) {
/* nothing to serialize*/
}
void Save(const memgraph::coordination::GetInstanceUUIDRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.uuid, builder);
}
void Load(memgraph::coordination::GetInstanceUUIDRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->uuid, reader);
}
} // namespace slk
} // namespace memgraph

View File

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

View File

@@ -25,7 +25,7 @@
namespace memgraph::coordination {
CoordinatorState::CoordinatorState() {
MG_ASSERT(!(FLAGS_raft_server_id && FLAGS_coordinator_server_port),
MG_ASSERT(!(FLAGS_coordinator && FLAGS_coordinator_server_port),
"Instance cannot be a coordinator and have registered coordinator server.");
spdlog::info("Executing coordinator constructor");
@@ -41,53 +41,43 @@ CoordinatorState::CoordinatorState() {
}
}
auto CoordinatorState::RegisterReplicationInstance(CoordinatorClientConfig config)
-> RegisterInstanceCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
auto CoordinatorState::RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorData>(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](CoordinatorInstance &coordinator_instance) {
return coordinator_instance.RegisterReplicationInstance(config);
}},
memgraph::utils::Overloaded{
[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
return RegisterInstanceCoordinatorStatus::NOT_COORDINATOR;
},
[config](CoordinatorData &coordinator_data) { return coordinator_data.RegisterInstance(config); }},
data_);
}
auto CoordinatorState::UnregisterReplicationInstance(std::string instance_name) -> UnregisterInstanceCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
"Coordinator cannot unregister instance since variant holds wrong alternative");
return std::visit(
memgraph::utils::Overloaded{[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
return UnregisterInstanceCoordinatorStatus::NOT_COORDINATOR;
},
[&instance_name](CoordinatorInstance &coordinator_instance) {
return coordinator_instance.UnregisterReplicationInstance(instance_name);
}},
data_);
}
auto CoordinatorState::SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
auto CoordinatorState::SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorData>(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](CoordinatorInstance &coordinator_instance) {
return coordinator_instance.SetReplicationInstanceToMain(instance_name);
[&instance_name](CoordinatorData &coordinator_data) {
return coordinator_data.SetInstanceToMain(instance_name);
}},
data_);
}
auto CoordinatorState::ShowInstances() const -> std::vector<InstanceStatus> {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
auto CoordinatorState::ShowInstances() const -> std::vector<CoordinatorInstanceStatus> {
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
"Can't call show instances on data_, as variant holds wrong alternative");
return std::get<CoordinatorInstance>(data_).ShowInstances();
return std::get<CoordinatorData>(data_).ShowInstances();
}
[[nodiscard]] auto CoordinatorState::DoFailover() -> DoFailoverStatus {
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_), "Cannot do failover since variant holds wrong alternative");
auto &coord_state = std::get<CoordinatorData>(data_);
return coord_state.DoFailover();
}
auto CoordinatorState::GetCoordinatorServer() const -> CoordinatorServer & {
@@ -95,13 +85,5 @@ 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

View File

@@ -1,111 +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 "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

View File

@@ -1,68 +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 "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

View File

@@ -15,21 +15,20 @@
#include "coordination/coordinator_config.hpp"
#include "rpc/client.hpp"
#include "rpc_errors.hpp"
#include "utils/result.hpp"
#include "utils/scheduler.hpp"
#include "utils/uuid.hpp"
namespace memgraph::coordination {
class CoordinatorInstance;
using HealthCheckCallback = std::function<void(CoordinatorInstance *, std::string_view)>;
using ReplicationClientsInfo = std::vector<ReplClientInfo>;
class CoordinatorData;
using HealthCheckCallback = std::function<void(CoordinatorData *, std::string_view)>;
class CoordinatorClient {
public:
explicit CoordinatorClient(CoordinatorInstance *coord_instance, CoordinatorClientConfig config,
HealthCheckCallback succ_cb, HealthCheckCallback fail_cb);
using ReplClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
using ReplicationClientsInfo = std::vector<ReplClientInfo>;
explicit CoordinatorClient(CoordinatorData *coord_data_, CoordinatorClientConfig config, HealthCheckCallback succ_cb,
HealthCheckCallback fail_cb);
~CoordinatorClient() = default;
@@ -47,28 +46,15 @@ class CoordinatorClient {
auto InstanceName() const -> std::string;
auto SocketAddress() const -> std::string;
[[nodiscard]] auto DemoteToReplica() const -> bool;
auto SendPromoteReplicaToMainRpc(ReplicationClientsInfo replication_clients_info) 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 SendUnregisterReplicaRpc(std::string const &instance_name) const -> bool;
auto SendEnableWritingOnMainRpc() const -> bool;
auto SendGetInstanceUUIDRpc() const -> memgraph::utils::BasicResult<GetInstanceUUIDError, std::optional<utils::UUID>>;
auto ReplicationClientInfo() const -> ReplClientInfo;
auto SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void;
auto RpcClient() -> rpc::Client & { return rpc_client_; }
auto InstanceDownTimeoutSec() const -> std::chrono::seconds;
auto InstanceGetUUIDFrequencySec() const -> std::chrono::seconds;
auto SetSuccCallback(HealthCheckCallback succ_cb) -> void;
auto SetFailCallback(HealthCheckCallback fail_cb) -> void;
friend bool operator==(CoordinatorClient const &first, CoordinatorClient const &second) {
return first.config_ == second.config_;
@@ -77,11 +63,12 @@ class CoordinatorClient {
private:
utils::Scheduler instance_checker_;
// TODO: (andi) Pimpl?
communication::ClientContext rpc_context_;
mutable rpc::Client rpc_client_;
CoordinatorClientConfig config_;
CoordinatorInstance *coord_instance_;
CoordinatorData *coord_data_;
HealthCheckCallback succ_cb_;
HealthCheckCallback fail_cb_;
};

View File

@@ -9,6 +9,14 @@
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
namespace memgraph::coordination {
enum class GetInstanceUUIDError { NO_RESPONSE, RPC_EXCEPTION };
struct CoordinatorClusterConfig {
static constexpr int alive_response_time_difference_sec_{5};
};
} // namespace memgraph::coordination
#endif

View File

@@ -28,13 +28,13 @@ struct CoordinatorClientConfig {
std::string instance_name;
std::string ip_address;
uint16_t port{};
std::chrono::seconds instance_health_check_frequency_sec{1};
std::chrono::seconds instance_down_timeout_sec{5};
std::chrono::seconds instance_get_uuid_frequency_sec{10};
std::chrono::seconds health_check_frequency_sec{1};
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;
@@ -43,6 +43,7 @@ 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 {
@@ -57,8 +58,6 @@ struct CoordinatorClientConfig {
friend bool operator==(CoordinatorClientConfig const &, CoordinatorClientConfig const &) = default;
};
using ReplClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
struct CoordinatorServerConfig {
std::string ip_address;
uint16_t port{};

View File

@@ -0,0 +1,49 @@
// 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

View File

@@ -16,71 +16,16 @@
#include "utils/exceptions.hpp"
namespace memgraph::coordination {
class CoordinatorRegisterInstanceException final : public utils::BasicException {
class CoordinatorFailoverException final : public utils::BasicException {
public:
explicit CoordinatorRegisterInstanceException(const std::string_view what) noexcept
: BasicException("Failed to create instance: " + std::string(what)) {}
explicit CoordinatorFailoverException(const std::string_view what) noexcept
: BasicException("Failover didn't complete successfully: " + std::string(what)) {}
template <class... Args>
explicit CoordinatorRegisterInstanceException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: CoordinatorRegisterInstanceException(fmt::format(fmt, std::forward<Args>(args)...)) {}
explicit CoordinatorFailoverException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: CoordinatorFailoverException(fmt::format(fmt, std::forward<Args>(args)...)) {}
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)
};
class RaftCouldNotParseFlagsException final : public utils::BasicException {
public:
explicit RaftCouldNotParseFlagsException(std::string_view what) noexcept : BasicException(what) {}
template <class... Args>
explicit RaftCouldNotParseFlagsException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: RaftCouldNotParseFlagsException(fmt::format(fmt, std::forward<Args>(args)...)) {}
SPECIALIZE_GET_EXCEPTION_NAME(RaftCouldNotParseFlagsException)
SPECIALIZE_GET_EXCEPTION_NAME(CoordinatorFailoverException)
};
} // namespace memgraph::coordination

View File

@@ -1,48 +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_server.hpp"
#include "replication_handler/replication_handler.hpp"
#include "slk/streams.hpp"
namespace memgraph::dbms {
class DbmsHandler;
class CoordinatorHandlers {
public:
static void Register(memgraph::coordination::CoordinatorServer &server,
replication::ReplicationHandler &replication_handler);
private:
static void PromoteReplicaToMainHandler(replication::ReplicationHandler &replication_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);
static void UnregisterReplicaHandler(replication::ReplicationHandler &replication_handler, slk::Reader *req_reader,
slk::Builder *res_builder);
static void EnableWritingOnMainHandler(replication::ReplicationHandler &replication_handler, slk::Reader *req_reader,
slk::Builder *res_builder);
static void GetInstanceUUIDHandler(replication::ReplicationHandler &replication_handler, slk::Reader *req_reader,
slk::Builder *res_builder);
};
} // namespace memgraph::dbms
#endif

View File

@@ -13,47 +13,64 @@
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_server.hpp"
#include "coordination/instance_status.hpp"
#include "coordination/raft_state.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>
#include "coordination/coordinator_client.hpp"
#include "coordination/coordinator_cluster_config.hpp"
#include "replication_coordination_glue/role.hpp"
namespace memgraph::coordination {
class CoordinatorData;
class CoordinatorInstance {
public:
CoordinatorInstance();
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) {}
[[nodiscard]] auto RegisterReplicationInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
[[nodiscard]] auto UnregisterReplicationInstance(std::string instance_name) -> UnregisterInstanceCoordinatorStatus;
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 SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
auto UpdateInstanceStatus() -> bool {
is_alive_ = std::chrono::duration_cast<std::chrono::seconds>(std::chrono::system_clock::now() - last_response_time_)
.count() < CoordinatorClusterConfig::alive_response_time_difference_sec_;
return is_alive_;
}
auto UpdateLastResponseTime() -> void { last_response_time_ = std::chrono::system_clock::now(); }
auto ShowInstances() const -> std::vector<InstanceStatus>;
auto InstanceName() const -> std::string { return client_.InstanceName(); }
auto SocketAddress() const -> std::string { return client_.SocketAddress(); }
auto IsAlive() const -> bool { return is_alive_; }
auto TryFailover() -> void;
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 AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
auto PrepareForFailover() -> void { client_.PauseFrequentCheck(); }
auto RestoreAfterFailedFailover() -> void { client_.ResumeFrequentCheck(); }
auto GetMainUUID() const -> utils::UUID;
auto PostFailover(HealthCheckCallback main_succ_cb, HealthCheckCallback main_fail_cb) -> void {
replication_role_ = replication_coordination_glue::ReplicationRole::MAIN;
client_.SetSuccCallback(std::move(main_succ_cb));
client_.SetFailCallback(std::move(main_fail_cb));
// Comment with Andi but we shouldn't delete this, what if this MAIN FAILS AGAIN
// client_.ResetReplicationClientInfo();
client_.ResumeFrequentCheck();
}
auto SetMainUUID(utils::UUID new_uuid) -> void;
CoordinatorClient client_;
replication_coordination_glue::ReplicationRole replication_role_;
std::chrono::system_clock::time_point last_response_time_{};
bool is_alive_{false};
private:
HealthCheckCallback main_succ_cb_, main_fail_cb_, replica_succ_cb_, replica_fail_cb_;
// 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_;
RaftState raft_state_;
friend bool operator==(CoordinatorInstance const &first, CoordinatorInstance const &second) {
return first.client_ == second.client_ && first.replication_role_ == second.replication_role_;
}
};
} // namespace memgraph::coordination

View File

@@ -19,13 +19,10 @@
namespace memgraph::coordination {
// 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 {
struct CoordinatorInstanceStatus {
std::string instance_name;
std::string raft_socket_address;
std::string coord_socket_address;
std::string cluster_role;
std::string socket_address;
std::string replication_role;
bool is_alive;
};

View File

@@ -11,7 +11,6 @@
#pragma once
#include "utils/uuid.hpp"
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_config.hpp"
@@ -27,13 +26,10 @@ struct PromoteReplicaToMainReq {
static void Load(PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader);
static void Save(const PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder);
explicit PromoteReplicaToMainReq(const utils::UUID &uuid,
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info)
: main_uuid_(uuid), replication_clients_info(std::move(replication_clients_info)) {}
explicit PromoteReplicaToMainReq(std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info)
: replication_clients_info(std::move(replication_clients_info)) {}
PromoteReplicaToMainReq() = default;
// get uuid here
utils::UUID main_uuid_;
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info;
};
@@ -52,145 +48,56 @@ struct PromoteReplicaToMainRes {
using PromoteReplicaToMainRpc = rpc::RequestResponse<PromoteReplicaToMainReq, PromoteReplicaToMainRes>;
struct DemoteMainToReplicaReq {
struct SetMainToReplicaReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader);
static void Save(const DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder);
static void Load(SetMainToReplicaReq *self, memgraph::slk::Reader *reader);
static void Save(const SetMainToReplicaReq &self, memgraph::slk::Builder *builder);
explicit DemoteMainToReplicaReq(CoordinatorClientConfig::ReplicationClientInfo replication_client_info)
explicit SetMainToReplicaReq(CoordinatorClientConfig::ReplicationClientInfo replication_client_info)
: replication_client_info(std::move(replication_client_info)) {}
DemoteMainToReplicaReq() = default;
SetMainToReplicaReq() = default;
CoordinatorClientConfig::ReplicationClientInfo replication_client_info;
};
struct DemoteMainToReplicaRes {
struct SetMainToReplicaRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader);
static void Save(const DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder);
static void Load(SetMainToReplicaRes *self, memgraph::slk::Reader *reader);
static void Save(const SetMainToReplicaRes &self, memgraph::slk::Builder *builder);
explicit DemoteMainToReplicaRes(bool success) : success(success) {}
DemoteMainToReplicaRes() = default;
explicit SetMainToReplicaRes(bool success) : success(success) {}
SetMainToReplicaRes() = default;
bool success;
};
using DemoteMainToReplicaRpc = rpc::RequestResponse<DemoteMainToReplicaReq, DemoteMainToReplicaRes>;
struct UnregisterReplicaReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(UnregisterReplicaReq *self, memgraph::slk::Reader *reader);
static void Save(UnregisterReplicaReq const &self, memgraph::slk::Builder *builder);
explicit UnregisterReplicaReq(std::string instance_name) : instance_name(std::move(instance_name)) {}
UnregisterReplicaReq() = default;
std::string instance_name;
};
struct UnregisterReplicaRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(UnregisterReplicaRes *self, memgraph::slk::Reader *reader);
static void Save(const UnregisterReplicaRes &self, memgraph::slk::Builder *builder);
explicit UnregisterReplicaRes(bool success) : success(success) {}
UnregisterReplicaRes() = default;
bool success;
};
using UnregisterReplicaRpc = rpc::RequestResponse<UnregisterReplicaReq, UnregisterReplicaRes>;
struct EnableWritingOnMainReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(EnableWritingOnMainReq *self, memgraph::slk::Reader *reader);
static void Save(EnableWritingOnMainReq const &self, memgraph::slk::Builder *builder);
EnableWritingOnMainReq() = default;
};
struct EnableWritingOnMainRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(EnableWritingOnMainRes *self, memgraph::slk::Reader *reader);
static void Save(EnableWritingOnMainRes const &self, memgraph::slk::Builder *builder);
explicit EnableWritingOnMainRes(bool success) : success(success) {}
EnableWritingOnMainRes() = default;
bool success;
};
using EnableWritingOnMainRpc = rpc::RequestResponse<EnableWritingOnMainReq, EnableWritingOnMainRes>;
struct GetInstanceUUIDReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(GetInstanceUUIDReq *self, memgraph::slk::Reader *reader);
static void Save(const GetInstanceUUIDReq &self, memgraph::slk::Builder *builder);
GetInstanceUUIDReq() = default;
};
struct GetInstanceUUIDRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(GetInstanceUUIDRes *self, memgraph::slk::Reader *reader);
static void Save(const GetInstanceUUIDRes &self, memgraph::slk::Builder *builder);
explicit GetInstanceUUIDRes(std::optional<utils::UUID> uuid) : uuid(uuid) {}
GetInstanceUUIDRes() = default;
std::optional<utils::UUID> uuid;
};
using GetInstanceUUIDRpc = rpc::RequestResponse<GetInstanceUUIDReq, GetInstanceUUIDRes>;
using SetMainToReplicaRpc = rpc::RequestResponse<SetMainToReplicaReq, SetMainToReplicaRes>;
} // 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);
// 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 Save(const memgraph::coordination::SetMainToReplicaRes &self, memgraph::slk::Builder *builder);
// GetInstanceUUIDRpc
void Save(const memgraph::coordination::GetInstanceUUIDReq &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::GetInstanceUUIDReq *self, memgraph::slk::Reader *reader);
void Save(const memgraph::coordination::GetInstanceUUIDRes &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::GetInstanceUUIDRes *self, memgraph::slk::Reader *reader);
// UnregisterReplicaRpc
void Save(memgraph::coordination::UnregisterReplicaRes const &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::UnregisterReplicaRes *self, memgraph::slk::Reader *reader);
void Save(memgraph::coordination::UnregisterReplicaReq const &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::UnregisterReplicaReq *self, memgraph::slk::Reader *reader);
void Load(memgraph::coordination::SetMainToReplicaRes *self, memgraph::slk::Reader *reader);
void Save(memgraph::coordination::EnableWritingOnMainRes const &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::EnableWritingOnMainRes *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

View File

@@ -13,9 +13,10 @@
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_instance.hpp"
#include "coordination/coordinator_data.hpp"
#include "coordination/coordinator_instance_status.hpp"
#include "coordination/coordinator_server.hpp"
#include "coordination/instance_status.hpp"
#include "coordination/failover_status.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
#include <variant>
@@ -27,30 +28,25 @@ class CoordinatorState {
CoordinatorState();
~CoordinatorState() = default;
CoordinatorState(CoordinatorState const &) = delete;
CoordinatorState &operator=(CoordinatorState const &) = delete;
CoordinatorState(const CoordinatorState &) = delete;
CoordinatorState &operator=(const CoordinatorState &) = delete;
CoordinatorState(CoordinatorState &&) noexcept = delete;
CoordinatorState &operator=(CoordinatorState &&) noexcept = delete;
[[nodiscard]] auto RegisterReplicationInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
[[nodiscard]] auto UnregisterReplicationInstance(std::string instance_name) -> UnregisterInstanceCoordinatorStatus;
[[nodiscard]] auto RegisterInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
[[nodiscard]] auto SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
[[nodiscard]] auto SetInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
auto ShowInstances() const -> std::vector<InstanceStatus>;
auto ShowInstances() const -> std::vector<CoordinatorInstanceStatus>;
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.
// The client code must check that the server exists before calling this method.
auto GetCoordinatorServer() const -> CoordinatorServer &;
private:
struct CoordinatorMainReplicaData {
std::unique_ptr<CoordinatorServer> coordinator_server_;
};
[[nodiscard]] auto DoFailover() -> DoFailoverStatus;
std::variant<CoordinatorInstance, CoordinatorMainReplicaData> data_;
private:
std::variant<CoordinatorData, CoordinatorMainReplicaData> data_;
};
} // namespace memgraph::coordination

View File

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

View File

@@ -1,79 +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 <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 RaftState {
private:
explicit RaftState(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb, uint32_t raft_server_id,
uint32_t raft_port, std::string raft_address);
auto InitRaftServer() -> void;
public:
RaftState() = delete;
RaftState(RaftState const &other) = default;
RaftState &operator=(RaftState const &other) = default;
RaftState(RaftState &&other) noexcept = default;
RaftState &operator=(RaftState &&other) noexcept = default;
~RaftState();
static auto MakeRaftState(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb) -> RaftState;
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>;
// TODO: (andi) I think variables below can be abstracted
uint32_t raft_server_id_;
uint32_t raft_port_;
std::string raft_address_;
ptr<state_machine> state_machine_;
ptr<state_mgr> state_manager_;
ptr<raft_server> raft_server_;
ptr<logger> logger_;
raft_launcher launcher_;
BecomeLeaderCb become_leader_cb_;
BecomeFollowerCb become_follower_cb_;
};
} // namespace memgraph::coordination
#endif

View File

@@ -19,31 +19,18 @@ namespace memgraph::coordination {
enum class RegisterInstanceCoordinatorStatus : uint8_t {
NAME_EXISTS,
ENDPOINT_EXISTS,
END_POINT_EXISTS,
COULD_NOT_BE_PERSISTED,
NOT_COORDINATOR,
RPC_FAILED,
NOT_LEADER,
RAFT_COULD_NOT_ACCEPT,
RAFT_COULD_NOT_APPEND,
SUCCESS
};
enum class UnregisterInstanceCoordinatorStatus : uint8_t {
NO_INSTANCE_WITH_NAME,
IS_MAIN,
NOT_COORDINATOR,
NOT_LEADER,
RPC_FAILED,
SUCCESS,
};
enum class SetInstanceToMainCoordinatorStatus : uint8_t {
NO_INSTANCE_WITH_NAME,
MAIN_ALREADY_EXISTS,
NOT_COORDINATOR,
SUCCESS,
COULD_NOT_PROMOTE_TO_MAIN,
SWAP_UUID_FAILED
};
} // namespace memgraph::coordination

View File

@@ -1,99 +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_client.hpp"
#include "coordination/coordinator_exceptions.hpp"
#include "replication_coordination_glue/role.hpp"
#include <libnuraft/nuraft.hxx>
#include "utils/result.hpp"
#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 IsReadyForUUIDPing() -> bool;
void UpdateReplicaLastResponseUUID();
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 EnsureReplicaHasCorrectMainUUID(utils::UUID const &curr_main_uuid) -> bool;
auto SendSwapAndUpdateUUID(const utils::UUID &new_main_uuid) -> bool;
auto SendUnregisterReplicaRpc(std::string const &instance_name) -> bool;
auto SendGetInstanceUUID() -> utils::BasicResult<coordination::GetInstanceUUIDError, std::optional<utils::UUID>>;
auto GetClient() -> CoordinatorClient &;
auto EnableWritingOnMain() -> bool;
auto SetNewMainUUID(utils::UUID const &main_uuid) -> void;
auto ResetMainUUID() -> void;
auto GetMainUUID() const -> 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};
std::chrono::system_clock::time_point last_check_of_uuid_{};
// 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

View File

@@ -1,70 +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 <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

View File

@@ -1,76 +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 <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

View File

@@ -1,66 +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 "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

View File

@@ -1,140 +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 "coordination/raft_state.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>>;
RaftState::RaftState(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb, uint32_t raft_server_id,
uint32_t raft_port, std::string raft_address)
: raft_server_id_(raft_server_id),
raft_port_(raft_port),
raft_address_(std::move(raft_address)),
state_machine_(cs_new<CoordinatorStateMachine>()),
state_manager_(
cs_new<CoordinatorStateManager>(raft_server_id_, raft_address_ + ":" + std::to_string(raft_port_))),
logger_(nullptr),
become_leader_cb_(std::move(become_leader_cb)),
become_follower_cb_(std::move(become_follower_cb)) {}
auto RaftState::InitRaftServer() -> void {
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_launcher launcher;
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_address_, raft_port_);
}
auto maybe_stop = utils::ResettableCounter<20>();
do {
if (raft_server_->is_initialized()) {
return;
}
std::this_thread::sleep_for(std::chrono::milliseconds(250));
} while (!maybe_stop());
throw RaftServerStartException("Failed to initialize raft server on {}:{}", raft_address_, raft_port_);
}
auto RaftState::MakeRaftState(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb) -> RaftState {
uint32_t raft_server_id{0};
uint32_t raft_port{0};
try {
raft_server_id = FLAGS_raft_server_id;
raft_port = FLAGS_raft_server_port;
} catch (std::exception const &e) {
throw RaftCouldNotParseFlagsException("Failed to parse flags: {}", e.what());
}
auto raft_state =
RaftState(std::move(become_leader_cb), std::move(become_follower_cb), raft_server_id, raft_port, "127.0.0.1");
raft_state.InitRaftServer();
return raft_state;
}
RaftState::~RaftState() { launcher_.shutdown(); }
auto RaftState::InstanceName() const -> std::string { return "coordinator_" + std::to_string(raft_server_id_); }
auto RaftState::RaftSocketAddress() const -> std::string { return raft_address_ + ":" + std::to_string(raft_port_); }
auto RaftState::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 RaftState::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 RaftState::IsLeader() const -> bool { return raft_server_->is_leader(); }
auto RaftState::RequestLeadership() -> bool { return raft_server_->is_leader() || raft_server_->request_leadership(); }
auto RaftState::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

View File

@@ -1,138 +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 "coordination/replication_instance.hpp"
#include "replication_coordination_glue/handler.hpp"
#include "utils/result.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 {
auto elapsed_time = std::chrono::system_clock::now() - last_response_time_;
is_alive_ = elapsed_time < client_.InstanceDownTimeoutSec();
return is_alive_;
}
auto ReplicationInstance::IsReadyForUUIDPing() -> bool {
return std::chrono::duration_cast<std::chrono::seconds>(std::chrono::system_clock::now() - last_check_of_uuid_) >
client_.InstanceGetUUIDFrequencySec();
}
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 new_uuid, ReplicationClientsInfo repl_clients_info,
HealthCheckCallback main_succ_cb, HealthCheckCallback main_fail_cb) -> bool {
if (!client_.SendPromoteReplicaToMainRpc(new_uuid, std::move(repl_clients_info))) {
return false;
}
replication_role_ = replication_coordination_glue::ReplicationRole::MAIN;
main_uuid_ = new_uuid;
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_; }
auto ReplicationInstance::SetNewMainUUID(utils::UUID const &main_uuid) -> void { main_uuid_ = main_uuid; }
auto ReplicationInstance::ResetMainUUID() -> void { main_uuid_ = std::nullopt; }
auto ReplicationInstance::GetMainUUID() const -> std::optional<utils::UUID> const & { return main_uuid_; }
auto ReplicationInstance::EnsureReplicaHasCorrectMainUUID(utils::UUID const &curr_main_uuid) -> bool {
if (!IsReadyForUUIDPing()) {
return true;
}
auto res = SendGetInstanceUUID();
if (res.HasError()) {
return false;
}
UpdateReplicaLastResponseUUID();
if (res.GetValue().has_value() && res.GetValue().value() == curr_main_uuid) {
return true;
}
return SendSwapAndUpdateUUID(curr_main_uuid);
}
auto ReplicationInstance::SendSwapAndUpdateUUID(const utils::UUID &new_main_uuid) -> bool {
if (!replication_coordination_glue::SendSwapMainUUIDRpc(client_.RpcClient(), new_main_uuid)) {
return false;
}
SetNewMainUUID(new_main_uuid);
return true;
}
auto ReplicationInstance::SendUnregisterReplicaRpc(std::string const &instance_name) -> bool {
return client_.SendUnregisterReplicaRpc(instance_name);
}
auto ReplicationInstance::EnableWritingOnMain() -> bool { return client_.SendEnableWritingOnMainRpc(); }
auto ReplicationInstance::SendGetInstanceUUID()
-> utils::BasicResult<coordination::GetInstanceUUIDError, std::optional<utils::UUID>> {
return client_.SendGetInstanceUUIDRpc();
}
void ReplicationInstance::UpdateReplicaLastResponseUUID() { last_check_of_uuid_ = std::chrono::system_clock::now(); }
} // namespace memgraph::coordination
#endif

View File

@@ -1,10 +1,2 @@
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)
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)

View File

@@ -16,4 +16,10 @@ namespace memgraph::dbms {
constexpr std::string_view kDefaultDB = "memgraph"; //!< Name of the default database
constexpr std::string_view kMultiTenantDir = "databases"; //!< Name of the multi-tenant directory
#ifdef MG_EXPERIMENTAL_REPLICATION_MULTITENANCY
constexpr bool allow_mt_repl = true;
#else
constexpr bool allow_mt_repl = false;
#endif
} // namespace memgraph::dbms

View File

@@ -9,41 +9,30 @@
// 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 "coordination/register_main_replica_coordinator_status.hpp"
#include "dbms/dbms_handler.hpp"
namespace memgraph::dbms {
CoordinatorHandler::CoordinatorHandler(coordination::CoordinatorState &coordinator_state)
: coordinator_state_(coordinator_state) {}
CoordinatorHandler::CoordinatorHandler(DbmsHandler &dbms_handler) : dbms_handler_(dbms_handler) {}
auto CoordinatorHandler::RegisterReplicationInstance(memgraph::coordination::CoordinatorClientConfig config)
auto CoordinatorHandler::RegisterInstance(memgraph::coordination::CoordinatorClientConfig config)
-> coordination::RegisterInstanceCoordinatorStatus {
return coordinator_state_.RegisterReplicationInstance(config);
return dbms_handler_.CoordinatorState().RegisterInstance(config);
}
auto CoordinatorHandler::UnregisterReplicationInstance(std::string instance_name)
-> coordination::UnregisterInstanceCoordinatorStatus {
return coordinator_state_.UnregisterReplicationInstance(std::move(instance_name));
}
auto CoordinatorHandler::SetReplicationInstanceToMain(std::string instance_name)
auto CoordinatorHandler::SetInstanceToMain(std::string instance_name)
-> coordination::SetInstanceToMainCoordinatorStatus {
return coordinator_state_.SetReplicationInstanceToMain(std::move(instance_name));
return dbms_handler_.CoordinatorState().SetInstanceToMain(std::move(instance_name));
}
auto CoordinatorHandler::ShowInstances() const -> std::vector<coordination::InstanceStatus> {
return coordinator_state_.ShowInstances();
auto CoordinatorHandler::ShowInstances() const -> std::vector<coordination::CoordinatorInstanceStatus> {
return dbms_handler_.CoordinatorState().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

View File

@@ -13,11 +13,15 @@
#ifdef MG_ENTERPRISE
#include "utils/result.hpp"
#include "coordination/coordinator_config.hpp"
#include "coordination/coordinator_state.hpp"
#include "coordination/instance_status.hpp"
#include "coordination/coordinator_instance_status.hpp"
#include "coordination/failover_status.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
#include <cstdint>
#include <optional>
#include <vector>
namespace memgraph::dbms {
@@ -26,23 +30,17 @@ class DbmsHandler;
class CoordinatorHandler {
public:
explicit CoordinatorHandler(coordination::CoordinatorState &coordinator_state);
explicit CoordinatorHandler(DbmsHandler &dbms_handler);
// TODO: (andi) When moving coordinator state on same instances, rename from RegisterReplicationInstance to
// RegisterInstance
auto RegisterReplicationInstance(coordination::CoordinatorClientConfig config)
auto RegisterInstance(coordination::CoordinatorClientConfig config)
-> coordination::RegisterInstanceCoordinatorStatus;
auto UnregisterReplicationInstance(std::string instance_name) -> coordination::UnregisterInstanceCoordinatorStatus;
auto SetInstanceToMain(std::string instance_name) -> coordination::SetInstanceToMainCoordinatorStatus;
auto SetReplicationInstanceToMain(std::string instance_name) -> coordination::SetInstanceToMainCoordinatorStatus;
auto ShowInstances() const -> std::vector<coordination::InstanceStatus>;
auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
auto ShowInstances() const -> std::vector<coordination::CoordinatorInstanceStatus>;
private:
coordination::CoordinatorState &coordinator_state_;
DbmsHandler &dbms_handler_;
};
} // namespace memgraph::dbms

View File

@@ -0,0 +1,153 @@
// 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

View File

@@ -11,22 +11,24 @@
#pragma once
#include "gflags/gflags.h"
#ifdef MG_ENTERPRISE
// Short help flag.
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_string(experimental_enabled);
#include "slk/serialization.hpp"
namespace memgraph::flags {
namespace memgraph::dbms {
// Each bit is an enabled experiment
// old experiments can be reused once code cleanup has happened
enum class Experiments : uint8_t {
SYSTEM_REPLICATION = 1 << 0,
class DbmsHandler;
class CoordinatorHandlers {
public:
static void Register(DbmsHandler &dbms_handler);
private:
static void PromoteReplicaToMainHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder);
static void SetMainToReplicaHandler(DbmsHandler &dbms_handler, slk::Reader *req_reader, slk::Builder *res_builder);
};
bool AreExperimentsEnabled(Experiments experiments);
} // namespace memgraph::dbms
void InitializeExperimental();
} // namespace memgraph::flags
#endif

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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
@@ -11,7 +11,10 @@
#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>;

View File

@@ -11,76 +11,29 @@
#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 "flags/experimental.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
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.");
}
constexpr std::string_view kDBPrefix = "database:"; // Key prefix for database durability
constexpr std::string_view kLastCommitedSystemTsKey = "last_commited_system_ts"; // Key for timestamp durability
} // namespace
#ifdef MG_ENTERPRISE
struct Durability {
enum class DurabilityVersion : uint8_t {
V0 = 0,
@@ -159,9 +112,11 @@ struct Durability {
}
};
DbmsHandler::DbmsHandler(storage::Config config, replication::ReplicationState &repl_state, auth::SynchedAuth &auth,
bool recovery_on_startup)
: default_config_{std::move(config)}, auth_{auth}, repl_state_{repl_state} {
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_)} {
// TODO: Decouple storage config from dbms config
// TODO: Save individual db configs inside the kvstore and restore from there
@@ -195,13 +150,19 @@ DbmsHandler::DbmsHandler(storage::Config config, replication::ReplicationState &
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, nullptr, rel_dir);
auto new_db = New_(name, uuid, 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) {
@@ -211,6 +172,8 @@ DbmsHandler::DbmsHandler(storage::Config config, replication::ReplicationState &
locked_auth->DeleteDatabase(name);
durability_->Delete(key);
}
// Delete the last timestamp
durability_->Delete(kLastCommitedSystemTsKey);
}
/*
@@ -235,31 +198,45 @@ DbmsHandler::DbmsHandler(storage::Config config, replication::ReplicationState &
*/
// 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!");
}
}
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) {
DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name) {
std::lock_guard<LockT> wr(lock_);
if (db_name == kDefaultDB) {
// MSG cannot delete the default db
@@ -296,10 +273,9 @@ DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name, syste
// Success
// Save delta
if (transaction) {
transaction->AddAction<DropDatabase>(uuid);
if (system_transaction_) {
system_transaction_->delta.emplace(SystemTransaction::Delta::drop_database, uuid);
}
return {};
}
@@ -320,50 +296,18 @@ DbmsHandler::DeleteResult DbmsHandler::Delete(utils::UUID uuid) {
return Delete_(db_name);
}
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) {
DbmsHandler::NewResultT DbmsHandler::New_(storage::Config storage_config) {
auto new_db = db_handler_.New(storage_config, repl_state_);
if (new_db.HasValue()) { // Success
// Save delta
UpdateDurability(storage_config);
if (txn) {
txn->AddAction<CreateDatabase>(storage_config.salient, new_db.GetValue());
if (system_transaction_) {
system_transaction_->delta.emplace(SystemTransaction::Delta::create_database, storage_config.salient);
}
UpdateDurability(storage_config);
return new_db.GetValue();
}
return new_db;
return new_db.GetError();
}
DbmsHandler::DeleteResult DbmsHandler::Delete_(std::string_view db_name) {
@@ -417,19 +361,89 @@ void DbmsHandler::UpdateDurability(const storage::Config &config, std::optional<
durability_->Put(key, val);
}
#endif
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;
void DbmsHandler::RecoverStorageReplication(DatabaseAccess db_acc, replication::RoleMainData &role_main_data) {
using enum memgraph::flags::Experiments;
auto const is_enterprise = license::global_license_checker.IsEnterpriseValidFast();
auto experimental_system_replication = flags::AreExperimentsEnabled(SYSTEM_REPLICATION);
if ((is_enterprise && experimental_system_replication) || 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!");
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
} // namespace memgraph::dbms

View File

@@ -25,25 +25,24 @@
#include "constants.hpp"
#include "dbms/database.hpp"
#include "dbms/inmemory/replication_handlers.hpp"
#include "dbms/rpc.hpp"
#include "dbms/replication_handler.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"
@@ -52,6 +51,11 @@
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
@@ -107,7 +111,8 @@ 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, replication::ReplicationState &repl_state, auth::SynchedAuth &auth,
DbmsHandler(storage::Config config,
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
bool recovery_on_startup); // TODO If more arguments are added use a config struct
#else
/**
@@ -115,13 +120,15 @@ class DbmsHandler {
*
* @param configs storage configuration
*/
DbmsHandler(storage::Config config, replication::ReplicationState &repl_state)
: repl_state_{repl_state},
DbmsHandler(storage::Config config)
: repl_state_{ReplicationStateRootPath(config)},
db_gatekeeper_{[&] {
config.salient.name = kDefaultDB;
return std::move(config);
}(),
repl_state_} {}
repl_state_} {
RecoverReplication(Get());
}
#endif
#ifdef MG_ENTERPRISE
@@ -131,10 +138,10 @@ class DbmsHandler {
* @param name name of the database
* @return NewResultT context on success, error on failure
*/
NewResultT New(const std::string &name, system::Transaction *txn = nullptr) {
NewResultT New(const std::string &name) {
std::lock_guard<LockT> wr(lock_);
const auto uuid = utils::UUID{};
return New_(name, uuid, txn);
return New_(name, uuid);
}
/**
@@ -227,7 +234,7 @@ class DbmsHandler {
* @param db_name database name
* @return DeleteResult error on failure
*/
DeleteResult TryDelete(std::string_view db_name, system::Transaction *transaction = nullptr);
DeleteResult TryDelete(std::string_view db_name);
/**
* @brief Delete or defer deletion of database.
@@ -267,29 +274,16 @@ 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
/**
* @brief Return all active databases.
*
* @return std::vector<std::string>
*/
auto Count() const -> std::size_t {
#ifdef MG_ENTERPRISE
std::shared_lock<LockT> rd(lock_);
return db_handler_.size();
#else
return 1;
#endif
}
/**
* @brief Return the statistics all databases.
*
* @return Statistics
*/
Statistics Stats(memgraph::replication_coordination_glue::ReplicationRole replication_role) {
Statistics Stats() {
auto const replication_role = repl_state_.GetRole();
Statistics stats{};
// TODO: Handle overflow?
#ifdef MG_ENTERPRISE
@@ -325,7 +319,8 @@ class DbmsHandler {
*
* @return std::vector<DatabaseInfo>
*/
std::vector<DatabaseInfo> Info(memgraph::replication_coordination_glue::ReplicationRole replication_role) {
std::vector<DatabaseInfo> Info() {
auto const replication_role = repl_state_.GetRole();
std::vector<DatabaseInfo> res;
#ifdef MG_ENTERPRISE
std::shared_lock<LockT> rd(lock_);
@@ -412,18 +407,99 @@ class DbmsHandler {
}
}
static void RecoverStorageReplication(DatabaseAccess db_acc, replication::RoleMainData &role_main_data);
auto default_config() const -> storage::Config const & {
#ifdef MG_ENTERPRISE
return default_config_;
#else
const auto acc = db_gatekeeper_.access();
MG_ASSERT(acc, "Failed to get default database!");
return acc->get()->config();
#endif
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); }
#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; });
}
#endif
private:
#ifdef MG_ENTERPRISE
/**
@@ -448,8 +524,7 @@ class DbmsHandler {
* @param uuid undelying RocksDB directory
* @return NewResultT context on success, error on failure
*/
NewResultT New_(std::string_view name, utils::UUID uuid, system::Transaction *txn = nullptr,
std::optional<std::filesystem::path> rel_dir = {}) {
NewResultT New_(std::string_view name, utils::UUID uuid, std::optional<std::filesystem::path> rel_dir = {}) {
auto config_copy = default_config_;
config_copy.salient.name = name;
config_copy.salient.uuid = uuid;
@@ -460,7 +535,7 @@ class DbmsHandler {
storage::UpdatePaths(config_copy,
default_config_.durability.storage_directory / kMultiTenantDir / std::string{uuid});
}
return New_(std::move(config_copy), txn);
return New_(std::move(config_copy));
}
/**
@@ -469,11 +544,11 @@ class DbmsHandler {
* @param config configuration to be used
* @return NewResultT context on success, error on failure
*/
NewResultT New_(const storage::SalientConfig &config, system::Transaction *txn = nullptr) {
NewResultT New_(const storage::SalientConfig &config) {
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), txn);
return New_(std::move(config_copy));
}
/**
@@ -482,7 +557,7 @@ class DbmsHandler {
* @param storage_config storage configuration
* @return NewResultT context on success, error on failure
*/
DbmsHandler::NewResultT New_(storage::Config storage_config, system::Transaction *txn = nullptr);
NewResultT New_(storage::Config storage_config);
// TODO: new overload of Delete_ with DatabaseAccess
DeleteResult Delete_(std::string_view db_name);
@@ -497,8 +572,7 @@ class DbmsHandler {
Get(kDefaultDB);
} catch (const UnknownDatabaseException &) {
// No default DB restored, create it
MG_ASSERT(New_(kDefaultDB, {/* random UUID */}, nullptr, ".").HasValue(),
"Failed while creating the default database");
MG_ASSERT(New_(kDefaultDB, {/* random UUID */}, ".").HasValue(), "Failed while creating the default database");
}
// For back-compatibility...
@@ -585,21 +659,35 @@ 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
// 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
std::unique_ptr<kvstore::KVStore> durability_; //!< list of active dbs (pointer so we can postpone its creation)
coordination::CoordinatorState coordinator_state_; //!< Replication coordinator
#endif
// TODO: Make an api
public:
utils::ResourceLock system_lock_{}; //!> Ensure exclusive access for system queries
private:
// 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
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
#ifndef MG_ENTERPRISE
mutable utils::Gatekeeper<Database> db_gatekeeper_; //!< Single databases gatekeeper
#endif

View File

@@ -144,8 +144,6 @@ class Handler {
auto cbegin() const { return items_.cbegin(); }
auto cend() const { return items_.cend(); }
auto size() const { return items_.size(); }
struct string_hash {
using is_transparent = void;
[[nodiscard]] size_t operator()(const char *s) const { return std::hash<std::string_view>{}(s); }

View File

@@ -76,110 +76,58 @@ 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::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);
});
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>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
[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);
InMemoryReplicationHandlers::AppendDeltasHandler(dbms_handler, 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::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) {
void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handler, 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) {
storage::replication::HeartbeatRes res{false, 0, ""};
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::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
if (!db_acc.has_value()) {
spdlog::warn("No database accessor");
storage::replication::HeartbeatRes res{false, 0, ""};
slk::Save(res, res_builder);
return;
}
auto const *storage = db_acc->get()->storage();
storage::replication::HeartbeatRes res{true, storage->repl_storage_state_.last_commit_timestamp_.load(),
std::string{storage->repl_storage_state_.epoch_.id()}};
slk::Save(res, 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) {
void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_handler, 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};
@@ -239,9 +187,8 @@ 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,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handler, 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);
@@ -250,12 +197,6 @@ 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);
@@ -329,9 +270,8 @@ void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handle
spdlog::debug("Replication recovery from snapshot finished!");
}
void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handler, 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);
@@ -340,12 +280,6 @@ 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);
@@ -364,9 +298,8 @@ 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,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_handler, 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);
@@ -376,13 +309,6 @@ 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());
@@ -444,9 +370,8 @@ void InMemoryReplicationHandlers::LoadWal(storage::InMemoryStorage *storage, sto
}
}
void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handler, 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);
@@ -456,13 +381,6 @@ 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()};

View File

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

View File

@@ -11,6 +11,10 @@
#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"

View File

@@ -0,0 +1,43 @@
// 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

View File

@@ -8,4 +8,14 @@
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "system/include/system/action.hpp"
#pragma once
#include "dbms/dbms_handler.hpp"
#include "replication/replication_client.hpp"
namespace memgraph::dbms {
void StartReplicaClient(DbmsHandler &dbms_handler, replication::ReplicationClient &client);
} // namespace memgraph::dbms

View File

@@ -0,0 +1,400 @@
// 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

View File

@@ -0,0 +1,82 @@
// 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

View File

@@ -1,205 +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_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

View File

@@ -1,42 +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 "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

View File

@@ -1,122 +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/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.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.config, builder);
}
void Load(memgraph::storage::replication::CreateDatabaseReq *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->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.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.uuid, builder);
}
void Load(memgraph::storage::replication::DropDatabaseReq *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->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

View File

@@ -1,122 +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 <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

View File

@@ -11,10 +11,7 @@
#pragma once
#include <list>
#include <memory>
#include <optional>
#include "auth/models.hpp"
#include "storage/v2/config.hpp"
namespace memgraph::dbms {
@@ -23,70 +20,17 @@ 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;
@@ -98,14 +42,8 @@ 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
}
}
@@ -113,20 +51,13 @@ 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) {}
std::list<Delta> deltas{};
// Currently system transitions support a single delta
std::optional<Delta> delta{};
uint64_t system_timestamp;
};

View File

@@ -0,0 +1,133 @@
// 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

View File

@@ -1,17 +1,12 @@
add_library(mg-flags STATIC
audit.cpp
bolt.cpp
general.cpp
isolation_level.cpp
log_level.cpp
memory_limit.cpp
run_time_configurable.cpp
storage_mode.cpp
query.cpp
replication.cpp
experimental.cpp
experimental.hpp)
target_include_directories(mg-flags PUBLIC include)
target_link_libraries(mg-flags
PUBLIC spdlog::spdlog mg-settings mg-utils
PRIVATE lib::rangev3)
add_library(mg-flags STATIC audit.cpp
bolt.cpp
general.cpp
isolation_level.cpp
log_level.cpp
memory_limit.cpp
run_time_configurable.cpp
storage_mode.cpp
query.cpp
replication.cpp)
target_include_directories(mg-flags PUBLIC ${CMAKE_SOURCE_DIR}/include)
target_link_libraries(mg-flags PUBLIC spdlog::spdlog mg-settings mg-utils)

View File

@@ -12,7 +12,6 @@
#include "flags/audit.hpp"
#include "flags/bolt.hpp"
#include "flags/experimental.hpp"
#include "flags/general.hpp"
#include "flags/isolation_level.hpp"
#include "flags/log_level.hpp"

View File

@@ -1,67 +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 "flags/experimental.hpp"
#include "range/v3/all.hpp"
#include "utils/string.hpp"
#include <map>
#include <string_view>
// Bolt server flags.
// NOLINTNEXTLINE (cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_string(experimental_enabled, "",
"Experimental features to be used, comma seperated. Options [system-replication]");
using namespace std::string_view_literals;
namespace memgraph::flags {
auto const mapping = std::map{std::pair{"system-replication"sv, Experiments::SYSTEM_REPLICATION}};
auto ExperimentsInstance() -> Experiments & {
static auto instance = Experiments{};
return instance;
}
bool AreExperimentsEnabled(Experiments experiments) {
using t = std::underlying_type_t<Experiments>;
auto actual = static_cast<t>(ExperimentsInstance());
auto check = static_cast<t>(experiments);
return (actual & check) == check;
}
void InitializeExperimental() {
namespace rv = ranges::views;
auto const connonicalize_string = [](auto &&rng) {
auto const is_space = [](auto c) { return c == ' '; };
auto const to_lower = [](unsigned char c) { return std::tolower(c); };
return rng | rv::drop_while(is_space) | rv::take_while(std::not_fn(is_space)) | rv::transform(to_lower) |
ranges::to<std::string>;
};
auto const mapping_end = mapping.cend();
using underlying_type = std::underlying_type_t<Experiments>;
auto to_set = underlying_type{};
for (auto &&experiment : FLAGS_experimental_enabled | rv::split(',') | rv::transform(connonicalize_string)) {
if (auto it = mapping.find(experiment); it != mapping_end) {
to_set |= static_cast<underlying_type>(it->second);
}
}
ExperimentsInstance() = static_cast<Experiments>(to_set);
}
} // namespace memgraph::flags

View File

@@ -13,17 +13,9 @@
#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.");
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_uint32(instance_down_timeout_sec, 5, "Time duration after which an instance is considered down.");
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_uint32(instance_health_check_frequency_sec, 1, "The time duration between two health checks/pings.");
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_uint32(instance_get_uuid_frequency_sec, 10, "The time duration between two instance uuid checks.");
#endif
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)

View File

@@ -15,17 +15,9 @@
#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);
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_uint32(instance_down_timeout_sec);
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_uint32(instance_health_check_frequency_sec);
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_uint32(instance_get_uuid_frequency_sec);
#endif
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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::auth::SynchedAuth *auth;
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth;
#if MG_ENTERPRISE
memgraph::audit::Log *audit_log;
#endif

View File

@@ -319,7 +319,8 @@ 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::auth::SynchedAuth *auth
memgraph::communication::v2::OutputStream *output_stream,
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth
#ifdef MG_ENTERPRISE
,
memgraph::audit::Log *audit_log

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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
@@ -25,7 +25,8 @@ 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::auth::SynchedAuth *auth
memgraph::communication::v2::OutputStream *output_stream,
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth
#ifdef MG_ENTERPRISE
,
memgraph::audit::Log *audit_log
@@ -87,7 +88,7 @@ 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::auth::SynchedAuth *auth_;
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth_;
memgraph::communication::v2::ServerEndpoint endpoint_;
std::optional<std::string> implicit_db_;
};

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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
@@ -66,7 +66,9 @@ bool IsUserAuthorizedEdgeType(const memgraph::auth::User &user, const memgraph::
#endif
namespace memgraph::glue {
AuthChecker::AuthChecker(memgraph::auth::SynchedAuth *auth) : auth_(auth) {}
AuthChecker::AuthChecker(
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth)
: auth_(auth) {}
bool AuthChecker::IsUserAuthorized(const std::optional<std::string> &username,
const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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
@@ -22,7 +22,8 @@ namespace memgraph::glue {
class AuthChecker : public query::AuthChecker {
public:
explicit AuthChecker(memgraph::auth::SynchedAuth *auth);
explicit AuthChecker(
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth);
bool IsUserAuthorized(const std::optional<std::string> &username,
const std::vector<query::AuthQuery::Privilege> &privileges,
@@ -40,7 +41,7 @@ class AuthChecker : public query::AuthChecker {
const std::string &db_name = "");
private:
memgraph::auth::SynchedAuth *auth_;
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth_;
mutable memgraph::utils::Synchronized<auth::User, memgraph::utils::SpinLock> user_; // cached user
};
#ifdef MG_ENTERPRISE

View File

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

View File

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

View File

@@ -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"

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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.

View File

@@ -37,7 +37,6 @@ 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!");

View File

@@ -11,12 +11,9 @@
#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"
@@ -27,19 +24,14 @@
#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"
@@ -47,6 +39,10 @@
#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";
@@ -134,7 +130,6 @@ int main(int argc, char **argv) {
}
memgraph::flags::InitializeLogger();
memgraph::flags::InitializeExperimental();
// Unhandled exception handler init.
std::set_terminate(&memgraph::utils::TerminateHandler);
@@ -356,85 +351,49 @@ int main(int argc, char **argv) {
memgraph::query::InterpreterConfig interp_config{
.query = {.allow_load_csv = FLAGS_allow_load_csv},
.replication_replica_check_frequency = std::chrono::seconds(FLAGS_replication_replica_check_frequency_sec),
#ifdef MG_ENTERPRISE
.instance_down_timeout_sec = std::chrono::seconds(FLAGS_instance_down_timeout_sec),
.instance_health_check_frequency_sec = std::chrono::seconds(FLAGS_instance_health_check_frequency_sec),
.instance_get_uuid_frequency_sec = std::chrono::seconds(FLAGS_instance_get_uuid_frequency_sec),
#endif
.default_kafka_bootstrap_servers = FLAGS_kafka_bootstrap_servers,
.default_pulsar_service_url = FLAGS_pulsar_service_url,
.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::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);
}
}
};
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);
}
}
};
memgraph::auth::Auth::Config auth_config{FLAGS_auth_user_or_role_name_regex, FLAGS_auth_password_strength_regex,
FLAGS_auth_password_permit_null};
memgraph::auth::SynchedAuth auth_{data_directory / "auth", auth_config};
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> 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);
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, repl_state
memgraph::dbms::DbmsHandler dbms_handler(db_config
#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, &repl_state, system,
#ifdef MG_ENTERPRISE
&coordinator_state,
#endif
auth_handler.get(), auth_checker.get(),
&replication_handler);
memgraph::query::InterpreterContext interpreter_context_(
interp_config, &dbms_handler, &dbms_handler.ReplicationState(), auth_handler.get(), auth_checker.get());
MG_ASSERT(db_acc, "Failed to access the main database");
memgraph::query::procedure::gModuleRegistry.SetModulesDirectory(memgraph::flags::ParseQueryModulesDirectory(),
@@ -501,9 +460,9 @@ 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_, repl_state);
telemetry->AddStorageCollector(dbms_handler, auth_);
#ifdef MG_ENTERPRISE
telemetry->AddDatabaseCollector(dbms_handler, repl_state);
telemetry->AddDatabaseCollector(dbms_handler);
#else
telemetry->AddDatabaseCollector();
#endif

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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
@@ -60,13 +60,10 @@ void *my_alloc(extent_hooks_t *extent_hooks, void *new_addr, size_t size, size_t
unsigned arena_ind) {
// This needs to be before, to throw exception in case of too big alloc
if (*commit) [[likely]] {
memgraph::utils::total_memory_tracker.Alloc(static_cast<int64_t>(size));
if (GetQueriesMemoryControl().IsThreadTracked()) [[unlikely]] {
bool ok = GetQueriesMemoryControl().TrackAllocOnCurrentThread(size);
if (!ok) return nullptr;
GetQueriesMemoryControl().TrackAllocOnCurrentThread(size);
}
// This needs to be here so it doesn't get incremented in case the first TrackAlloc throws an exception
bool ok = memgraph::utils::total_memory_tracker.Alloc(static_cast<int64_t>(size));
if (!ok) return nullptr;
}
auto *ptr = old_hooks->alloc(extent_hooks, new_addr, size, alignment, zero, commit, arena_ind);
@@ -120,15 +117,11 @@ static bool my_commit(extent_hooks_t *extent_hooks, void *addr, size_t size, siz
return err;
}
[[maybe_unused]] auto blocker = memgraph::utils::MemoryTracker::OutOfMemoryExceptionBlocker{};
memgraph::utils::total_memory_tracker.Alloc(static_cast<int64_t>(length));
if (GetQueriesMemoryControl().IsThreadTracked()) [[unlikely]] {
bool ok = GetQueriesMemoryControl().TrackAllocOnCurrentThread(length);
DMG_ASSERT(ok);
GetQueriesMemoryControl().TrackAllocOnCurrentThread(size);
}
auto ok = memgraph::utils::total_memory_tracker.Alloc(static_cast<int64_t>(length));
DMG_ASSERT(ok);
return false;
}

View File

@@ -1,4 +1,4 @@
// Copyright 2024 Memgraph Ltd.
// Copyright 2023 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
@@ -28,12 +28,6 @@ void *newImpl(const std::size_t size) {
return ptr;
}
[[maybe_unused]] auto blocker = memgraph::utils::MemoryTracker::OutOfMemoryExceptionBlocker{};
auto maybe_msg = memgraph::utils::MemoryErrorStatus().msg();
if (maybe_msg) {
throw memgraph::utils::OutOfMemoryException{std::move(*maybe_msg)};
}
throw std::bad_alloc{};
}
@@ -43,21 +37,11 @@ void *newImpl(const std::size_t size, const std::align_val_t align) {
return ptr;
}
[[maybe_unused]] auto blocker = memgraph::utils::MemoryTracker::OutOfMemoryExceptionBlocker{};
auto maybe_msg = memgraph::utils::MemoryErrorStatus().msg();
if (maybe_msg) {
throw memgraph::utils::OutOfMemoryException{std::move(*maybe_msg)};
}
throw std::bad_alloc{};
}
void *newNoExcept(const std::size_t size) noexcept {
[[maybe_unused]] auto blocker = memgraph::utils::MemoryTracker::OutOfMemoryExceptionBlocker{};
return malloc(size);
}
void *newNoExcept(const std::size_t size) noexcept { return malloc(size); }
void *newNoExcept(const std::size_t size, const std::align_val_t align) noexcept {
[[maybe_unused]] auto blocker = memgraph::utils::MemoryTracker::OutOfMemoryExceptionBlocker{};
return aligned_alloc(size, static_cast<std::size_t>(align));
}

View File

@@ -54,14 +54,14 @@ void QueriesMemoryControl::EraseThreadToTransactionId(const std::thread::id &thr
}
}
bool QueriesMemoryControl::TrackAllocOnCurrentThread(size_t size) {
void QueriesMemoryControl::TrackAllocOnCurrentThread(size_t size) {
auto thread_id_to_transaction_id_accessor = thread_id_to_transaction_id.access();
// we might be just constructing mapping between thread id and transaction id
// so we miss this allocation
auto thread_id_to_transaction_id_elem = thread_id_to_transaction_id_accessor.find(std::this_thread::get_id());
if (thread_id_to_transaction_id_elem == thread_id_to_transaction_id_accessor.end()) {
return true;
return;
}
auto transaction_id_to_tracker_accessor = transaction_id_to_tracker.access();
@@ -71,10 +71,10 @@ bool QueriesMemoryControl::TrackAllocOnCurrentThread(size_t size) {
// It can happen that some allocation happens between mapping thread to
// transaction id, so we miss this allocation
if (transaction_id_to_tracker == transaction_id_to_tracker_accessor.end()) [[unlikely]] {
return true;
return;
}
auto &query_tracker = transaction_id_to_tracker->tracker;
return query_tracker.TrackAlloc(size);
query_tracker.TrackAlloc(size);
}
void QueriesMemoryControl::TrackFreeOnCurrentThread(size_t size) {
@@ -110,7 +110,7 @@ void QueriesMemoryControl::CreateTransactionIdTracker(uint64_t transaction_id, s
bool QueriesMemoryControl::EraseTransactionIdTracker(uint64_t transaction_id) {
auto transaction_id_to_tracker_accessor = transaction_id_to_tracker.access();
auto removed = transaction_id_to_tracker_accessor.remove(transaction_id);
auto removed = transaction_id_to_tracker.access().remove(transaction_id);
return removed;
}

View File

@@ -62,7 +62,7 @@ class QueriesMemoryControl {
// Find tracker for current thread if exists, track
// query allocation and procedure allocation if
// necessary
bool TrackAllocOnCurrentThread(size_t size);
void TrackAllocOnCurrentThread(size_t size);
// Find tracker for current thread if exists, track
// query allocation and procedure allocation if

View File

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

Some files were not shown because too many files have changed in this diff Show More