Compare commits

...

27 Commits

Author SHA1 Message Date
Andi Skrgat
8f9e044fcd Add unreachable replica state 2024-02-09 07:53:21 +01:00
Andi
efd3257479 Merge branch 'master' into replication-status 2024-02-08 16:45:29 +01:00
Marko Barišić
dcdbd0a19a Fix primary urls (#1700) 2024-02-08 14:19:30 +01:00
Andi Skrgat
189895370b Adapt tests to current state 2024-02-08 11:05:40 +01:00
Andi Skrgat
4f873a6b4d Test for distributed AF 2024-02-08 10:36:31 +01:00
Andi Skrgat
ec6d35ff67 Added InMemoryLogStore 2024-02-08 10:35:55 +01:00
Andi Skrgat
e125c5cd98 Tests for creating cluster 2024-02-08 10:35:54 +01:00
Andi Skrgat
1ecf6ddab2 Request leadership on registering instance 2024-02-08 10:34:38 +01:00
Andi Skrgat
17ad671773 Callbacks for leadership change 2024-02-08 10:34:38 +01:00
Andi Skrgat
7386b786a9 Remove CoordinatorData 2024-02-08 10:34:37 +01:00
Andi Skrgat
6e758d3b5a Only leader performing callbacks 2024-02-08 10:24:21 +01:00
Andi
cf80687d1d HA: Organize Raft coordinator group (#1687) 2024-02-08 09:11:33 +00:00
Aidar Samerkhanov
2fa8e00124 Fix accumulated path evaluation in builtin algorithms. (#1642)
Fix accumulated path evaluation in DFS, BFS, WeghtedShortestPath and AllShortestPath algorithm.
2024-02-08 10:48:54 +04:00
Antonio Filipovic
c15b62a88d HA: Disable replication from old main (#1674) 2024-02-07 11:20:47 +01:00
Gareth Andrew Lloyd
4ef6a1f9c3 Improve memory handling of Deltas (#1688)
- Reduce delta from 104B to 80B
- Hold and pass them around as in a deque
- Detect and deleted deltas within commit if safe to do so
2024-02-06 18:07:38 +01:00
andrejtonev
7ead00f23e Adding authentication data replication (#1666)
* Add AUTH system tx deltas
* Add auth data RPC and handlers
* Support multiple system deltas in a single transaction
* Added e2e test
* Bugfix: KVStore segfault after move

---------

Co-authored-by: Gareth Lloyd <gareth.lloyd@memgraph.io>
2024-02-05 10:37:00 +00:00
Marko Budiselić
c46dad18fe Add RocksDB ADR (#1659) 2024-02-03 19:21:13 +01:00
Andi
cb7b88ad92 HA: Support restart of instances (#1672) 2024-02-01 11:55:48 +01:00
Marko Barišić
b443934b68 Update release CI job timeouts (#1683)
* Set timeout for jobs in release CI to 60 minutes

* Set timout for stress test large to 12 hours
2024-01-31 19:19:05 +01:00
Andi
6ab4235cc9 Add NuRaft library (#1678) 2024-01-31 13:13:59 +01:00
Marko Barišić
a9ef28c68e Upgrade deprecated github actions (#1673)
* upgrade actions/checkout from v3 to v4

* upgrade actions/upload-artifact from v3 to v4

* upgrade actions/download-artifact from v2 and v3 to v4

* Fix duplicate artifact names in diff.yaml

* Fix duplicate artifact names in release_debian10.yaml and release_ubuntu2004.yaml
2024-01-30 22:52:56 +01:00
Marko Barišić
79361e9205 Disable e2e tests in all workflows (#1677)
* Disable e2e tests in diff.yaml

* Disable e2e tests in release builds
2024-01-30 19:37:00 +01:00
Gareth Andrew Lloyd
97b1e67d80 Fix auth durability (#1644)
* Change auth durability to store hash algorithm
* Add Salt to SHA256
2024-01-30 18:17:05 +00:00
Andi
78a88737f8 HA: Add automatic failover (#1646)
Co-authored-by: antoniofilipovic <filipovicantonio1998@gmail.com>
2024-01-29 15:34:00 +01:00
andrejtonev
ff44d68843 Simplify auth::Auth (#1663)
Moved various auth flags under a single config
Moved all regex logic under auth::Auth
2024-01-29 12:52:32 +00:00
Marko Barišić
f1484240a0 Make release CI predictable (#1658)
* Move stress test large to a separate workflow
* Combine Debian10 and Ubuntu20.04 stress tests into one workflow
* Remove BigMemory tag from release_debian10
* Move e2e tests to a separate job
* Move debug interation tests to a separate job
* Move release durability and stress tests to a separate job
* Move release benchamarks to separate job
* Add 90 min timeout restriction to all jobs
* Move env variables to workflow level
* Move BUILD_TYPE env var to workflow level
---------

Co-authored-by: Aidar Samerkhanov <aidar.samerkhanov@memgraph.io>
2024-01-26 17:01:25 +01:00
Marko Barišić
1c95c3dc59 Add step to refresh jepsen cluster before test (#1667)
refresh jepsen during diff workflow
2024-01-26 10:50:03 +00:00
268 changed files with 10731 additions and 4604 deletions

View File

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

View File

@@ -16,7 +16,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)

View File

@@ -27,7 +27,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -65,7 +65,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -96,7 +96,7 @@ jobs:
- name: Python code analysis
run: |
CHANGED_FILES=$(git diff -U0 ${{ env.BASE_BRANCH }}... --name-only)
CHANGED_FILES=$(git diff -U0 ${{ env.BASE_BRANCH }}... --name-only --diff-filter=d)
for file in ${CHANGED_FILES}; do
echo ${file}
if [[ ${file} == *.py ]]; then
@@ -137,9 +137,9 @@ jobs:
tar -czf code_coverage.tar.gz coverage.json html report.json summary.rmu
- name: Save code coverage
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Code coverage"
name: "Code coverage(Code analysis)"
path: tools/github/generated/code_coverage.tar.gz
- name: Run clang-tidy
@@ -162,7 +162,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -208,9 +208,9 @@ jobs:
./cppcheck_and_clang_format diff
- name: Save cppcheck and clang-format errors
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Code coverage"
name: "Code coverage(Debug build)"
path: tools/github/cppcheck_and_clang_format.txt
release_build:
@@ -223,7 +223,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -251,7 +251,7 @@ jobs:
./continuous_integration
- name: Save quality assurance status
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "GQL Behave Status"
path: |
@@ -268,6 +268,7 @@ jobs:
ctest -R memgraph__unit --output-on-failure -j$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
@@ -283,7 +285,7 @@ jobs:
./run.sh
- name: Ensure Kafka and Pulsar are down
if: always()
if: false
run: |
cd tests/e2e/streams/kafka
docker-compose down
@@ -323,16 +325,16 @@ jobs:
cpack -G DEB --config ../CPackConfig.cmake
- name: Save enterprise DEB package
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Enterprise DEB package"
path: build/output/memgraph*.deb
- name: Save test data
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
if: always()
with:
name: "Test data"
name: "Test data(Release build)"
path: |
# multiple paths could be defined
build/logs
@@ -347,7 +349,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -366,6 +368,7 @@ jobs:
cd build
ctest -R memgraph__unit --output-on-failure -j$THREADS
- name: Run e2e tests
if: false
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
@@ -375,10 +378,10 @@ jobs:
./run.sh "Client initiated failover"
./run.sh "Uninitialized cluster"
- name: Save test data
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
if: always()
with:
name: "Test data"
name: "Test data(High availability build)"
path: |
# multiple paths could be defined
build/logs
@@ -393,7 +396,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -423,6 +426,7 @@ jobs:
ctest -R memgraph__unit --output-on-failure -j$THREADS
- name: Run e2e tests
if: false
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
@@ -439,10 +443,10 @@ jobs:
./run.sh "Constraints"
- name: Save test data
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
if: always()
with:
name: "Test data"
name: "Test data(MultiTenancy replication build)"
path: |
# multiple paths could be defined
build/logs
@@ -458,7 +462,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -475,13 +479,18 @@ jobs:
cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo ..
make -j$THREADS memgraph
- name: Refresh Jepsen Cluster
run: |
cd tests/jepsen
./run.sh cluster-refresh
- name: Run Jepsen tests
run: |
cd tests/jepsen
./run.sh test-all-individually --binary ../../build/memgraph --ignore-run-stdout-logs --ignore-run-stderr-logs
- name: Save Jepsen report
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
if: ${{ always() }}
with:
name: "Jepsen Report"
@@ -497,7 +506,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)

View File

@@ -14,7 +14,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)

View File

@@ -42,14 +42,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package amzn-2 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: amzn-2
path: build/output/amzn-2/memgraph*.rpm
@@ -60,14 +60,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package centos-7 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: centos-7
path: build/output/centos-7/memgraph*.rpm
@@ -78,14 +78,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package centos-9 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: centos-9
path: build/output/centos-9/memgraph*.rpm
@@ -96,14 +96,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package debian-10 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: debian-10
path: build/output/debian-10/memgraph*.deb
@@ -114,14 +114,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package debian-11 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: debian-11
path: build/output/debian-11/memgraph*.deb
@@ -132,14 +132,14 @@ jobs:
timeout-minutes: 120
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package debian-11-arm ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: debian-11-aarch64
path: build/output/debian-11-arm/memgraph*.deb
@@ -150,14 +150,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package debian-11 ${{ github.event.inputs.build_type }} --for-platform
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: debian-11-platform
path: build/output/debian-11/memgraph*.deb
@@ -168,7 +168,7 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
@@ -177,7 +177,7 @@ jobs:
./run.sh package debian-11 ${{ github.event.inputs.build_type }} --for-docker
./run.sh docker
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: docker
path: build/output/docker/memgraph*.tar.gz
@@ -188,14 +188,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package fedora-36 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: fedora-36
path: build/output/fedora-36/memgraph*.rpm
@@ -206,14 +206,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package ubuntu-18.04 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: ubuntu-18.04
path: build/output/ubuntu-18.04/memgraph*.deb
@@ -224,14 +224,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package ubuntu-20.04 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: ubuntu-20.04
path: build/output/ubuntu-20.04/memgraph*.deb
@@ -242,14 +242,14 @@ jobs:
timeout-minutes: 60
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package ubuntu-22.04 ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: ubuntu-22.04
path: build/output/ubuntu-22.04/memgraph*.deb
@@ -260,14 +260,14 @@ jobs:
timeout-minutes: 120
steps:
- name: "Set up repository"
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
fetch-depth: 0 # Required because of release/get_version.py
- name: "Build package"
run: |
./release/package/run.sh package ubuntu-22.04-arm ${{ github.event.inputs.build_type }}
- name: "Upload package"
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: ubuntu-22.04-aarch64
path: build/output/ubuntu-22.04-arm/memgraph*.deb
@@ -279,7 +279,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Download artifacts
uses: actions/download-artifact@v3
uses: actions/download-artifact@v4
with:
# name: # if name input parameter is not provided, all artifacts are downloaded
# and put in directories named after each one.

View File

@@ -14,7 +14,7 @@ jobs:
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)

View File

@@ -14,19 +14,21 @@ on:
schedule:
- cron: "0 22 * * *"
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
BUILD_TYPE: ${{ github.event.inputs.build_type || 'Release' }}
jobs:
community_build:
name: "Community build"
runs-on: [self-hosted, Linux, X64, Debian10]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 960
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -40,10 +42,6 @@ jobs:
# Initialize dependencies.
./init
# Set default build_type to Release
INPUT_BUILD_TYPE=${{ github.event.inputs.build_type }}
BUILD_TYPE=${INPUT_BUILD_TYPE:-"Release"}
# Build community binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE -DMG_ENTERPRISE=OFF ..
@@ -65,10 +63,11 @@ jobs:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -110,22 +109,19 @@ jobs:
tar -czf code_coverage.tar.gz coverage.json html report.json summary.rmu
- name: Save code coverage
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Code coverage"
name: "Code coverage(Coverage build)"
path: tools/github/generated/code_coverage.tar.gz
debug_build:
name: "Debug build"
runs-on: [self-hosted, Linux, X64, Debian10]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -157,10 +153,6 @@ jobs:
run: |
./tests/drivers/run.sh
- name: Run integration tests
run: |
tests/integration/run.sh
- name: Run cppcheck and clang-format
run: |
# Activate toolchain.
@@ -171,23 +163,49 @@ jobs:
./cppcheck_and_clang_format diff
- name: Save cppcheck and clang-format errors
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Code coverage"
name: "Code coverage(Debug build)"
path: tools/github/cppcheck_and_clang_format.txt
release_build:
name: "Release build"
runs-on: [self-hosted, Linux, X64, Debian10, BigMemory]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 960
debug_integration_test:
name: "Debug integration tests"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Build debug binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Build debug binaries.
cd build
cmake ..
make -j$THREADS
- name: Run integration tests
run: |
tests/integration/run.sh
release_build:
name: "Release build"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
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)
@@ -201,10 +219,6 @@ jobs:
# Initialize dependencies.
./init
# Set default build_type to Release
INPUT_BUILD_TYPE=${{ github.event.inputs.build_type }}
BUILD_TYPE=${INPUT_BUILD_TYPE:-"Release"}
# Build release binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
@@ -226,11 +240,60 @@ jobs:
cpack -G DEB --config ../CPackConfig.cmake
- name: Save enterprise DEB package
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Enterprise DEB package"
path: build/output/memgraph*.deb
- name: Run GQL Behave tests
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
cd gql_behave
./continuous_integration
- name: Save quality assurance status
uses: actions/upload-artifact@v4
with:
name: "GQL Behave Status"
path: |
tests/gql_behave/gql_behave_status.csv
tests/gql_behave/gql_behave_status.html
- name: Run unit tests
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Run unit tests.
cd build
ctest -R memgraph__unit --output-on-failure
release_benchmark_tests:
name: "Release Benchmark Tests"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Build release binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Build release binaries
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Run micro benchmark tests
run: |
# Activate toolchain.
@@ -257,29 +320,31 @@ jobs:
--num-database-workers 9 --num-clients-workers 30 \
--no-strict
- name: Run GQL Behave tests
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
cd gql_behave
./continuous_integration
release_e2e_test:
if: false
name: "Release End-to-end Test"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
- name: Save quality assurance status
uses: actions/upload-artifact@v3
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
name: "GQL Behave Status"
path: |
tests/gql_behave/gql_behave_status.csv
tests/gql_behave/gql_behave_status.html
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Run unit tests
- name: Build release binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Run unit tests.
# Build release binaries
cd build
ctest -R memgraph__unit --output-on-failure
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Ensure Kafka and Pulsar are up
run: |
@@ -304,6 +369,32 @@ jobs:
cd ../pulsar
docker-compose down
release_durability_stress_tests:
name: "Release durability and stress tests"
runs-on: [self-hosted, Linux, X64, Debian10]
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Build release binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Build release binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Run stress test (plain)
run: |
cd tests/stress
@@ -314,11 +405,6 @@ jobs:
cd tests/stress
./continuous_integration --use-ssl
- name: Run stress test (large)
run: |
cd tests/stress
./continuous_integration --large-dataset
- name: Run durability test (plain)
run: |
cd tests/stress
@@ -334,15 +420,11 @@ jobs:
release_jepsen_test:
name: "Release Jepsen Test"
runs-on: [self-hosted, Linux, X64, Debian10, JepsenControl]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -355,22 +437,23 @@ jobs:
# Initialize dependencies.
./init
# Set default build_type to Release
INPUT_BUILD_TYPE=${{ github.event.inputs.build_type }}
BUILD_TYPE=${INPUT_BUILD_TYPE:-"Release"}
# Build only memgraph release binary.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS memgraph
- name: Refresh Jepsen Cluster
run: |
cd tests/jepsen
./run.sh cluster-refresh
- name: Run Jepsen tests
run: |
cd tests/jepsen
./run.sh test-all-individually --binary ../../build/memgraph --ignore-run-stdout-logs --ignore-run-stderr-logs
- name: Save Jepsen report
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
if: ${{ always() }}
with:
name: "Jepsen Report"

View File

@@ -19,7 +19,7 @@ jobs:
DOCKER_REPOSITORY_NAME: memgraph
steps:
- name: Checkout
uses: actions/checkout@v3
uses: actions/checkout@v4
- name: Set up QEMU
uses: docker/setup-qemu-action@v2

View File

@@ -20,7 +20,7 @@ jobs:
steps:
- name: Checkout
uses: actions/checkout@v3
uses: actions/checkout@v4
- name: Set up QEMU
uses: docker/setup-qemu-action@v2

View File

@@ -14,19 +14,21 @@ on:
schedule:
- cron: "0 22 * * *"
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
BUILD_TYPE: ${{ github.event.inputs.build_type || 'Release' }}
jobs:
community_build:
name: "Community build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 960
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -40,10 +42,6 @@ jobs:
# Initialize dependencies.
./init
# Set default build_type to Release
INPUT_BUILD_TYPE=${{ github.event.inputs.build_type }}
BUILD_TYPE=${INPUT_BUILD_TYPE:-"Release"}
# Build community binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE -DMG_ENTERPRISE=OFF ..
@@ -61,14 +59,11 @@ jobs:
coverage_build:
name: "Coverage build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -110,22 +105,19 @@ jobs:
tar -czf code_coverage.tar.gz coverage.json html report.json summary.rmu
- name: Save code coverage
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Code coverage"
name: "Code coverage(Coverage build)"
path: tools/github/generated/code_coverage.tar.gz
debug_build:
name: "Debug build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -157,10 +149,6 @@ jobs:
run: |
./tests/drivers/run.sh
- name: Run integration tests
run: |
tests/integration/run.sh
- name: Run cppcheck and clang-format
run: |
# Activate toolchain.
@@ -171,23 +159,49 @@ jobs:
./cppcheck_and_clang_format diff
- name: Save cppcheck and clang-format errors
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Code coverage"
name: "Code coverage(Debug build)"
path: tools/github/cppcheck_and_clang_format.txt
debug_integration_test:
name: "Debug integration tests"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Build debug binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Build debug binaries.
cd build
cmake ..
make -j$THREADS
- name: Run integration tests
run: |
tests/integration/run.sh
release_build:
name: "Release build"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
timeout-minutes: 960
timeout-minutes: 60
steps:
- name: Set up repository
uses: actions/checkout@v3
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
@@ -201,10 +215,6 @@ jobs:
# Initialize dependencies.
./init
# Set default build_type to Release
INPUT_BUILD_TYPE=${{ github.event.inputs.build_type }}
BUILD_TYPE=${INPUT_BUILD_TYPE:-"Release"}
# Build release binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
@@ -226,11 +236,60 @@ jobs:
cpack -G DEB --config ../CPackConfig.cmake
- name: Save enterprise DEB package
uses: actions/upload-artifact@v3
uses: actions/upload-artifact@v4
with:
name: "Enterprise DEB package"
path: build/output/memgraph*.deb
- name: Run GQL Behave tests
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
cd gql_behave
./continuous_integration
- name: Save quality assurance status
uses: actions/upload-artifact@v4
with:
name: "GQL Behave Status"
path: |
tests/gql_behave/gql_behave_status.csv
tests/gql_behave/gql_behave_status.html
- name: Run unit tests
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Run unit tests.
cd build
ctest -R memgraph__unit --output-on-failure
release_benchmark_tests:
name: "Release Benchmark Tests"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
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 release binaries
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Run micro benchmark tests
run: |
# Activate toolchain.
@@ -257,29 +316,31 @@ jobs:
--num-database-workers 9 --num-clients-workers 30 \
--no-strict
- name: Run GQL Behave tests
run: |
cd tests
./setup.sh /opt/toolchain-v4/activate
cd gql_behave
./continuous_integration
release_e2e_test:
if: false
name: "Release End-to-end Test"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
- name: Save quality assurance status
uses: actions/upload-artifact@v3
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
name: "GQL Behave Status"
path: |
tests/gql_behave/gql_behave_status.csv
tests/gql_behave/gql_behave_status.html
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Run unit tests
- name: Build release binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Run unit tests.
# Build release binaries
cd build
ctest -R memgraph__unit --output-on-failure
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Ensure Kafka and Pulsar are up
run: |
@@ -304,6 +365,32 @@ jobs:
cd ../pulsar
docker-compose down
release_durability_stress_tests:
name: "Release durability and stress tests"
runs-on: [self-hosted, Linux, X64, Ubuntu20.04]
timeout-minutes: 60
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 release binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Run stress test (plain)
run: |
cd tests/stress
@@ -314,11 +401,6 @@ jobs:
cd tests/stress
./continuous_integration --use-ssl
- name: Run stress test (large)
run: |
cd tests/stress
./continuous_integration --large-dataset
- name: Run durability test (plain)
run: |
cd tests/stress

View File

@@ -0,0 +1,62 @@
name: Stress test large
on:
workflow_dispatch:
inputs:
build_type:
type: choice
description: "Memgraph Build type. Default value is Release."
default: 'Release'
options:
- Release
- RelWithDebInfo
schedule:
- cron: "0 22 * * *"
env:
THREADS: 24
MEMGRAPH_ENTERPRISE_LICENSE: ${{ secrets.MEMGRAPH_ENTERPRISE_LICENSE }}
MEMGRAPH_ORGANIZATION_NAME: ${{ secrets.MEMGRAPH_ORGANIZATION_NAME }}
BUILD_TYPE: ${{ github.event.inputs.build_type || 'Release' }}
jobs:
stress_test_large:
name: "Stress test large"
timeout-minutes: 720
strategy:
matrix:
os: [Debian10, Ubuntu20.04]
extra: [BigMemory, Gen8]
exclude:
- os: Debian10
extra: Gen8
- os: Ubuntu20.04
extra: BigMemory
runs-on: [self-hosted, Linux, X64, "${{ matrix.os }}", "${{ matrix.extra }}"]
steps:
- name: Set up repository
uses: actions/checkout@v4
with:
# Number of commits to fetch. `0` indicates all history for all
# branches and tags. (default: 1)
fetch-depth: 0
- name: Build release binaries
run: |
# Activate toolchain.
source /opt/toolchain-v4/activate
# Initialize dependencies.
./init
# Build release binaries.
cd build
cmake -DCMAKE_BUILD_TYPE=$BUILD_TYPE ..
make -j$THREADS
- name: Run stress test (large)
run: |
cd tests/stress
./continuous_integration --large-dataset

View File

@@ -15,7 +15,7 @@ jobs:
runs-on: ubuntu-latest
steps:
- name: Download artifacts
uses: dawidd6/action-download-artifact@v2
uses: dawidd6/action-download-artifact@v4
with:
workflow: package_all.yaml
workflow_conclusion: success

38
ADRs/003_rocksdb.md Normal file
View File

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

View File

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

View File

@@ -294,3 +294,10 @@ 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

@@ -123,9 +123,10 @@ declare -A primary_urls=(
["pulsar"]="http://$local_cache_host/git/pulsar.git"
["librdtsc"]="http://$local_cache_host/git/librdtsc.git"
["ctre"]="http://$local_cache_host/file/hanickadot/compile-time-regular-expressions/v3.7.2/single-header/ctre.hpp"
["absl"]="https://$local_cache_host/git/abseil-cpp.git"
["jemalloc"]="https://$local_cache_host/git/jemalloc.git"
["range-v3"]="https://$local_cache_host/git/ericniebler/range-v3.git"
["absl"]="http://$local_cache_host/git/abseil-cpp.git"
["jemalloc"]="http://$local_cache_host/git/jemalloc.git"
["range-v3"]="http://$local_cache_host/git/range-v3.git"
["nuraft"]="http://$local_cache_host/git/NuRaft.git"
)
# The goal of secondary urls is to have links to the "source of truth" of
@@ -155,6 +156,7 @@ 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
@@ -268,6 +270,8 @@ pushd jemalloc
MALLOC_CONF="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000" \
./configure \
--disable-cxx \
--with-lg-page=12 \
--with-lg-hugepage=21 \
--enable-shared=no --prefix=$working_dir \
--with-malloc-conf="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000"
@@ -277,3 +281,10 @@ popd
#range-v3 release-0.12.0
range_v3_ref="release-0.12.0"
repo_clone_try_double "${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

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

View File

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

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Licensed as a Memgraph Enterprise file under the Memgraph Enterprise
// License (the "License"); by using this file, you agree to be bound by the terms of the License, and you may not use
@@ -8,17 +8,18 @@
#include "auth/auth.hpp"
#include <cstring>
#include <iostream>
#include <limits>
#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/logging.hpp"
#include "utils/message.hpp"
#include "utils/settings.hpp"
#include "utils/string.hpp"
@@ -43,9 +44,89 @@ DEFINE_VALIDATED_int32(auth_module_timeout_ms, 10000,
FLAG_IN_RANGE(100, 1800000));
namespace memgraph::auth {
namespace {
#ifdef MG_ENTERPRISE
/**
* REPLICATION SYSTEM ACTION IMPLEMENTATIONS
*/
struct UpdateAuthData : memgraph::system::ISystemAction {
explicit UpdateAuthData(User user) : user_{std::move(user)}, role_{std::nullopt} {}
explicit UpdateAuthData(Role role) : user_{std::nullopt}, role_{std::move(role)} {}
void DoDurability() override { /* Done during Auth execution */
}
bool DoReplication(replication::ReplicationClient &client, const utils::UUID &main_uuid,
replication::ReplicationEpoch const &epoch,
memgraph::system::Transaction const &txn) const override {
auto check_response = [](const replication::UpdateAuthDataRes &response) { return response.success; };
if (user_) {
return client.SteamAndFinalizeDelta<replication::UpdateAuthDataRpc>(
check_response, main_uuid, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(),
*user_);
}
if (role_) {
return client.SteamAndFinalizeDelta<replication::UpdateAuthDataRpc>(
check_response, main_uuid, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(),
*role_);
}
// Should never get here
MG_ASSERT(false, "Trying to update auth data that is not a user nor a role");
return {};
}
void PostReplication(replication::RoleMainData &mainData) const override {}
private:
std::optional<User> user_;
std::optional<Role> role_;
};
struct DropAuthData : memgraph::system::ISystemAction {
enum class AuthDataType { USER, ROLE };
explicit DropAuthData(AuthDataType type, std::string_view name) : type_{type}, name_{name} {}
void DoDurability() override { /* Done during Auth execution */
}
bool DoReplication(replication::ReplicationClient &client, const utils::UUID &main_uuid,
replication::ReplicationEpoch const &epoch,
memgraph::system::Transaction const &txn) const override {
auto check_response = [](const replication::DropAuthDataRes &response) { return response.success; };
memgraph::replication::DropAuthDataReq::DataType type{};
switch (type_) {
case AuthDataType::USER:
type = memgraph::replication::DropAuthDataReq::DataType::USER;
break;
case AuthDataType::ROLE:
type = memgraph::replication::DropAuthDataReq::DataType::ROLE;
break;
}
return client.SteamAndFinalizeDelta<replication::DropAuthDataRpc>(
check_response, main_uuid, std::string{epoch.id()}, txn.last_committed_system_timestamp(), txn.timestamp(),
type, name_);
}
void PostReplication(replication::RoleMainData &mainData) const override {}
private:
AuthDataType type_;
std::string name_;
};
#endif
/**
* CONSTANTS
*/
const std::string kUserPrefix = "user:";
const std::string kRolePrefix = "role:";
const std::string kLinkPrefix = "link:";
const std::string kVersion = "version";
static constexpr auto kVersionV1 = "V1";
} // namespace
/**
* All data stored in the `Auth` storage is stored in an underlying
@@ -64,7 +145,59 @@ const std::string kLinkPrefix = "link:";
* key="link:<username>", value="<rolename>"
*/
Auth::Auth(const std::string &storage_directory) : storage_(storage_directory), module_(FLAGS_auth_module_executable) {}
namespace {
void MigrateVersions(kvstore::KVStore &store) {
static constexpr auto kPasswordHashV0V1 = "password_hash";
auto version_str = store.Get(kVersion);
if (!version_str) {
using namespace std::string_literals;
// pre versioning, add version to the store
auto puts = std::map<std::string, std::string>{{kVersion, kVersionV1}};
// also add hash kind into durability
auto it = store.begin(kUserPrefix);
auto const e = store.end(kUserPrefix);
if (it != e) {
const auto hash_algo = CurrentHashAlgorithm();
spdlog::info("Updating auth durability, assuming previously stored as {}", AsString(hash_algo));
for (; it != e; ++it) {
auto const &[key, value] = *it;
try {
auto user_data = nlohmann::json::parse(value);
auto password_hash = user_data[kPasswordHashV0V1];
if (!password_hash.is_string()) {
throw AuthException("Couldn't load user data!");
}
// upgrade the password_hash to include the hash algortihm
if (password_hash.empty()) {
user_data[kPasswordHashV0V1] = nullptr;
} else {
user_data[kPasswordHashV0V1] = HashedPassword{hash_algo, password_hash};
}
puts.emplace(key, user_data.dump());
} catch (const nlohmann::json::parse_error &e) {
throw AuthException("Couldn't load user data!");
}
}
}
// Perform migration to V1
store.PutMultiple(puts);
version_str = kVersionV1;
}
}
}; // namespace
Auth::Auth(std::string storage_directory, Config config)
: storage_(std::move(storage_directory)), module_(FLAGS_auth_module_executable), config_{std::move(config)} {
MigrateVersions(storage_);
}
std::optional<User> Auth::Authenticate(const std::string &username, const std::string &password) {
if (module_.IsUsed()) {
@@ -95,6 +228,12 @@ std::optional<User> Auth::Authenticate(const std::string &username, const std::s
// Authenticate the user.
if (!is_authenticated) return std::nullopt;
/**
* TODO
* The auth module should not update auth data.
* There is now way to replicate it and we should not be storing sensitive data if we don't have to.
*/
// Find or create the user and return it.
auto user = GetUser(username);
if (!user) {
@@ -113,7 +252,7 @@ std::optional<User> Auth::Authenticate(const std::string &username, const std::s
return std::nullopt;
}
} else {
user->UpdatePassword(password);
UpdatePassword(*user, password);
}
if (FLAGS_auth_module_manage_roles) {
if (!rolename.empty()) {
@@ -155,6 +294,10 @@ std::optional<User> Auth::Authenticate(const std::string &username, const std::s
username, "https://memgr.ph/auth"));
return std::nullopt;
}
if (user->UpgradeHash(password)) {
SaveUser(*user);
}
return user;
}
}
@@ -183,7 +326,7 @@ std::optional<User> Auth::GetUser(const std::string &username_orig) const {
return user;
}
void Auth::SaveUser(const User &user) {
void Auth::SaveUser(const User &user, system::Transaction *system_tx) {
bool success = false;
if (const auto *role = user.role(); role != nullptr) {
success = storage_.PutMultiple(
@@ -195,26 +338,72 @@ void Auth::SaveUser(const User &user) {
if (!success) {
throw AuthException("Couldn't save user '{}'!", user.username());
}
// All changes to the user end up calling this function, so no need to add a delta anywhere else
if (system_tx) {
#ifdef MG_ENTERPRISE
system_tx->AddAction<UpdateAuthData>(user);
#endif
}
}
std::optional<User> Auth::AddUser(const std::string &username, const std::optional<std::string> &password) {
void Auth::UpdatePassword(auth::User &user, const std::optional<std::string> &password) {
// Check if null
if (!password) {
if (!config_.password_permit_null) {
throw AuthException("Null passwords aren't permitted!");
}
} else {
// Check if compliant with our filter
if (config_.custom_password_regex) {
if (const auto license_check_result = license::global_license_checker.IsEnterpriseValid(utils::global_settings);
license_check_result.HasError()) {
throw AuthException(
"Custom password regex is a Memgraph Enterprise feature. Please set the config "
"(\"--auth-password-strength-regex\") to its default value (\"{}\") or remove the flag.\n{}",
glue::kDefaultPasswordRegex,
license::LicenseCheckErrorToString(license_check_result.GetError(), "password regex"));
}
}
if (!std::regex_match(*password, config_.password_regex)) {
throw AuthException(
"The user password doesn't conform to the required strength! Regex: "
"\"{}\"",
config_.password_regex_str);
}
}
// All checks passed; update
user.UpdatePassword(password);
}
std::optional<User> Auth::AddUser(const std::string &username, const std::optional<std::string> &password,
system::Transaction *system_tx) {
if (!NameRegexMatch(username)) {
throw AuthException("Invalid user name.");
}
auto existing_user = GetUser(username);
if (existing_user) return std::nullopt;
auto existing_role = GetRole(username);
if (existing_role) return std::nullopt;
auto new_user = User(username);
new_user.UpdatePassword(password);
SaveUser(new_user);
UpdatePassword(new_user, password);
SaveUser(new_user, system_tx);
return new_user;
}
bool Auth::RemoveUser(const std::string &username_orig) {
bool Auth::RemoveUser(const std::string &username_orig, system::Transaction *system_tx) {
auto username = utils::ToLowerCase(username_orig);
if (!storage_.Get(kUserPrefix + username)) return false;
std::vector<std::string> keys({kLinkPrefix + username, kUserPrefix + username});
if (!storage_.DeleteMultiple(keys)) {
throw AuthException("Couldn't remove user '{}'!", username);
}
// Handling drop user delta
if (system_tx) {
#ifdef MG_ENTERPRISE
system_tx->AddAction<DropAuthData>(DropAuthData::AuthDataType::USER, username);
#endif
}
return true;
}
@@ -231,6 +420,19 @@ std::vector<auth::User> Auth::AllUsers() const {
return ret;
}
std::vector<std::string> Auth::AllUsernames() const {
std::vector<std::string> ret;
for (auto it = storage_.begin(kUserPrefix); it != storage_.end(kUserPrefix); ++it) {
auto username = it->first.substr(kUserPrefix.size());
if (username != utils::ToLowerCase(username)) continue;
auto user = GetUser(username);
if (user) {
ret.push_back(username);
}
}
return ret;
}
bool Auth::HasUsers() const { return storage_.begin(kUserPrefix) != storage_.end(kUserPrefix); }
std::optional<Role> Auth::GetRole(const std::string &rolename_orig) const {
@@ -248,21 +450,30 @@ std::optional<Role> Auth::GetRole(const std::string &rolename_orig) const {
return Role::Deserialize(data);
}
void Auth::SaveRole(const Role &role) {
void Auth::SaveRole(const Role &role, system::Transaction *system_tx) {
if (!storage_.Put(kRolePrefix + role.rolename(), role.Serialize().dump())) {
throw AuthException("Couldn't save role '{}'!", role.rolename());
}
// All changes to the role end up calling this function, so no need to add a delta anywhere else
if (system_tx) {
#ifdef MG_ENTERPRISE
system_tx->AddAction<UpdateAuthData>(role);
#endif
}
}
std::optional<Role> Auth::AddRole(const std::string &rolename) {
std::optional<Role> Auth::AddRole(const std::string &rolename, system::Transaction *system_tx) {
if (!NameRegexMatch(rolename)) {
throw AuthException("Invalid role name.");
}
if (auto existing_role = GetRole(rolename)) return std::nullopt;
if (auto existing_user = GetUser(rolename)) return std::nullopt;
auto new_role = Role(rolename);
SaveRole(new_role);
SaveRole(new_role, system_tx);
return new_role;
}
bool Auth::RemoveRole(const std::string &rolename_orig) {
bool Auth::RemoveRole(const std::string &rolename_orig, system::Transaction *system_tx) {
auto rolename = utils::ToLowerCase(rolename_orig);
if (!storage_.Get(kRolePrefix + rolename)) return false;
std::vector<std::string> keys;
@@ -275,6 +486,12 @@ bool Auth::RemoveRole(const std::string &rolename_orig) {
if (!storage_.DeleteMultiple(keys)) {
throw AuthException("Couldn't remove role '{}'!", rolename);
}
// Handling drop role delta
if (system_tx) {
#ifdef MG_ENTERPRISE
system_tx->AddAction<DropAuthData>(DropAuthData::AuthDataType::ROLE, rolename);
#endif
}
return true;
}
@@ -292,6 +509,18 @@ std::vector<auth::Role> Auth::AllRoles() const {
return ret;
}
std::vector<std::string> Auth::AllRolenames() const {
std::vector<std::string> ret;
for (auto it = storage_.begin(kRolePrefix); it != storage_.end(kRolePrefix); ++it) {
auto rolename = it->first.substr(kRolePrefix.size());
if (rolename != utils::ToLowerCase(rolename)) continue;
if (auto role = GetRole(rolename)) {
ret.push_back(rolename);
}
}
return ret;
}
std::vector<auth::User> Auth::AllUsersForRole(const std::string &rolename_orig) const {
const auto rolename = utils::ToLowerCase(rolename_orig);
std::vector<auth::User> ret;
@@ -311,52 +540,67 @@ std::vector<auth::User> Auth::AllUsersForRole(const std::string &rolename_orig)
}
#ifdef MG_ENTERPRISE
bool Auth::GrantDatabaseToUser(const std::string &db, const std::string &name) {
bool Auth::GrantDatabaseToUser(const std::string &db, const std::string &name, system::Transaction *system_tx) {
if (auto user = GetUser(name)) {
if (db == kAllDatabases) {
user->db_access().GrantAll();
} else {
user->db_access().Add(db);
}
SaveUser(*user);
SaveUser(*user, system_tx);
return true;
}
return false;
}
bool Auth::RevokeDatabaseFromUser(const std::string &db, const std::string &name) {
bool Auth::RevokeDatabaseFromUser(const std::string &db, const std::string &name, system::Transaction *system_tx) {
if (auto user = GetUser(name)) {
if (db == kAllDatabases) {
user->db_access().DenyAll();
} else {
user->db_access().Remove(db);
}
SaveUser(*user);
SaveUser(*user, system_tx);
return true;
}
return false;
}
void Auth::DeleteDatabase(const std::string &db) {
void Auth::DeleteDatabase(const std::string &db, system::Transaction *system_tx) {
for (auto it = storage_.begin(kUserPrefix); it != storage_.end(kUserPrefix); ++it) {
auto username = it->first.substr(kUserPrefix.size());
if (auto user = GetUser(username)) {
user->db_access().Delete(db);
SaveUser(*user);
SaveUser(*user, system_tx);
}
}
}
bool Auth::SetMainDatabase(std::string_view db, const std::string &name) {
bool Auth::SetMainDatabase(std::string_view db, const std::string &name, system::Transaction *system_tx) {
if (auto user = GetUser(name)) {
if (!user->db_access().SetDefault(db)) {
throw AuthException("Couldn't set default database '{}' for user '{}'!", db, name);
}
SaveUser(*user);
SaveUser(*user, system_tx);
return true;
}
return false;
}
#endif
bool Auth::NameRegexMatch(const std::string &user_or_role) const {
if (config_.custom_name_regex) {
if (const auto license_check_result =
memgraph::license::global_license_checker.IsEnterpriseValid(memgraph::utils::global_settings);
license_check_result.HasError()) {
throw memgraph::auth::AuthException(
"Custom user/role regex is a Memgraph Enterprise feature. Please set the config "
"(\"--auth-user-or-role-name-regex\") to its default value (\"{}\") or remove the flag.\n{}",
glue::kDefaultUserRoleRegex,
memgraph::license::LicenseCheckErrorToString(license_check_result.GetError(), "user/role regex"));
}
}
return std::regex_match(user_or_role, config_.name_regex);
}
} // namespace memgraph::auth

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Licensed as a Memgraph Enterprise file under the Memgraph Enterprise
// License (the "License"); by using this file, you agree to be bound by the terms of the License, and you may not use
@@ -10,16 +10,23 @@
#include <mutex>
#include <optional>
#include <regex>
#include <vector>
#include "auth/exceptions.hpp"
#include "auth/models.hpp"
#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 = "*";
/**
@@ -31,7 +38,47 @@ static const constexpr char *const kAllDatabases = "*";
*/
class Auth final {
public:
explicit Auth(const std::string &storage_directory);
struct Config {
Config() {}
Config(std::string name_regex, std::string password_regex, bool password_permit_null)
: name_regex_str{std::move(name_regex)},
password_regex_str{std::move(password_regex)},
password_permit_null{password_permit_null},
custom_name_regex{name_regex_str != glue::kDefaultUserRoleRegex},
name_regex{name_regex_str},
custom_password_regex{password_regex_str != glue::kDefaultPasswordRegex},
password_regex{password_regex_str} {}
std::string name_regex_str{glue::kDefaultUserRoleRegex};
std::string password_regex_str{glue::kDefaultPasswordRegex};
bool password_permit_null{true};
private:
friend class Auth;
bool custom_name_regex{false};
std::regex name_regex{name_regex_str};
bool custom_password_regex{false};
std::regex password_regex{password_regex_str};
};
explicit Auth(std::string storage_directory, Config config);
/**
* @brief Set the Config object
*
* @param config
*/
void SetConfig(Config config) {
// NOTE: The Auth class itself is not thread-safe, higher-level code needs to synchronize it when using it.
config_ = std::move(config);
}
/**
* @brief
*
* @return Config
*/
Config GetConfig() const { return config_; }
/**
* Authenticates a user using his username and password.
@@ -61,7 +108,7 @@ class Auth final {
*
* @throw AuthException if unable to save the user.
*/
void SaveUser(const User &user);
void SaveUser(const User &user, system::Transaction *system_tx = nullptr);
/**
* Creates a user if the user doesn't exist.
@@ -72,7 +119,8 @@ class Auth final {
* @return a user when the user is created, nullopt if the user exists
* @throw AuthException if unable to save the user.
*/
std::optional<User> AddUser(const std::string &username, const std::optional<std::string> &password = std::nullopt);
std::optional<User> AddUser(const std::string &username, const std::optional<std::string> &password = std::nullopt,
system::Transaction *system_tx = nullptr);
/**
* Removes a user from the storage.
@@ -83,7 +131,15 @@ class Auth final {
* doesn't exist
* @throw AuthException if unable to remove the user.
*/
bool RemoveUser(const std::string &username);
bool RemoveUser(const std::string &username, system::Transaction *system_tx = nullptr);
/**
* @brief
*
* @param user
* @param password
*/
void UpdatePassword(auth::User &user, const std::optional<std::string> &password);
/**
* Gets all users from the storage.
@@ -93,6 +149,13 @@ class Auth final {
*/
std::vector<User> AllUsers() const;
/**
* @brief
*
* @return std::vector<std::string>
*/
std::vector<std::string> AllUsernames() const;
/**
* Returns whether there are users in the storage.
*
@@ -117,7 +180,7 @@ class Auth final {
*
* @throw AuthException if unable to save the role.
*/
void SaveRole(const Role &role);
void SaveRole(const Role &role, system::Transaction *system_tx = nullptr);
/**
* Creates a role if the role doesn't exist.
@@ -127,7 +190,7 @@ class Auth final {
* @return a role when the role is created, nullopt if the role exists
* @throw AuthException if unable to save the role.
*/
std::optional<Role> AddRole(const std::string &rolename);
std::optional<Role> AddRole(const std::string &rolename, system::Transaction *system_tx = nullptr);
/**
* Removes a role from the storage.
@@ -138,7 +201,7 @@ class Auth final {
* doesn't exist
* @throw AuthException if unable to remove the role.
*/
bool RemoveRole(const std::string &rolename);
bool RemoveRole(const std::string &rolename, system::Transaction *system_tx = nullptr);
/**
* Gets all roles from the storage.
@@ -148,6 +211,13 @@ class Auth final {
*/
std::vector<Role> AllRoles() const;
/**
* @brief
*
* @return std::vector<std::string>
*/
std::vector<std::string> AllRolenames() const;
/**
* Gets all users for a role from the storage.
*
@@ -167,7 +237,7 @@ class Auth final {
* @return true on success
* @throw AuthException if unable to find or update the user
*/
bool RevokeDatabaseFromUser(const std::string &db, const std::string &name);
bool RevokeDatabaseFromUser(const std::string &db, const std::string &name, system::Transaction *system_tx = nullptr);
/**
* @brief Grant access to individual database for a user.
@@ -177,7 +247,7 @@ class Auth final {
* @return true on success
* @throw AuthException if unable to find or update the user
*/
bool GrantDatabaseToUser(const std::string &db, const std::string &name);
bool GrantDatabaseToUser(const std::string &db, const std::string &name, system::Transaction *system_tx = nullptr);
/**
* @brief Delete a database from all users.
@@ -185,7 +255,7 @@ class Auth final {
* @param db name of the database to delete
* @throw AuthException if unable to read data
*/
void DeleteDatabase(const std::string &db);
void DeleteDatabase(const std::string &db, system::Transaction *system_tx = nullptr);
/**
* @brief Set main database for an individual user.
@@ -195,14 +265,24 @@ class Auth final {
* @return true on success
* @throw AuthException if unable to find or update the user
*/
bool SetMainDatabase(std::string_view db, const std::string &name);
bool SetMainDatabase(std::string_view db, const std::string &name, system::Transaction *system_tx = nullptr);
#endif
private:
/**
* @brief
*
* @param user_or_role
* @return true
* @return false
*/
bool NameRegexMatch(const std::string &user_or_role) const;
// Even though the `kvstore::KVStore` class is guaranteed to be thread-safe,
// Auth is not thread-safe because modifying users and roles might require
// more than one operation on the storage.
kvstore::KVStore storage_;
auth::Module module_;
Config config_;
};
} // namespace memgraph::auth

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Licensed as a Memgraph Enterprise file under the Memgraph Enterprise
// License (the "License"); by using this file, you agree to be bound by the terms of the License, and you may not use
@@ -22,10 +22,14 @@
namespace {
using namespace std::literals;
inline constexpr std::array password_encryption_mappings{
std::pair{"bcrypt"sv, memgraph::auth::PasswordEncryptionAlgorithm::BCRYPT},
std::pair{"sha256"sv, memgraph::auth::PasswordEncryptionAlgorithm::SHA256},
std::pair{"sha256-multiple"sv, memgraph::auth::PasswordEncryptionAlgorithm::SHA256_MULTIPLE}};
constexpr auto kHashAlgo = "hash_algo";
constexpr auto kPasswordHash = "password_hash";
inline constexpr std::array password_hash_mappings{
std::pair{"bcrypt"sv, memgraph::auth::PasswordHashAlgorithm::BCRYPT},
std::pair{"sha256"sv, memgraph::auth::PasswordHashAlgorithm::SHA256},
std::pair{"sha256-multiple"sv, memgraph::auth::PasswordHashAlgorithm::SHA256_MULTIPLE}};
inline constexpr uint64_t ONE_SHA_ITERATION = 1;
inline constexpr uint64_t MULTIPLE_SHA_ITERATIONS = 1024;
@@ -35,7 +39,7 @@ inline constexpr uint64_t MULTIPLE_SHA_ITERATIONS = 1024;
DEFINE_VALIDATED_string(password_encryption_algorithm, "bcrypt",
"The password encryption algorithm used for authentication.", {
if (const auto result =
memgraph::utils::IsValidEnumValueString(value, password_encryption_mappings);
memgraph::utils::IsValidEnumValueString(value, password_hash_mappings);
result.HasError()) {
const auto error = result.GetError();
switch (error) {
@@ -45,7 +49,7 @@ DEFINE_VALIDATED_string(password_encryption_algorithm, "bcrypt",
}
case memgraph::utils::ValidationError::InvalidValue: {
std::cout << "Invalid value for password encryption algorithm. Allowed values: "
<< memgraph::utils::GetAllowedEnumValuesString(password_encryption_mappings)
<< memgraph::utils::GetAllowedEnumValuesString(password_hash_mappings)
<< std::endl;
break;
}
@@ -58,7 +62,7 @@ DEFINE_VALIDATED_string(password_encryption_algorithm, "bcrypt",
namespace memgraph::auth {
namespace BCrypt {
std::string EncryptPassword(const std::string &password) {
std::string HashPassword(const std::string &password) {
char salt[BCRYPT_HASHSIZE];
char hash[BCRYPT_HASHSIZE];
@@ -86,16 +90,30 @@ bool VerifyPassword(const std::string &password, const std::string &hash) {
} // namespace BCrypt
namespace SHA {
namespace {
constexpr auto SHA_LENGTH = 64U;
constexpr auto SALT_SIZE = 16U;
constexpr auto SALT_SIZE_DURABLE = SALT_SIZE * 2;
#if OPENSSL_VERSION_MAJOR >= 3
std::string EncryptPasswordOpenSSL3(const std::string &password, const uint64_t number_of_iterations) {
std::string HashPasswordOpenSSL3(std::string_view password, const uint64_t number_of_iterations,
std::string_view salt) {
unsigned char hash[SHA256_DIGEST_LENGTH];
EVP_MD_CTX *ctx = EVP_MD_CTX_new();
EVP_MD *md = EVP_MD_fetch(nullptr, "SHA2-256", nullptr);
EVP_DigestInit_ex(ctx, md, nullptr);
if (!salt.empty()) {
DMG_ASSERT(salt.size() == SALT_SIZE);
EVP_DigestUpdate(ctx, salt.data(), salt.size());
}
for (auto i = 0; i < number_of_iterations; i++) {
EVP_DigestUpdate(ctx, password.c_str(), password.size());
EVP_DigestUpdate(ctx, password.data(), password.size());
}
EVP_DigestFinal_ex(ctx, hash, nullptr);
@@ -103,6 +121,11 @@ std::string EncryptPasswordOpenSSL3(const std::string &password, const uint64_t
EVP_MD_CTX_free(ctx);
std::stringstream result_stream;
for (unsigned char salt_char : salt) {
result_stream << std::hex << std::setw(2) << std::setfill('0') << (((unsigned int)salt_char) & 0xFFU);
}
for (auto hash_char : hash) {
result_stream << std::hex << std::setw(2) << std::setfill('0') << (int)hash_char;
}
@@ -110,17 +133,27 @@ std::string EncryptPasswordOpenSSL3(const std::string &password, const uint64_t
return result_stream.str();
}
#else
std::string EncryptPasswordOpenSSL1_1(const std::string &password, const uint64_t number_of_iterations) {
std::string HashPasswordOpenSSL1_1(std::string_view password, const uint64_t number_of_iterations,
std::string_view salt) {
unsigned char hash[SHA256_DIGEST_LENGTH];
SHA256_CTX sha256;
SHA256_Init(&sha256);
if (!salt.empty()) {
DMG_ASSERT(salt.size() == SALT_SIZE);
SHA256_Update(&sha256, salt.data(), salt.size());
}
for (auto i = 0; i < number_of_iterations; i++) {
SHA256_Update(&sha256, password.c_str(), password.size());
SHA256_Update(&sha256, password.data(), password.size());
}
SHA256_Final(hash, &sha256);
std::stringstream ss;
for (unsigned char salt_char : salt) {
ss << std::hex << std::setw(2) << std::setfill('0') << (((unsigned int)salt_char) & 0xFFU);
}
for (auto hash_char : hash) {
ss << std::hex << std::setw(2) << std::setfill('0') << (int)hash_char;
}
@@ -129,55 +162,144 @@ std::string EncryptPasswordOpenSSL1_1(const std::string &password, const uint64_
}
#endif
std::string EncryptPassword(const std::string &password, const uint64_t number_of_iterations) {
std::string HashPassword(std::string_view password, const uint64_t number_of_iterations, std::string_view salt) {
#if OPENSSL_VERSION_MAJOR >= 3
return EncryptPasswordOpenSSL3(password, number_of_iterations);
return HashPasswordOpenSSL3(password, number_of_iterations, salt);
#else
return EncryptPasswordOpenSSL1_1(password, number_of_iterations);
return HashPasswordOpenSSL1_1(password, number_of_iterations, salt);
#endif
}
bool VerifyPassword(const std::string &password, const std::string &hash, const uint64_t number_of_iterations) {
auto password_hash = EncryptPassword(password, number_of_iterations);
auto ExtractSalt(std::string_view salt_durable) -> std::array<char, SALT_SIZE> {
static_assert(SALT_SIZE_DURABLE % 2 == 0);
static_assert(SALT_SIZE_DURABLE / 2 == SALT_SIZE);
MG_ASSERT(salt_durable.size() == SALT_SIZE_DURABLE);
auto const *b = salt_durable.cbegin();
auto const *const e = salt_durable.cend();
auto salt = std::array<char, SALT_SIZE>{};
auto *inserter = salt.begin();
auto const toval = [](char a) -> uint8_t {
if ('0' <= a && a <= '9') {
return a - '0';
}
if ('a' <= a && a <= 'f') {
return 10 + (a - 'a');
}
MG_ASSERT(false, "Currupt hash, can't extract salt");
__builtin_unreachable();
};
for (; b != e; b += 2, ++inserter) {
*inserter = static_cast<char>(static_cast<uint8_t>(toval(b[0]) << 4U) | toval(b[1]));
}
return salt;
}
bool IsSalted(std::string_view hash) { return hash.size() == SHA_LENGTH + SALT_SIZE_DURABLE; }
bool VerifyPassword(std::string_view password, std::string_view hash, const uint64_t number_of_iterations) {
auto password_hash = std::invoke([&] {
if (hash.size() == SHA_LENGTH) [[unlikely]] {
// Just SHA256
return HashPassword(password, number_of_iterations, {});
} else {
// SHA256 + SALT
MG_ASSERT(IsSalted(hash));
auto const salt_durable = std::string_view{hash.data(), SALT_SIZE_DURABLE};
std::array<char, SALT_SIZE> salt = ExtractSalt(salt_durable);
return HashPassword(password, number_of_iterations, {salt.data(), salt.size()});
}
});
return password_hash == hash;
}
} // namespace
} // namespace SHA
bool VerifyPassword(const std::string &password, const std::string &hash) {
const auto password_encryption_algorithm = utils::StringToEnum<PasswordEncryptionAlgorithm>(
FLAGS_password_encryption_algorithm, password_encryption_mappings);
HashedPassword HashPassword(const std::string &password, std::optional<PasswordHashAlgorithm> override_algo) {
auto const hash_algo = override_algo.value_or(CurrentHashAlgorithm());
auto password_hash = std::invoke([&] {
switch (hash_algo) {
case PasswordHashAlgorithm::BCRYPT: {
return BCrypt::HashPassword(password);
}
case PasswordHashAlgorithm::SHA256:
case PasswordHashAlgorithm::SHA256_MULTIPLE: {
auto gen = std::mt19937(std::random_device{}());
auto salt = std::array<char, SHA::SALT_SIZE>{};
auto dis = std::uniform_int_distribution<unsigned char>(0, 255);
std::generate(salt.begin(), salt.end(), [&]() { return dis(gen); });
auto iterations = (hash_algo == PasswordHashAlgorithm::SHA256) ? ONE_SHA_ITERATION : MULTIPLE_SHA_ITERATIONS;
return SHA::HashPassword(password, iterations, {salt.data(), salt.size()});
}
}
});
return HashedPassword{hash_algo, std::move(password_hash)};
};
if (!password_encryption_algorithm.has_value()) {
throw AuthException("Invalid password encryption flag '{}'!", FLAGS_password_encryption_algorithm);
namespace {
auto InternalParseHashAlgorithm(std::string_view algo) -> PasswordHashAlgorithm {
auto maybe_parsed = utils::StringToEnum<PasswordHashAlgorithm>(algo, password_hash_mappings);
if (!maybe_parsed) {
throw AuthException("Invalid password encryption '{}'!", algo);
}
switch (password_encryption_algorithm.value()) {
case PasswordEncryptionAlgorithm::BCRYPT:
return BCrypt::VerifyPassword(password, hash);
case PasswordEncryptionAlgorithm::SHA256:
return SHA::VerifyPassword(password, hash, ONE_SHA_ITERATION);
case PasswordEncryptionAlgorithm::SHA256_MULTIPLE:
return SHA::VerifyPassword(password, hash, MULTIPLE_SHA_ITERATIONS);
}
throw AuthException("Invalid password encryption flag '{}'!", FLAGS_password_encryption_algorithm);
return *maybe_parsed;
}
std::string EncryptPassword(const std::string &password) {
const auto password_encryption_algorithm = utils::StringToEnum<PasswordEncryptionAlgorithm>(
FLAGS_password_encryption_algorithm, password_encryption_mappings);
PasswordHashAlgorithm &InternalCurrentHashAlgorithm() {
static auto current = PasswordHashAlgorithm::BCRYPT;
static std::once_flag flag;
std::call_once(flag, [] { current = InternalParseHashAlgorithm(FLAGS_password_encryption_algorithm); });
return current;
}
} // namespace
if (!password_encryption_algorithm.has_value()) {
throw AuthException("Invalid password encryption flag '{}'!", FLAGS_password_encryption_algorithm);
auto CurrentHashAlgorithm() -> PasswordHashAlgorithm { return InternalCurrentHashAlgorithm(); }
void SetHashAlgorithm(std::string_view algo) {
auto &current = InternalCurrentHashAlgorithm();
current = InternalParseHashAlgorithm(algo);
}
auto AsString(PasswordHashAlgorithm hash_algo) -> std::string_view {
return *utils::EnumToString<PasswordHashAlgorithm>(hash_algo, password_hash_mappings);
}
bool HashedPassword::VerifyPassword(const std::string &password) {
switch (hash_algo) {
case PasswordHashAlgorithm::BCRYPT:
return BCrypt::VerifyPassword(password, password_hash);
case PasswordHashAlgorithm::SHA256:
return SHA::VerifyPassword(password, password_hash, ONE_SHA_ITERATION);
case PasswordHashAlgorithm::SHA256_MULTIPLE:
return SHA::VerifyPassword(password, password_hash, MULTIPLE_SHA_ITERATIONS);
}
}
switch (password_encryption_algorithm.value()) {
case PasswordEncryptionAlgorithm::BCRYPT:
return BCrypt::EncryptPassword(password);
case PasswordEncryptionAlgorithm::SHA256:
return SHA::EncryptPassword(password, ONE_SHA_ITERATION);
case PasswordEncryptionAlgorithm::SHA256_MULTIPLE:
return SHA::EncryptPassword(password, MULTIPLE_SHA_ITERATIONS);
void to_json(nlohmann::json &j, const HashedPassword &p) {
j = nlohmann::json{{kHashAlgo, p.hash_algo}, {kPasswordHash, p.password_hash}};
}
void from_json(const nlohmann::json &j, HashedPassword &p) {
// NOLINTNEXTLINE(cppcoreguidelines-init-variables)
PasswordHashAlgorithm hash_algo;
j.at(kHashAlgo).get_to(hash_algo);
auto password_hash = j.value(kPasswordHash, std::string());
p = HashedPassword{hash_algo, std::move(password_hash)};
}
bool HashedPassword::IsSalted() const {
switch (hash_algo) {
case PasswordHashAlgorithm::BCRYPT:
return true;
case PasswordHashAlgorithm::SHA256:
case PasswordHashAlgorithm::SHA256_MULTIPLE:
return SHA::IsSalted(password_hash);
}
}

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Licensed as a Memgraph Enterprise file under the Memgraph Enterprise
// License (the "License"); by using this file, you agree to be bound by the terms of the License, and you may not use
@@ -8,14 +8,45 @@
#pragma once
#include <json/json.hpp>
#include <optional>
#include <string>
namespace memgraph::auth {
enum class PasswordEncryptionAlgorithm : uint8_t { BCRYPT, SHA256, SHA256_MULTIPLE };
/// Need to be stable, auth durability depends on this
enum class PasswordHashAlgorithm : uint8_t { BCRYPT = 0, SHA256 = 1, SHA256_MULTIPLE = 2 };
/// @throw AuthException if unable to encrypt the password.
std::string EncryptPassword(const std::string &password);
void SetHashAlgorithm(std::string_view algo);
/// @throw AuthException if unable to verify the password.
bool VerifyPassword(const std::string &password, const std::string &hash);
auto CurrentHashAlgorithm() -> PasswordHashAlgorithm;
auto AsString(PasswordHashAlgorithm hash_algo) -> std::string_view;
struct HashedPassword {
HashedPassword() = default;
HashedPassword(PasswordHashAlgorithm hash_algo, std::string password_hash)
: hash_algo{hash_algo}, password_hash{std::move(password_hash)} {}
HashedPassword(HashedPassword const &) = default;
HashedPassword(HashedPassword &&) = default;
HashedPassword &operator=(HashedPassword const &) = default;
HashedPassword &operator=(HashedPassword &&) = default;
friend bool operator==(HashedPassword const &, HashedPassword const &) = default;
bool VerifyPassword(const std::string &password);
bool IsSalted() const;
auto HashAlgo() const -> PasswordHashAlgorithm { return hash_algo; }
friend void to_json(nlohmann::json &j, const HashedPassword &p);
friend void from_json(const nlohmann::json &j, HashedPassword &p);
private:
PasswordHashAlgorithm hash_algo{PasswordHashAlgorithm::BCRYPT};
std::string password_hash{};
};
/// @throw AuthException if unable to hash the password.
HashedPassword HashPassword(const std::string &password, std::optional<PasswordHashAlgorithm> override_algo = {});
} // namespace memgraph::auth

View File

@@ -9,7 +9,6 @@
#include "auth/models.hpp"
#include <cstdint>
#include <regex>
#include <utility>
#include <gflags/gflags.h>
@@ -21,22 +20,26 @@
#include "query/constants.hpp"
#include "spdlog/spdlog.h"
#include "utils/cast.hpp"
#include "utils/logging.hpp"
#include "utils/settings.hpp"
#include "utils/string.hpp"
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_bool(auth_password_permit_null, true, "Set to false to disable null passwords.");
inline constexpr std::string_view default_password_regex = ".+";
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_string(auth_password_strength_regex, default_password_regex.data(),
"The regular expression that should be used to match the entire "
"entered password to ensure its strength.");
namespace memgraph::auth {
namespace {
constexpr auto kRoleName = "rolename";
constexpr auto kPermissions = "permissions";
constexpr auto kGrants = "grants";
constexpr auto kDenies = "denies";
constexpr auto kUsername = "username";
constexpr auto kPasswordHash = "password_hash";
#ifdef MG_ENTERPRISE
constexpr auto kGlobalPermission = "global_permission";
constexpr auto kFineGrainedAccessHandler = "fine_grained_access_handler";
constexpr auto kAllowAll = "allow_all";
constexpr auto kDefault = "default";
constexpr auto kDatabases = "databases";
#endif
// Constant list of all available permissions.
const std::vector<Permission> kPermissionsAll = {Permission::MATCH,
Permission::CREATE,
@@ -245,8 +248,9 @@ std::vector<Permission> Permissions::GetDenies() const {
nlohmann::json Permissions::Serialize() const {
nlohmann::json data = nlohmann::json::object();
data["grants"] = grants_;
data["denies"] = denies_;
data[kGrants] = grants_;
data[kDenies] = denies_;
return data;
}
@@ -254,10 +258,10 @@ Permissions Permissions::Deserialize(const nlohmann::json &data) {
if (!data.is_object()) {
throw AuthException("Couldn't load permissions data!");
}
if (!data["grants"].is_number_unsigned() || !data["denies"].is_number_unsigned()) {
if (!data[kGrants].is_number_unsigned() || !data[kDenies].is_number_unsigned()) {
throw AuthException("Couldn't load permissions data!");
}
return Permissions{data["grants"], data["denies"]};
return Permissions{data[kGrants], data[kDenies]};
}
uint64_t Permissions::grants() const { return grants_; }
@@ -319,8 +323,8 @@ nlohmann::json FineGrainedAccessPermissions::Serialize() const {
return {};
}
nlohmann::json data = nlohmann::json::object();
data["permissions"] = permissions_;
data["global_permission"] = global_permission_.has_value() ? global_permission_.value() : -1;
data[kPermissions] = permissions_;
data[kGlobalPermission] = global_permission_.has_value() ? global_permission_.value() : -1;
return data;
}
@@ -333,13 +337,13 @@ FineGrainedAccessPermissions FineGrainedAccessPermissions::Deserialize(const nlo
}
std::optional<uint64_t> global_permission;
if (data["global_permission"].empty() || data["global_permission"] == -1) {
if (data[kGlobalPermission].empty() || data[kGlobalPermission] == -1) {
global_permission = std::nullopt;
} else {
global_permission = data["global_permission"];
global_permission = data[kGlobalPermission];
}
return FineGrainedAccessPermissions(data["permissions"], global_permission);
return FineGrainedAccessPermissions(data[kPermissions], global_permission);
}
const std::unordered_map<std::string, uint64_t> &FineGrainedAccessPermissions::GetPermissions() const {
@@ -445,13 +449,13 @@ const FineGrainedAccessPermissions &Role::GetFineGrainedAccessEdgeTypePermission
nlohmann::json Role::Serialize() const {
nlohmann::json data = nlohmann::json::object();
data["rolename"] = rolename_;
data["permissions"] = permissions_.Serialize();
data[kRoleName] = rolename_;
data[kPermissions] = permissions_.Serialize();
#ifdef MG_ENTERPRISE
if (memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
data["fine_grained_access_handler"] = fine_grained_access_handler_.Serialize();
data[kFineGrainedAccessHandler] = fine_grained_access_handler_.Serialize();
} else {
data["fine_grained_access_handler"] = {};
data[kFineGrainedAccessHandler] = {};
}
#endif
return data;
@@ -461,21 +465,21 @@ Role Role::Deserialize(const nlohmann::json &data) {
if (!data.is_object()) {
throw AuthException("Couldn't load role data!");
}
if (!data["rolename"].is_string() || !data["permissions"].is_object()) {
if (!data[kRoleName].is_string() || !data[kPermissions].is_object()) {
throw AuthException("Couldn't load role data!");
}
auto permissions = Permissions::Deserialize(data["permissions"]);
auto permissions = Permissions::Deserialize(data[kPermissions]);
#ifdef MG_ENTERPRISE
if (memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
FineGrainedAccessHandler fine_grained_access_handler;
// We can have an empty fine_grained if the user was created without a valid license
if (data["fine_grained_access_handler"].is_object()) {
fine_grained_access_handler = FineGrainedAccessHandler::Deserialize(data["fine_grained_access_handler"]);
if (data[kFineGrainedAccessHandler].is_object()) {
fine_grained_access_handler = FineGrainedAccessHandler::Deserialize(data[kFineGrainedAccessHandler]);
}
return {data["rolename"], permissions, std::move(fine_grained_access_handler)};
return {data[kRoleName], permissions, std::move(fine_grained_access_handler)};
}
#endif
return {data["rolename"], permissions};
return {data[kRoleName], permissions};
}
bool operator==(const Role &first, const Role &second) {
@@ -545,10 +549,10 @@ const std::string &Databases::GetDefault() const {
nlohmann::json Databases::Serialize() const {
nlohmann::json data = nlohmann::json::object();
data["grants"] = grants_dbs_;
data["denies"] = denies_dbs_;
data["allow_all"] = allow_all_;
data["default"] = default_db_;
data[kGrants] = grants_dbs_;
data[kDenies] = denies_dbs_;
data[kAllowAll] = allow_all_;
data[kDefault] = default_db_;
return data;
}
@@ -556,22 +560,22 @@ Databases Databases::Deserialize(const nlohmann::json &data) {
if (!data.is_object()) {
throw AuthException("Couldn't load database data!");
}
if (!data["grants"].is_structured() || !data["denies"].is_structured() || !data["allow_all"].is_boolean() ||
!data["default"].is_string()) {
if (!data[kGrants].is_structured() || !data[kDenies].is_structured() || !data[kAllowAll].is_boolean() ||
!data[kDefault].is_string()) {
throw AuthException("Couldn't load database data!");
}
return {data["allow_all"], data["grants"], data["denies"], data["default"]};
return {data[kAllowAll], data[kGrants], data[kDenies], data[kDefault]};
}
#endif
User::User() = default;
User::User(const std::string &username) : username_(utils::ToLowerCase(username)) {}
User::User(const std::string &username, std::string password_hash, const Permissions &permissions)
User::User(const std::string &username, std::optional<HashedPassword> password_hash, const Permissions &permissions)
: username_(utils::ToLowerCase(username)), password_hash_(std::move(password_hash)), permissions_(permissions) {}
#ifdef MG_ENTERPRISE
User::User(const std::string &username, std::string password_hash, const Permissions &permissions,
User::User(const std::string &username, std::optional<HashedPassword> password_hash, const Permissions &permissions,
FineGrainedAccessHandler fine_grained_access_handler, Databases db_access)
: username_(utils::ToLowerCase(username)),
password_hash_(std::move(password_hash)),
@@ -581,38 +585,16 @@ User::User(const std::string &username, std::string password_hash, const Permiss
#endif
bool User::CheckPassword(const std::string &password) {
if (password_hash_.empty()) return true;
return VerifyPassword(password, password_hash_);
return password_hash_ ? password_hash_->VerifyPassword(password) : true;
}
void User::UpdatePassword(const std::optional<std::string> &password) {
void User::UpdatePassword(const std::optional<std::string> &password,
std::optional<PasswordHashAlgorithm> algo_override) {
if (!password) {
if (!FLAGS_auth_password_permit_null) {
throw AuthException("Null passwords aren't permitted!");
}
password_hash_ = "";
password_hash_.reset();
return;
}
if (FLAGS_auth_password_strength_regex != default_password_regex) {
if (const auto license_check_result = license::global_license_checker.IsEnterpriseValid(utils::global_settings);
license_check_result.HasError()) {
throw AuthException(
"Custom password regex is a Memgraph Enterprise feature. Please set the config "
"(\"--auth-password-strength-regex\") to its default value (\"{}\") or remove the flag.\n{}",
default_password_regex,
license::LicenseCheckErrorToString(license_check_result.GetError(), "password regex"));
}
}
std::regex re(FLAGS_auth_password_strength_regex);
if (!std::regex_match(*password, re)) {
throw AuthException(
"The user password doesn't conform to the required strength! Regex: "
"\"{}\"",
FLAGS_auth_password_strength_regex);
}
password_hash_ = EncryptPassword(*password);
password_hash_ = HashPassword(*password, algo_override);
}
void User::SetRole(const Role &role) { role_.emplace(role); }
@@ -629,27 +611,49 @@ Permissions User::GetPermissions() const {
#ifdef MG_ENTERPRISE
FineGrainedAccessPermissions User::GetFineGrainedAccessLabelPermissions() const {
return Merge(GetUserFineGrainedAccessLabelPermissions(), GetRoleFineGrainedAccessLabelPermissions());
}
FineGrainedAccessPermissions User::GetFineGrainedAccessEdgeTypePermissions() const {
return Merge(GetUserFineGrainedAccessEdgeTypePermissions(), GetRoleFineGrainedAccessEdgeTypePermissions());
}
FineGrainedAccessPermissions User::GetUserFineGrainedAccessEdgeTypePermissions() const {
if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
return FineGrainedAccessPermissions{};
}
if (role_) {
return Merge(role()->fine_grained_access_handler().label_permissions(),
fine_grained_access_handler_.label_permissions());
return fine_grained_access_handler_.edge_type_permissions();
}
FineGrainedAccessPermissions User::GetUserFineGrainedAccessLabelPermissions() const {
if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
return FineGrainedAccessPermissions{};
}
return fine_grained_access_handler_.label_permissions();
}
FineGrainedAccessPermissions User::GetFineGrainedAccessEdgeTypePermissions() const {
FineGrainedAccessPermissions User::GetRoleFineGrainedAccessEdgeTypePermissions() const {
if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
return FineGrainedAccessPermissions{};
}
if (role_) {
return Merge(role()->fine_grained_access_handler().edge_type_permissions(),
fine_grained_access_handler_.edge_type_permissions());
return role()->fine_grained_access_handler().edge_type_permissions();
}
return fine_grained_access_handler_.edge_type_permissions();
return FineGrainedAccessPermissions{};
}
FineGrainedAccessPermissions User::GetRoleFineGrainedAccessLabelPermissions() const {
if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
return FineGrainedAccessPermissions{};
}
if (role_) {
return role()->fine_grained_access_handler().label_permissions();
}
return FineGrainedAccessPermissions{};
}
#endif
@@ -671,16 +675,20 @@ const Role *User::role() const {
nlohmann::json User::Serialize() const {
nlohmann::json data = nlohmann::json::object();
data["username"] = username_;
data["password_hash"] = password_hash_;
data["permissions"] = permissions_.Serialize();
data[kUsername] = username_;
if (password_hash_.has_value()) {
data[kPasswordHash] = *password_hash_;
} else {
data[kPasswordHash] = nullptr;
}
data[kPermissions] = permissions_.Serialize();
#ifdef MG_ENTERPRISE
if (memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
data["fine_grained_access_handler"] = fine_grained_access_handler_.Serialize();
data["databases"] = database_access_.Serialize();
data[kFineGrainedAccessHandler] = fine_grained_access_handler_.Serialize();
data[kDatabases] = database_access_.Serialize();
} else {
data["fine_grained_access_handler"] = {};
data["databases"] = {};
data[kFineGrainedAccessHandler] = {};
data[kDatabases] = {};
}
#endif
// The role shouldn't be serialized here, it is stored as a foreign key.
@@ -691,15 +699,23 @@ User User::Deserialize(const nlohmann::json &data) {
if (!data.is_object()) {
throw AuthException("Couldn't load user data!");
}
if (!data["username"].is_string() || !data["password_hash"].is_string() || !data["permissions"].is_object()) {
auto password_hash_json = data[kPasswordHash];
if (!data[kUsername].is_string() || !(password_hash_json.is_object() || password_hash_json.is_null()) ||
!data[kPermissions].is_object()) {
throw AuthException("Couldn't load user data!");
}
auto permissions = Permissions::Deserialize(data["permissions"]);
std::optional<HashedPassword> password_hash{};
if (password_hash_json.is_object()) {
password_hash = password_hash_json.get<HashedPassword>();
}
auto permissions = Permissions::Deserialize(data[kPermissions]);
#ifdef MG_ENTERPRISE
if (memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
Databases db_access;
if (data["databases"].is_structured()) {
db_access = Databases::Deserialize(data["databases"]);
if (data[kDatabases].is_structured()) {
db_access = Databases::Deserialize(data[kDatabases]);
} else {
// Back-compatibility
spdlog::warn("User without specified database access. Given access to the default database.");
@@ -708,13 +724,13 @@ User User::Deserialize(const nlohmann::json &data) {
}
FineGrainedAccessHandler fine_grained_access_handler;
// We can have an empty fine_grained if the user was created without a valid license
if (data["fine_grained_access_handler"].is_object()) {
fine_grained_access_handler = FineGrainedAccessHandler::Deserialize(data["fine_grained_access_handler"]);
if (data[kFineGrainedAccessHandler].is_object()) {
fine_grained_access_handler = FineGrainedAccessHandler::Deserialize(data[kFineGrainedAccessHandler]);
}
return {data["username"], data["password_hash"], permissions, std::move(fine_grained_access_handler), db_access};
return {data[kUsername], std::move(password_hash), permissions, std::move(fine_grained_access_handler), db_access};
}
#endif
return {data["username"], data["password_hash"], permissions};
return {data[kUsername], std::move(password_hash), permissions};
}
bool operator==(const User &first, const User &second) {

View File

@@ -15,6 +15,7 @@
#include <json/json.hpp>
#include <utility>
#include "crypto.hpp"
#include "dbms/constants.hpp"
#include "utils/logging.hpp"
@@ -206,6 +207,8 @@ bool operator==(const FineGrainedAccessHandler &first, const FineGrainedAccessHa
class Role final {
public:
Role() = default;
explicit Role(const std::string &rolename);
Role(const std::string &rolename, const Permissions &permissions);
#ifdef MG_ENTERPRISE
@@ -332,9 +335,9 @@ class User final {
User();
explicit User(const std::string &username);
User(const std::string &username, std::string password_hash, const Permissions &permissions);
User(const std::string &username, std::optional<HashedPassword> password_hash, const Permissions &permissions);
#ifdef MG_ENTERPRISE
User(const std::string &username, std::string password_hash, const Permissions &permissions,
User(const std::string &username, std::optional<HashedPassword> password_hash, const Permissions &permissions,
FineGrainedAccessHandler fine_grained_access_handler, Databases db_access = {});
#endif
User(const User &) = default;
@@ -346,8 +349,18 @@ class User final {
/// @throw AuthException if unable to verify the password.
bool CheckPassword(const std::string &password);
bool UpgradeHash(const std::string password) {
if (!password_hash_) return false;
if (password_hash_->IsSalted()) return false;
auto const algo = password_hash_->HashAlgo();
UpdatePassword(password, algo);
return true;
}
/// @throw AuthException if unable to set the password.
void UpdatePassword(const std::optional<std::string> &password = std::nullopt);
void UpdatePassword(const std::optional<std::string> &password = {},
std::optional<PasswordHashAlgorithm> algo_override = std::nullopt);
void SetRole(const Role &role);
@@ -358,6 +371,10 @@ class User final {
#ifdef MG_ENTERPRISE
FineGrainedAccessPermissions GetFineGrainedAccessLabelPermissions() const;
FineGrainedAccessPermissions GetFineGrainedAccessEdgeTypePermissions() const;
FineGrainedAccessPermissions GetUserFineGrainedAccessLabelPermissions() const;
FineGrainedAccessPermissions GetUserFineGrainedAccessEdgeTypePermissions() const;
FineGrainedAccessPermissions GetRoleFineGrainedAccessLabelPermissions() const;
FineGrainedAccessPermissions GetRoleFineGrainedAccessEdgeTypePermissions() const;
const FineGrainedAccessHandler &fine_grained_access_handler() const;
FineGrainedAccessHandler &fine_grained_access_handler();
#endif
@@ -382,7 +399,7 @@ class User final {
private:
std::string username_;
std::string password_hash_;
std::optional<HashedPassword> password_hash_;
Permissions permissions_;
#ifdef MG_ENTERPRISE
FineGrainedAccessHandler fine_grained_access_handler_;

View File

@@ -0,0 +1,190 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "auth/replication_handlers.hpp"
#include "auth/auth.hpp"
#include "auth/rpc.hpp"
#include "license/license.hpp"
namespace memgraph::auth {
void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
std::string_view rpc_req) {
spdlog::error(fmt::format("Received {} with main_id: {} != current_main_uuid: {}", rpc_req, std::string(main_req_id),
current_main_uuid.has_value() ? std::string(current_main_uuid.value()) : ""));
}
#ifdef MG_ENTERPRISE
void UpdateAuthDataHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, auth::SynchedAuth &auth,
slk::Reader *req_reader, slk::Builder *res_builder) {
replication::UpdateAuthDataReq req;
memgraph::slk::Load(&req, req_reader);
using memgraph::replication::UpdateAuthDataRes;
UpdateAuthDataRes res(false);
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, replication::UpdateAuthDataReq::kType.name);
memgraph::slk::Save(res, res_builder);
return;
}
// Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
// of the set of databases. Hence no history exists to maintain regarding epoch change.
// If MAIN has changed we need to check this new group_timestamp is consistent with
// what we have so far.
if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
spdlog::debug("UpdateAuthDataHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
system_state_access.LastCommitedTS());
memgraph::slk::Save(res, res_builder);
return;
}
try {
// Update
if (req.user) auth->SaveUser(*req.user);
if (req.role) auth->SaveRole(*req.role);
// Success
system_state_access.SetLastCommitedTS(req.new_group_timestamp);
res = UpdateAuthDataRes(true);
spdlog::debug("UpdateAuthDataHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
} catch (const auth::AuthException & /* not used */) {
// Failure
}
memgraph::slk::Save(res, res_builder);
}
void DropAuthDataHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, auth::SynchedAuth &auth,
slk::Reader *req_reader, slk::Builder *res_builder) {
replication::DropAuthDataReq req;
memgraph::slk::Load(&req, req_reader);
using memgraph::replication::DropAuthDataRes;
DropAuthDataRes res(false);
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, replication::DropAuthDataRes::kType.name);
memgraph::slk::Save(res, res_builder);
return;
}
// Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
// of the set of databases. Hence no history exists to maintain regarding epoch change.
// If MAIN has changed we need to check this new group_timestamp is consistent with
// what we have so far.
if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
spdlog::debug("DropAuthDataHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
system_state_access.LastCommitedTS());
memgraph::slk::Save(res, res_builder);
return;
}
try {
// Remove
switch (req.type) {
case replication::DropAuthDataReq::DataType::USER:
auth->RemoveUser(req.name);
break;
case replication::DropAuthDataReq::DataType::ROLE:
auth->RemoveRole(req.name);
break;
}
// Success
system_state_access.SetLastCommitedTS(req.new_group_timestamp);
res = DropAuthDataRes(true);
spdlog::debug("DropAuthDataHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
} catch (const auth::AuthException & /* not used */) {
// Failure
}
memgraph::slk::Save(res, res_builder);
}
bool SystemRecoveryHandler(auth::SynchedAuth &auth, auth::Auth::Config auth_config,
const std::vector<auth::User> &users, const std::vector<auth::Role> &roles) {
return auth.WithLock([&](auto &locked_auth) {
// Update config
locked_auth.SetConfig(std::move(auth_config));
// Get all current users
auto old_users = locked_auth.AllUsernames();
// Save incoming users
for (const auto &user : users) {
// Missing users
try {
locked_auth.SaveUser(user);
} catch (const auth::AuthException &) {
spdlog::debug("SystemRecoveryHandler: Failed to save user");
return false;
}
const auto it = std::find(old_users.begin(), old_users.end(), user.username());
if (it != old_users.end()) old_users.erase(it);
}
// Delete all the leftover users
for (const auto &user : old_users) {
if (!locked_auth.RemoveUser(user)) {
spdlog::debug("SystemRecoveryHandler: Failed to remove user \"{}\".", user);
return false;
}
}
// Roles are only supported with a license
if (license::global_license_checker.IsEnterpriseValidFast()) {
// Get all current roles
auto old_roles = locked_auth.AllRolenames();
// Save incoming users
for (const auto &role : roles) {
// Missing users
try {
locked_auth.SaveRole(role);
} catch (const auth::AuthException &) {
spdlog::debug("SystemRecoveryHandler: Failed to save user");
return false;
}
const auto it = std::find(old_roles.begin(), old_roles.end(), role.rolename());
if (it != old_roles.end()) old_roles.erase(it);
}
// Delete all the leftover users
for (const auto &role : old_roles) {
if (!locked_auth.RemoveRole(role)) {
spdlog::debug("SystemRecoveryHandler: Failed to remove user \"{}\".", role);
return false;
}
}
}
// Success
return true;
});
}
void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
auth::SynchedAuth &auth) {
// NOTE: Register even without license as the user could add a license at run-time
data.server->rpc_server_.Register<replication::UpdateAuthDataRpc>(
[&data, system_state_access, &auth](auto *req_reader, auto *res_builder) mutable {
spdlog::debug("Received UpdateAuthDataRpc");
UpdateAuthDataHandler(system_state_access, data.uuid_, auth, req_reader, res_builder);
});
data.server->rpc_server_.Register<replication::DropAuthDataRpc>(
[&data, system_state_access, &auth](auto *req_reader, auto *res_builder) mutable {
spdlog::debug("Received DropAuthDataRpc");
DropAuthDataHandler(system_state_access, data.uuid_, auth, req_reader, res_builder);
});
}
#endif
} // namespace memgraph::auth

View File

@@ -0,0 +1,37 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#include "auth/auth.hpp"
#include "replication/state.hpp"
#include "slk/streams.hpp"
#include "system/state.hpp"
namespace memgraph::auth {
void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
std::string_view rpc_req);
#ifdef MG_ENTERPRISE
void UpdateAuthDataHandler(system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, auth::SynchedAuth &auth,
slk::Reader *req_reader, slk::Builder *res_builder);
void DropAuthDataHandler(system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, auth::SynchedAuth &auth,
slk::Reader *req_reader, slk::Builder *res_builder);
bool SystemRecoveryHandler(auth::SynchedAuth &auth, auth::Auth::Config auth_config,
const std::vector<auth::User> &users, const std::vector<auth::Role> &roles);
void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
auth::SynchedAuth &auth);
#endif
} // namespace memgraph::auth

182
src/auth/rpc.cpp Normal file
View File

@@ -0,0 +1,182 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "auth/rpc.hpp"
#include <json/json.hpp>
#include "auth/auth.hpp"
#include "slk/serialization.hpp"
#include "slk/streams.hpp"
#include "utils/enum.hpp"
namespace memgraph::slk {
// Serialize code for auth::Role
void Save(const auth::Role &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.Serialize().dump(), builder);
}
namespace {
auth::Role LoadAuthRole(memgraph::slk::Reader *reader) {
std::string tmp;
memgraph::slk::Load(&tmp, reader);
const auto json = nlohmann::json::parse(tmp);
return memgraph::auth::Role::Deserialize(json);
}
} // namespace
// Deserialize code for auth::Role
void Load(auth::Role *self, memgraph::slk::Reader *reader) { *self = LoadAuthRole(reader); }
// Special case for optional<Role>
template <>
inline void Load<auth::Role>(std::optional<auth::Role> *obj, Reader *reader) {
bool exists = false;
Load(&exists, reader);
if (exists) {
obj->emplace(LoadAuthRole(reader));
} else {
*obj = std::nullopt;
}
}
// Serialize code for auth::User
void Save(const auth::User &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.Serialize().dump(), builder);
std::optional<auth::Role> role{};
if (const auto *role_ptr = self.role(); role_ptr) {
role.emplace(*role_ptr);
}
memgraph::slk::Save(role, builder);
}
// Deserialize code for auth::User
void Load(auth::User *self, memgraph::slk::Reader *reader) {
std::string tmp;
memgraph::slk::Load(&tmp, reader);
const auto json = nlohmann::json::parse(tmp);
*self = memgraph::auth::User::Deserialize(json);
std::optional<auth::Role> role{};
memgraph::slk::Load(&role, reader);
if (role)
self->SetRole(*role);
else
self->ClearRole();
}
// Serialize code for auth::Auth::Config
void Save(const auth::Auth::Config &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.name_regex_str, builder);
memgraph::slk::Save(self.password_regex_str, builder);
memgraph::slk::Save(self.password_permit_null, builder);
}
// Deserialize code for auth::Auth::Config
void Load(auth::Auth::Config *self, memgraph::slk::Reader *reader) {
std::string name_regex_str{};
std::string password_regex_str{};
bool password_permit_null{};
memgraph::slk::Load(&name_regex_str, reader);
memgraph::slk::Load(&password_regex_str, reader);
memgraph::slk::Load(&password_permit_null, reader);
*self = auth::Auth::Config{std::move(name_regex_str), std::move(password_regex_str), password_permit_null};
}
// Serialize code for UpdateAuthDataReq
void Save(const memgraph::replication::UpdateAuthDataReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.main_uuid, builder);
memgraph::slk::Save(self.epoch_id, builder);
memgraph::slk::Save(self.expected_group_timestamp, builder);
memgraph::slk::Save(self.new_group_timestamp, builder);
memgraph::slk::Save(self.user, builder);
memgraph::slk::Save(self.role, builder);
}
void Load(memgraph::replication::UpdateAuthDataReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->main_uuid, reader);
memgraph::slk::Load(&self->epoch_id, reader);
memgraph::slk::Load(&self->expected_group_timestamp, reader);
memgraph::slk::Load(&self->new_group_timestamp, reader);
memgraph::slk::Load(&self->user, reader);
memgraph::slk::Load(&self->role, reader);
}
// Serialize code for UpdateAuthDataRes
void Save(const memgraph::replication::UpdateAuthDataRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
void Load(memgraph::replication::UpdateAuthDataRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->success, reader);
}
// Serialize code for DropAuthDataReq
void Save(const memgraph::replication::DropAuthDataReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.main_uuid, builder);
memgraph::slk::Save(self.epoch_id, builder);
memgraph::slk::Save(self.expected_group_timestamp, builder);
memgraph::slk::Save(self.new_group_timestamp, builder);
memgraph::slk::Save(utils::EnumToNum<2, uint8_t>(self.type), builder);
memgraph::slk::Save(self.name, builder);
}
void Load(memgraph::replication::DropAuthDataReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->main_uuid, reader);
memgraph::slk::Load(&self->epoch_id, reader);
memgraph::slk::Load(&self->expected_group_timestamp, reader);
memgraph::slk::Load(&self->new_group_timestamp, reader);
uint8_t type_tmp = 0;
memgraph::slk::Load(&type_tmp, reader);
if (!utils::NumToEnum<2>(type_tmp, self->type)) {
throw SlkReaderException("Unexpected result line:{}!", __LINE__);
}
memgraph::slk::Load(&self->name, reader);
}
// Serialize code for DropAuthDataRes
void Save(const memgraph::replication::DropAuthDataRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
void Load(memgraph::replication::DropAuthDataRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->success, reader);
}
} // namespace memgraph::slk
namespace memgraph::replication {
constexpr utils::TypeInfo UpdateAuthDataReq::kType{utils::TypeId::REP_UPDATE_AUTH_DATA_REQ, "UpdateAuthDataReq",
nullptr};
constexpr utils::TypeInfo UpdateAuthDataRes::kType{utils::TypeId::REP_UPDATE_AUTH_DATA_RES, "UpdateAuthDataRes",
nullptr};
constexpr utils::TypeInfo DropAuthDataReq::kType{utils::TypeId::REP_DROP_AUTH_DATA_REQ, "DropAuthDataReq", nullptr};
constexpr utils::TypeInfo DropAuthDataRes::kType{utils::TypeId::REP_DROP_AUTH_DATA_RES, "DropAuthDataRes", nullptr};
void UpdateAuthDataReq::Save(const UpdateAuthDataReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void UpdateAuthDataReq::Load(UpdateAuthDataReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void UpdateAuthDataRes::Save(const UpdateAuthDataRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void UpdateAuthDataRes::Load(UpdateAuthDataRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void DropAuthDataReq::Save(const DropAuthDataReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void DropAuthDataReq::Load(DropAuthDataReq *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
void DropAuthDataRes::Save(const DropAuthDataRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void DropAuthDataRes::Load(DropAuthDataRes *self, memgraph::slk::Reader *reader) { memgraph::slk::Load(self, reader); }
} // namespace memgraph::replication

127
src/auth/rpc.hpp Normal file
View File

@@ -0,0 +1,127 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#include <optional>
#include "auth/auth.hpp"
#include "auth/models.hpp"
#include "rpc/messages.hpp"
#include "slk/streams.hpp"
namespace memgraph::replication {
struct UpdateAuthDataReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(UpdateAuthDataReq *self, memgraph::slk::Reader *reader);
static void Save(const UpdateAuthDataReq &self, memgraph::slk::Builder *builder);
UpdateAuthDataReq() = default;
UpdateAuthDataReq(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_ts, uint64_t new_ts,
auth::User user)
: main_uuid(main_uuid),
epoch_id{std::move(epoch_id)},
expected_group_timestamp{expected_ts},
new_group_timestamp{new_ts},
user{std::move(user)} {}
UpdateAuthDataReq(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_ts, uint64_t new_ts,
auth::Role role)
: main_uuid(main_uuid),
epoch_id{std::move(epoch_id)},
expected_group_timestamp{expected_ts},
new_group_timestamp{new_ts},
role{std::move(role)} {}
utils::UUID main_uuid;
std::string epoch_id;
uint64_t expected_group_timestamp;
uint64_t new_group_timestamp;
std::optional<auth::User> user;
std::optional<auth::Role> role;
};
struct UpdateAuthDataRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(UpdateAuthDataRes *self, memgraph::slk::Reader *reader);
static void Save(const UpdateAuthDataRes &self, memgraph::slk::Builder *builder);
UpdateAuthDataRes() = default;
explicit UpdateAuthDataRes(bool success) : success{success} {}
bool success;
};
using UpdateAuthDataRpc = rpc::RequestResponse<UpdateAuthDataReq, UpdateAuthDataRes>;
struct DropAuthDataReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(DropAuthDataReq *self, memgraph::slk::Reader *reader);
static void Save(const DropAuthDataReq &self, memgraph::slk::Builder *builder);
DropAuthDataReq() = default;
enum class DataType { USER, ROLE };
DropAuthDataReq(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_ts, uint64_t new_ts,
DataType type, std::string_view name)
: main_uuid(main_uuid),
epoch_id{std::move(epoch_id)},
expected_group_timestamp{expected_ts},
new_group_timestamp{new_ts},
type{type},
name{name} {}
utils::UUID main_uuid;
std::string epoch_id;
uint64_t expected_group_timestamp;
uint64_t new_group_timestamp;
DataType type;
std::string name;
};
struct DropAuthDataRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(DropAuthDataRes *self, memgraph::slk::Reader *reader);
static void Save(const DropAuthDataRes &self, memgraph::slk::Builder *builder);
DropAuthDataRes() = default;
explicit DropAuthDataRes(bool success) : success{success} {}
bool success;
};
using DropAuthDataRpc = rpc::RequestResponse<DropAuthDataReq, DropAuthDataRes>;
} // namespace memgraph::replication
namespace memgraph::slk {
void Save(const auth::Role &self, memgraph::slk::Builder *builder);
void Load(auth::Role *self, memgraph::slk::Reader *reader);
void Save(const auth::User &self, memgraph::slk::Builder *builder);
void Load(auth::User *self, memgraph::slk::Reader *reader);
void Save(const auth::Auth::Config &self, memgraph::slk::Builder *builder);
void Load(auth::Auth::Config *self, memgraph::slk::Reader *reader);
void Save(const memgraph::replication::UpdateAuthDataRes &self, memgraph::slk::Builder *builder);
void Load(memgraph::replication::UpdateAuthDataRes *self, memgraph::slk::Reader *reader);
void Save(const memgraph::replication::UpdateAuthDataReq & /*self*/, memgraph::slk::Builder * /*builder*/);
void Load(memgraph::replication::UpdateAuthDataReq * /*self*/, memgraph::slk::Reader * /*reader*/);
void Save(const memgraph::replication::DropAuthDataRes &self, memgraph::slk::Builder *builder);
void Load(memgraph::replication::DropAuthDataRes *self, memgraph::slk::Reader *reader);
void Save(const memgraph::replication::DropAuthDataReq & /*self*/, memgraph::slk::Builder * /*builder*/);
void Load(memgraph::replication::DropAuthDataReq * /*self*/, memgraph::slk::Reader * /*reader*/);
} // namespace memgraph::slk

View File

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

View File

@@ -7,19 +7,36 @@ target_sources(mg-coordination
include/coordination/coordinator_rpc.hpp
include/coordination/coordinator_server.hpp
include/coordination/coordinator_config.hpp
include/coordination/coordinator_entity_info.hpp
include/coordination/coordinator_exceptions.hpp
include/coordination/coordinator_slk.hpp
include/coordination/coordinator_instance.hpp
include/coordination/coordinator_cluster_config.hpp
include/coordination/coordinator_handlers.hpp
include/coordination/constants.hpp
include/coordination/instance_status.hpp
include/coordination/replication_instance.hpp
include/coordination/raft_instance.hpp
include/nuraft/coordinator_log_store.hpp
include/nuraft/coordinator_state_machine.hpp
include/nuraft/coordinator_state_manager.hpp
PRIVATE
coordinator_client.cpp
coordinator_state.cpp
coordinator_rpc.cpp
coordinator_server.cpp
coordinator_handlers.cpp
coordinator_instance.cpp
replication_instance.cpp
raft_instance.cpp
coordinator_log_store.cpp
coordinator_state_machine.cpp
coordinator_state_manager.cpp
)
target_include_directories(mg-coordination PUBLIC include)
target_link_libraries(mg-coordination
PUBLIC mg::utils mg::rpc mg::slk mg::io mg::repl_coord_glue
PUBLIC mg::utils mg::rpc mg::slk mg::io mg::repl_coord_glue lib::rangev3 nuraft mg-replication_handler
)

View File

@@ -9,6 +9,7 @@
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "utils/uuid.hpp"
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_client.hpp"
@@ -20,91 +21,102 @@
namespace memgraph::coordination {
namespace {
auto CreateClientContext(const memgraph::coordination::CoordinatorClientConfig &config)
auto CreateClientContext(memgraph::coordination::CoordinatorClientConfig const &config)
-> communication::ClientContext {
return (config.ssl) ? communication::ClientContext{config.ssl->key_file, config.ssl->cert_file}
: communication::ClientContext{};
}
} // namespace
CoordinatorClient::CoordinatorClient(const CoordinatorClientConfig &config)
CoordinatorClient::CoordinatorClient(CoordinatorInstance *coord_instance, CoordinatorClientConfig config,
HealthCheckCallback succ_cb, HealthCheckCallback fail_cb)
: rpc_context_{CreateClientContext(config)},
rpc_client_{io::network::Endpoint(io::network::Endpoint::needs_resolving, config.ip_address, config.port),
&rpc_context_},
config_{config} {}
config_{std::move(config)},
coord_instance_{coord_instance},
succ_cb_{std::move(succ_cb)},
fail_cb_{std::move(fail_cb)} {}
CoordinatorClient::~CoordinatorClient() {
auto exit_job = utils::OnScopeExit([&] {
StopFrequentCheck();
thread_pool_.Shutdown();
});
const auto endpoint = rpc_client_.Endpoint();
// Logging can throw
spdlog::trace("Closing replication client on {}:{}", endpoint.address, endpoint.port);
}
auto CoordinatorClient::InstanceName() const -> std::string { return config_.instance_name; }
auto CoordinatorClient::SocketAddress() const -> std::string { return rpc_client_.Endpoint().SocketAddress(); }
void CoordinatorClient::StartFrequentCheck() {
if (instance_checker_.IsRunning()) {
return;
}
MG_ASSERT(config_.health_check_frequency_sec > std::chrono::seconds(0),
"Health check frequency must be greater than 0");
replica_checker_.Run(
"Coord checker", config_.health_check_frequency_sec,
[last_response_time = &last_response_time_, rpc_client = &rpc_client_] {
instance_checker_.Run(
config_.instance_name, config_.health_check_frequency_sec, [this, instance_name = config_.instance_name] {
try {
{
auto stream{rpc_client->Stream<memgraph::replication_coordination_glue::FrequentHeartbeatRpc>()};
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();
last_response_time->store(std::chrono::system_clock::now(), std::memory_order_acq_rel);
}
} catch (const rpc::RpcFailedException &) {
// Nothing to do...wait for a reconnect
succ_cb_(coord_instance_, instance_name);
} catch (rpc::RpcFailedException const &) {
fail_cb_(coord_instance_, instance_name);
}
});
}
void CoordinatorClient::StopFrequentCheck() { replica_checker_.Stop(); }
void CoordinatorClient::StopFrequentCheck() { instance_checker_.Stop(); }
void CoordinatorClient::PauseFrequentCheck() { instance_checker_.Pause(); }
void CoordinatorClient::ResumeFrequentCheck() { instance_checker_.Resume(); }
bool CoordinatorClient::DoHealthCheck() const {
auto current_time = std::chrono::system_clock::now();
auto duration = std::chrono::duration_cast<std::chrono::seconds>(current_time -
last_response_time_.load(std::memory_order_acquire));
return duration.count() <= alive_response_time_difference_sec_;
auto CoordinatorClient::SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void {
succ_cb_ = std::move(succ_cb);
fail_cb_ = std::move(fail_cb);
}
auto CoordinatorClient::InstanceName() const -> std::string_view { return config_.instance_name; }
auto CoordinatorClient::Endpoint() const -> io::network::Endpoint const & { return rpc_client_.Endpoint(); }
auto CoordinatorClient::Config() const -> CoordinatorClientConfig const & { return config_; }
auto CoordinatorClient::ReplicationClientInfo() const -> ReplClientInfo { return config_.replication_client_info; }
auto CoordinatorClient::ReplicationClientInfo() const -> CoordinatorClientConfig::ReplicationClientInfo const & {
MG_ASSERT(config_.replication_client_info.has_value(), "No ReplicationClientInfo for MAIN instance!");
return *config_.replication_client_info;
}
////// AF design choice
auto CoordinatorClient::ReplicationClientInfo() -> std::optional<CoordinatorClientConfig::ReplicationClientInfo> & {
MG_ASSERT(config_.replication_client_info.has_value(), "No ReplicationClientInfo for MAIN instance!");
return config_.replication_client_info;
}
void CoordinatorClient::UpdateTimeCheck(const std::chrono::system_clock::time_point &last_checked_time) {
last_response_time_.store(last_checked_time, std::memory_order_acq_rel);
}
auto CoordinatorClient::GetLastTimeResponse() -> std::chrono::system_clock::time_point { return last_response_time_; }
auto CoordinatorClient::SendPromoteReplicaToMainRpc(
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info) const -> bool {
auto CoordinatorClient::SendPromoteReplicaToMainRpc(const utils::UUID &uuid,
ReplicationClientsInfo replication_clients_info) const -> bool {
try {
{
auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(std::move(replication_clients_info))};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to perform failover!");
return false;
}
spdlog::info("Sent failover RPC from coordinator to new main!");
return true;
auto stream{rpc_client_.Stream<PromoteReplicaToMainRpc>(uuid, std::move(replication_clients_info))};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful PromoteReplicaToMainRpc response!");
return false;
}
return true;
} catch (rpc::RpcFailedException const &) {
spdlog::error("RPC error occurred while sending PromoteReplicaToMainRpc!");
}
return false;
}
auto CoordinatorClient::DemoteToReplica() const -> bool {
auto const &instance_name = config_.instance_name;
try {
auto stream{rpc_client_.Stream<DemoteMainToReplicaRpc>(config_.replication_client_info)};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful RPC response for setting instance {} to replica!", instance_name);
return false;
}
spdlog::info("Sent request RPC from coordinator to instance to set it as replica!");
return true;
} catch (rpc::RpcFailedException const &) {
spdlog::error("Failed to set instance {} to replica!", instance_name);
}
return false;
}
auto CoordinatorClient::SendSwapMainUUIDRpc(const utils::UUID &uuid) const -> bool {
try {
auto stream{rpc_client_.Stream<replication_coordination_glue::SwapMainUUIDRpc>(uuid)};
if (!stream.AwaitResponse().success) {
spdlog::error("Failed to receive successful RPC swapping of uuid response!");
return false;
}
return true;
} catch (const rpc::RpcFailedException &) {
spdlog::error("Failed to send failover RPC from coordinator to new main!");
spdlog::error("RPC error occurred while sending swapping uuid RPC!");
}
return false;
}

View File

@@ -0,0 +1,150 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_handlers.hpp"
#include <range/v3/view.hpp>
#include "coordination/coordinator_rpc.hpp"
#include "coordination/include/coordination/coordinator_server.hpp"
#include "replication/state.hpp"
namespace memgraph::dbms {
void CoordinatorHandlers::Register(memgraph::coordination::CoordinatorServer &server,
replication::ReplicationHandler &replication_handler) {
server.Register<coordination::PromoteReplicaToMainRpc>(
[&](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
spdlog::info("Received PromoteReplicaToMainRpc");
CoordinatorHandlers::PromoteReplicaToMainHandler(replication_handler, req_reader, res_builder);
});
server.Register<coordination::DemoteMainToReplicaRpc>(
[&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
spdlog::info("Received DemoteMainToReplicaRpc from coordinator server");
CoordinatorHandlers::DemoteMainToReplicaHandler(replication_handler, req_reader, res_builder);
});
server.Register<replication_coordination_glue::SwapMainUUIDRpc>(
[&replication_handler](slk::Reader *req_reader, slk::Builder *res_builder) -> void {
spdlog::info("Received SwapMainUUIDRPC on coordinator server");
CoordinatorHandlers::SwapMainUUIDHandler(replication_handler, req_reader, res_builder);
});
}
void CoordinatorHandlers::SwapMainUUIDHandler(replication::ReplicationHandler &replication_handler,
slk::Reader *req_reader, slk::Builder *res_builder) {
if (!replication_handler.IsReplica()) {
spdlog::error("Setting main uuid must be performed on replica.");
slk::Save(replication_coordination_glue::SwapMainUUIDRes{false}, res_builder);
return;
}
replication_coordination_glue::SwapMainUUIDReq req;
slk::Load(&req, req_reader);
spdlog::info(fmt::format("Set replica data UUID to main uuid {}", std::string(req.uuid)));
std::get<memgraph::replication::RoleReplicaData>(replication_handler.GetReplState().ReplicationData()).uuid_ =
req.uuid;
slk::Save(replication_coordination_glue::SwapMainUUIDRes{true}, res_builder);
}
void CoordinatorHandlers::DemoteMainToReplicaHandler(replication::ReplicationHandler &replication_handler,
slk::Reader *req_reader, slk::Builder *res_builder) {
spdlog::info("Executing DemoteMainToReplicaHandler");
if (!replication_handler.IsMain()) {
spdlog::error("Setting to replica must be performed on main.");
slk::Save(coordination::DemoteMainToReplicaRes{false}, res_builder);
return;
}
coordination::DemoteMainToReplicaReq req;
slk::Load(&req, req_reader);
const replication::ReplicationServerConfig clients_config{
.ip_address = req.replication_client_info.replication_ip_address,
.port = req.replication_client_info.replication_port};
if (!replication_handler.SetReplicationRoleReplica(clients_config, std::nullopt)) {
spdlog::error("Demoting main to replica failed!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
}
slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
}
void CoordinatorHandlers::PromoteReplicaToMainHandler(replication::ReplicationHandler &replication_handler,
slk::Reader *req_reader, slk::Builder *res_builder) {
if (!replication_handler.IsReplica()) {
spdlog::error("Failover must be performed on replica!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
}
coordination::PromoteReplicaToMainReq req;
slk::Load(&req, req_reader);
// This can fail because of disk. If it does, the cluster state could get inconsistent.
// We don't handle disk issues.
if (const bool success = replication_handler.DoReplicaToMainPromotion(req.main_uuid_); !success) {
spdlog::error("Promoting replica to main failed!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
}
auto const converter = [](const auto &repl_info_config) {
return replication::ReplicationClientConfig{
.name = repl_info_config.instance_name,
.mode = repl_info_config.replication_mode,
.ip_address = repl_info_config.replication_ip_address,
.port = repl_info_config.replication_port,
};
};
// registering replicas
for (auto const &config : req.replication_clients_info | ranges::views::transform(converter)) {
auto instance_client = replication_handler.RegisterReplica(config, false);
if (instance_client.HasError()) {
using enum memgraph::replication::RegisterReplicaError;
switch (instance_client.GetError()) {
// Can't happen, checked on the coordinator side
case memgraph::query::RegisterReplicaError::NAME_EXISTS:
spdlog::error("Replica with the same name already exists!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
// Can't happen, checked on the coordinator side
case memgraph::query::RegisterReplicaError::ENDPOINT_EXISTS:
spdlog::error("Replica with the same endpoint already exists!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
// We don't handle disk issues
case memgraph::query::RegisterReplicaError::COULD_NOT_BE_PERSISTED:
spdlog::error("Registered replica could not be persisted!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
case memgraph::query::RegisterReplicaError::ERROR_ACCEPTING_MAIN:
spdlog::error("Replica didn't accept change of main!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
case memgraph::query::RegisterReplicaError::CONNECTION_FAILED:
// Connection failure is not a fatal error
break;
}
}
}
spdlog::error(fmt::format("FICO : Promote replica to main was success {}", std::string(req.main_uuid_)));
slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
}
} // namespace memgraph::dbms
#endif

View File

@@ -0,0 +1,289 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_instance.hpp"
#include "coordination/coordinator_exceptions.hpp"
#include "nuraft/coordinator_state_machine.hpp"
#include "nuraft/coordinator_state_manager.hpp"
#include "utils/counter.hpp"
#include <range/v3/view.hpp>
#include <shared_mutex>
namespace memgraph::coordination {
using nuraft::ptr;
using nuraft::srv_config;
CoordinatorInstance::CoordinatorInstance()
: self_([this] { std::ranges::for_each(repl_instances_, &ReplicationInstance::StartFrequentCheck); },
[this] { std::ranges::for_each(repl_instances_, &ReplicationInstance::StopFrequentCheck); }) {
auto find_instance = [](CoordinatorInstance *coord_instance,
std::string_view instance_name) -> ReplicationInstance & {
auto instance = std::ranges::find_if(
coord_instance->repl_instances_,
[instance_name](ReplicationInstance const &instance) { return instance.InstanceName() == instance_name; });
MG_ASSERT(instance != coord_instance->repl_instances_.end(), "Instance {} not found during callback!",
instance_name);
return *instance;
};
replica_succ_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
spdlog::trace("Instance {} performing replica successful callback", instance_name);
find_instance(coord_instance, instance_name).OnSuccessPing();
};
replica_fail_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
spdlog::trace("Instance {} performing replica failure callback", instance_name);
find_instance(coord_instance, instance_name).OnFailPing();
};
main_succ_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
spdlog::trace("Instance {} performing main successful callback", instance_name);
auto &instance = find_instance(coord_instance, instance_name);
if (instance.IsAlive()) {
instance.OnSuccessPing();
return;
}
bool const is_latest_main = !coord_instance->ClusterHasAliveMain_();
if (is_latest_main) {
spdlog::info("Instance {} is the latest main", instance_name);
instance.OnSuccessPing();
return;
}
bool const demoted = instance.DemoteToReplica(coord_instance->replica_succ_cb_, coord_instance->replica_fail_cb_);
if (demoted) {
instance.OnSuccessPing();
spdlog::info("Instance {} demoted to replica", instance_name);
} else {
spdlog::error("Instance {} failed to become replica", instance_name);
}
};
main_fail_cb_ = [find_instance](CoordinatorInstance *coord_instance, std::string_view instance_name) -> void {
auto lock = std::lock_guard{coord_instance->coord_instance_lock_};
spdlog::trace("Instance {} performing main failure callback", instance_name);
find_instance(coord_instance, instance_name).OnFailPing();
if (!coord_instance->ClusterHasAliveMain_()) {
spdlog::info("Cluster without main instance, trying automatic failover");
coord_instance->TryFailover();
}
};
}
auto CoordinatorInstance::ClusterHasAliveMain_() const -> bool {
auto const alive_main = [](ReplicationInstance const &instance) { return instance.IsMain() && instance.IsAlive(); };
return std::ranges::any_of(repl_instances_, alive_main);
}
auto CoordinatorInstance::TryFailover() -> void {
auto alive_replicas = repl_instances_ | ranges::views::filter(&ReplicationInstance::IsReplica) |
ranges::views::filter(&ReplicationInstance::IsAlive);
if (ranges::empty(alive_replicas)) {
spdlog::warn("Failover failed since all replicas are down!");
return;
}
// TODO: Smarter choice
auto chosen_replica_instance = ranges::begin(alive_replicas);
chosen_replica_instance->PauseFrequentCheck();
utils::OnScopeExit scope_exit{[&chosen_replica_instance] { chosen_replica_instance->ResumeFrequentCheck(); }};
auto const potential_new_main_uuid = utils::UUID{};
auto const is_not_chosen_replica_instance = [&chosen_replica_instance](ReplicationInstance &instance) {
return instance != *chosen_replica_instance;
};
// If for some replicas swap fails, for others on successful ping we will revert back on next change
// or we will do failover first again and then it will be consistent again
for (auto &other_replica_instance : alive_replicas | ranges::views::filter(is_not_chosen_replica_instance)) {
if (!other_replica_instance.SendSwapAndUpdateUUID(potential_new_main_uuid)) {
spdlog::error(fmt::format("Failed to swap uuid for instance {} which is alive, aborting failover",
other_replica_instance.InstanceName()));
return;
}
}
std::vector<ReplClientInfo> repl_clients_info;
repl_clients_info.reserve(repl_instances_.size() - 1);
std::ranges::transform(repl_instances_ | ranges::views::filter(is_not_chosen_replica_instance),
std::back_inserter(repl_clients_info), &ReplicationInstance::ReplicationClientInfo);
if (!chosen_replica_instance->PromoteToMain(potential_new_main_uuid, std::move(repl_clients_info), main_succ_cb_,
main_fail_cb_)) {
spdlog::warn("Failover failed since promoting replica to main failed!");
return;
}
chosen_replica_instance->SetNewMainUUID(potential_new_main_uuid);
main_uuid_ = potential_new_main_uuid;
spdlog::info("Failover successful! Instance {} promoted to main.", chosen_replica_instance->InstanceName());
}
auto CoordinatorInstance::ShowInstances() const -> std::vector<InstanceStatus> {
auto const coord_instances = self_.GetAllCoordinators();
std::vector<InstanceStatus> instances_status;
instances_status.reserve(repl_instances_.size() + coord_instances.size());
auto const stringify_repl_role = [](ReplicationInstance const &instance) -> std::string {
if (!instance.IsAlive()) return "unknown";
if (instance.IsMain()) return "main";
return "replica";
};
auto const repl_instance_to_status = [&stringify_repl_role](ReplicationInstance const &instance) -> InstanceStatus {
return {.instance_name = instance.InstanceName(),
.coord_socket_address = instance.SocketAddress(),
.cluster_role = stringify_repl_role(instance),
.is_alive = instance.IsAlive()};
};
auto const coord_instance_to_status = [](ptr<srv_config> const &instance) -> InstanceStatus {
return {.instance_name = "coordinator_" + std::to_string(instance->get_id()),
.raft_socket_address = instance->get_endpoint(),
.cluster_role = "coordinator",
.is_alive = true}; // TODO: (andi) Get this info from RAFT and test it or when we will move
// CoordinatorState to every instance, we can be smarter about this using our RPC.
};
std::ranges::transform(coord_instances, std::back_inserter(instances_status), coord_instance_to_status);
{
auto lock = std::shared_lock{coord_instance_lock_};
std::ranges::transform(repl_instances_, std::back_inserter(instances_status), repl_instance_to_status);
}
return instances_status;
}
// TODO: (andi) Make sure you cannot put coordinator instance to the main
auto CoordinatorInstance::SetReplicationInstanceToMain(std::string instance_name)
-> SetInstanceToMainCoordinatorStatus {
auto lock = std::lock_guard{coord_instance_lock_};
auto const is_new_main = [&instance_name](ReplicationInstance const &instance) {
return instance.InstanceName() == instance_name;
};
auto new_main = std::ranges::find_if(repl_instances_, is_new_main);
if (new_main == repl_instances_.end()) {
spdlog::error("Instance {} not registered. Please register it using REGISTER INSTANCE {}", instance_name,
instance_name);
return SetInstanceToMainCoordinatorStatus::NO_INSTANCE_WITH_NAME;
}
new_main->PauseFrequentCheck();
utils::OnScopeExit scope_exit{[&new_main] { new_main->ResumeFrequentCheck(); }};
ReplicationClientsInfo repl_clients_info;
repl_clients_info.reserve(repl_instances_.size() - 1);
auto const is_not_new_main = [&instance_name](ReplicationInstance const &instance) {
return instance.InstanceName() != instance_name;
};
auto potential_new_main_uuid = utils::UUID{};
spdlog::trace("Generated potential new main uuid");
for (auto &other_instance : repl_instances_ | ranges::views::filter(is_not_new_main)) {
if (!other_instance.SendSwapAndUpdateUUID(potential_new_main_uuid)) {
spdlog::error(
fmt::format("Failed to swap uuid for instance {}, aborting failover", other_instance.InstanceName()));
return SetInstanceToMainCoordinatorStatus::SWAP_UUID_FAILED;
}
}
std::ranges::transform(repl_instances_ | ranges::views::filter(is_not_new_main),
std::back_inserter(repl_clients_info),
[](const ReplicationInstance &instance) { return instance.ReplicationClientInfo(); });
if (!new_main->PromoteToMain(potential_new_main_uuid, std::move(repl_clients_info), main_succ_cb_, main_fail_cb_)) {
return SetInstanceToMainCoordinatorStatus::COULD_NOT_PROMOTE_TO_MAIN;
}
new_main->SetNewMainUUID(potential_new_main_uuid);
main_uuid_ = potential_new_main_uuid;
spdlog::info("Instance {} promoted to main", instance_name);
return SetInstanceToMainCoordinatorStatus::SUCCESS;
}
auto CoordinatorInstance::RegisterReplicationInstance(CoordinatorClientConfig config)
-> RegisterInstanceCoordinatorStatus {
auto lock = std::lock_guard{coord_instance_lock_};
auto const name_matches = [&config](ReplicationInstance const &instance) {
return instance.InstanceName() == config.instance_name;
};
if (std::ranges::any_of(repl_instances_, name_matches)) {
return RegisterInstanceCoordinatorStatus::NAME_EXISTS;
}
auto const socket_address_matches = [&config](ReplicationInstance const &instance) {
return instance.SocketAddress() == config.SocketAddress();
};
if (std::ranges::any_of(repl_instances_, socket_address_matches)) {
return RegisterInstanceCoordinatorStatus::ENDPOINT_EXISTS;
}
if (!self_.RequestLeadership()) {
return RegisterInstanceCoordinatorStatus::NOT_LEADER;
}
auto const res = self_.AppendRegisterReplicationInstance(config.instance_name);
if (!res->get_accepted()) {
spdlog::error(
"Failed to accept request for registering instance {}. Most likely the reason is that the instance is not the "
"leader.",
config.instance_name);
return RegisterInstanceCoordinatorStatus::RAFT_COULD_NOT_ACCEPT;
}
spdlog::info("Request for registering instance {} accepted", config.instance_name);
try {
repl_instances_.emplace_back(this, std::move(config), replica_succ_cb_, replica_fail_cb_);
} catch (CoordinatorRegisterInstanceException const &) {
return RegisterInstanceCoordinatorStatus::RPC_FAILED;
}
if (res->get_result_code() != nuraft::cmd_result_code::OK) {
spdlog::error("Failed to register instance {} with error code {}", config.instance_name, res->get_result_code());
return RegisterInstanceCoordinatorStatus::RAFT_COULD_NOT_APPEND;
}
spdlog::info("Instance {} registered", config.instance_name);
return RegisterInstanceCoordinatorStatus::SUCCESS;
}
auto CoordinatorInstance::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
-> void {
self_.AddCoordinatorInstance(raft_server_id, raft_port, std::move(raft_address));
}
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,201 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "nuraft/coordinator_log_store.hpp"
#include "coordination/coordinator_exceptions.hpp"
namespace memgraph::coordination {
using nuraft::cs_new;
namespace {
ptr<log_entry> MakeClone(const ptr<log_entry> &entry) {
return cs_new<log_entry>(entry->get_term(), buffer::clone(entry->get_buf()), entry->get_val_type(),
entry->get_timestamp());
}
} // namespace
CoordinatorLogStore::CoordinatorLogStore() : start_idx_(1) {
ptr<buffer> buf = buffer::alloc(sizeof(uint64_t));
logs_[0] = cs_new<log_entry>(0, buf);
}
CoordinatorLogStore::~CoordinatorLogStore() {}
auto CoordinatorLogStore::FindOrDefault_(uint64_t index) const -> ptr<log_entry> {
auto entry = logs_.find(index);
if (entry == logs_.end()) {
entry = logs_.find(0);
}
return entry->second;
}
uint64_t CoordinatorLogStore::next_slot() const {
auto lock = std::lock_guard{logs_lock_};
return start_idx_ + logs_.size() - 1;
}
uint64_t CoordinatorLogStore::start_index() const { return start_idx_; }
ptr<log_entry> CoordinatorLogStore::last_entry() const {
auto lock = std::lock_guard{logs_lock_};
uint64_t const last_idx = start_idx_ + logs_.size() - 1;
auto const last_src = FindOrDefault_(last_idx - 1);
return MakeClone(last_src);
}
uint64_t CoordinatorLogStore::append(ptr<log_entry> &entry) {
ptr<log_entry> clone = MakeClone(entry);
uint64_t next_slot{0};
{
auto lock = std::lock_guard{logs_lock_};
next_slot = start_idx_ + logs_.size() - 1;
logs_[next_slot] = clone;
}
return next_slot;
}
void CoordinatorLogStore::write_at(uint64_t index, ptr<log_entry> &entry) {
ptr<log_entry> clone = MakeClone(entry);
// Discard all logs equal to or greater than `index.
{
auto lock = std::lock_guard{logs_lock_};
auto itr = logs_.lower_bound(index);
while (itr != logs_.end()) {
itr = logs_.erase(itr);
}
logs_[index] = clone;
}
}
ptr<std::vector<ptr<log_entry>>> CoordinatorLogStore::log_entries(uint64_t start, uint64_t end) {
auto ret = cs_new<std::vector<ptr<log_entry>>>();
ret->resize(end - start);
for (uint64_t i = start, curr_index = 0; i < end; ++i, ++curr_index) {
ptr<log_entry> src = nullptr;
{
auto lock = std::lock_guard{logs_lock_};
if (auto const entry = logs_.find(i); entry != logs_.end()) {
src = entry->second;
} else {
throw RaftCouldNotFindEntryException("Could not find entry at index {}", i);
}
}
(*ret)[curr_index] = MakeClone(src);
}
return ret;
}
ptr<log_entry> CoordinatorLogStore::entry_at(uint64_t index) {
ptr<log_entry> src = nullptr;
{
auto lock = std::lock_guard{logs_lock_};
src = FindOrDefault_(index);
}
return MakeClone(src);
}
uint64_t CoordinatorLogStore::term_at(uint64_t index) {
uint64_t term = 0;
{
auto lock = std::lock_guard{logs_lock_};
term = FindOrDefault_(index)->get_term();
}
return term;
}
ptr<buffer> CoordinatorLogStore::pack(uint64_t index, int32 cnt) {
std::vector<ptr<buffer>> logs;
size_t size_total = 0;
uint64_t const end_index = index + cnt;
for (uint64_t i = index; i < end_index; ++i) {
ptr<log_entry> le = nullptr;
{
auto lock = std::lock_guard{logs_lock_};
le = logs_[i];
}
assert(le.get());
auto buf = le->serialize();
size_total += buf->size();
logs.push_back(buf);
}
auto buf_out = buffer::alloc(sizeof(int32) + cnt * sizeof(int32) + size_total);
buf_out->pos(0);
buf_out->put((int32)cnt);
for (auto &entry : logs) {
auto &bb = entry; // TODO: (andi) This smells like not needed
buf_out->put(static_cast<int32>(bb->size()));
buf_out->put(*bb);
}
return buf_out;
}
void CoordinatorLogStore::apply_pack(uint64_t index, buffer &pack) {
pack.pos(0);
int32 const num_logs = pack.get_int();
for (int32 i = 0; i < num_logs; ++i) {
uint64_t cur_idx = index + i;
int32 buf_size = pack.get_int();
ptr<buffer> buf_local = buffer::alloc(buf_size);
pack.get(buf_local);
ptr<log_entry> le = log_entry::deserialize(*buf_local);
{
auto lock = std::lock_guard{logs_lock_};
logs_[cur_idx] = le;
}
}
{
auto lock = std::lock_guard{logs_lock_};
auto const entry = logs_.upper_bound(0);
if (entry != logs_.end()) {
start_idx_ = entry->first;
} else {
start_idx_ = 1;
}
}
}
// NOTE: Remove all logs up to given 'last_log_index' (inclusive).
bool CoordinatorLogStore::compact(uint64_t last_log_index) {
auto lock = std::lock_guard{logs_lock_};
for (uint64_t ii = start_idx_; ii <= last_log_index; ++ii) {
auto const entry = logs_.find(ii);
if (entry != logs_.end()) {
logs_.erase(entry);
}
}
if (start_idx_ <= last_log_index) {
start_idx_ = last_log_index + 1;
}
return true;
}
bool CoordinatorLogStore::flush() { return true; }
} // namespace memgraph::coordination
#endif

View File

@@ -36,6 +36,22 @@ void PromoteReplicaToMainRes::Load(PromoteReplicaToMainRes *self, memgraph::slk:
memgraph::slk::Load(self, reader);
}
void DemoteMainToReplicaReq::Save(const DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void DemoteMainToReplicaReq::Load(DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
void DemoteMainToReplicaRes::Save(const DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self, builder);
}
void DemoteMainToReplicaRes::Load(DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(self, reader);
}
} // namespace coordination
constexpr utils::TypeInfo coordination::PromoteReplicaToMainReq::kType{utils::TypeId::COORD_FAILOVER_REQ,
@@ -44,6 +60,12 @@ 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::DemoteMainToReplicaRes::kType{utils::TypeId::COORD_SET_REPL_MAIN_RES,
"CoordDemoteToReplicaRes", nullptr};
namespace slk {
void Save(const memgraph::coordination::PromoteReplicaToMainRes &self, memgraph::slk::Builder *builder) {
@@ -55,13 +77,31 @@ void Load(memgraph::coordination::PromoteReplicaToMainRes *self, memgraph::slk::
}
void Save(const memgraph::coordination::PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.main_uuid_, builder);
memgraph::slk::Save(self.replication_clients_info, builder);
}
void Load(memgraph::coordination::PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->main_uuid_, reader);
memgraph::slk::Load(&self->replication_clients_info, reader);
}
void Save(const memgraph::coordination::DemoteMainToReplicaReq &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.replication_client_info, builder);
}
void Load(memgraph::coordination::DemoteMainToReplicaReq *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->replication_client_info, reader);
}
void Save(const memgraph::coordination::DemoteMainToReplicaRes &self, memgraph::slk::Builder *builder) {
memgraph::slk::Save(self.success, builder);
}
void Load(memgraph::coordination::DemoteMainToReplicaRes *self, memgraph::slk::Reader *reader) {
memgraph::slk::Load(&self->success, reader);
}
} // namespace slk
} // namespace memgraph

View File

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

View File

@@ -9,207 +9,71 @@
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "coordination/coordinator_state.hpp"
#include <span>
#include "coordination/coordinator_client.hpp"
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_state.hpp"
#include "coordination/coordinator_config.hpp"
#include "coordination/coordinator_entity_info.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
#include "flags/replication.hpp"
#include "spdlog/spdlog.h"
#include "utils/logging.hpp"
#include "utils/variant_helpers.hpp"
#include <atomic>
#include <exception>
#include <optional>
#include <algorithm>
namespace memgraph::coordination {
namespace {
bool CheckName(const std::list<CoordinatorClient> &replicas, const CoordinatorClientConfig &config) {
auto name_matches = [&instance_name = config.instance_name](auto const &replica) {
return replica.InstanceName() == instance_name;
};
return std::any_of(replicas.begin(), replicas.end(), name_matches);
};
} // namespace
CoordinatorState::CoordinatorState() {
MG_ASSERT(!(FLAGS_coordinator && FLAGS_coordinator_server_port),
MG_ASSERT(!(FLAGS_raft_server_id && FLAGS_coordinator_server_port),
"Instance cannot be a coordinator and have registered coordinator server.");
spdlog::info("Executing coordinator constructor");
if (FLAGS_coordinator_server_port) {
spdlog::info("Coordinator server port set");
auto const config = CoordinatorServerConfig{
.ip_address = kDefaultReplicationServerIp,
.port = static_cast<uint16_t>(FLAGS_coordinator_server_port),
};
spdlog::info("Executing coordinator constructor main replica");
data_ = CoordinatorMainReplicaData{.coordinator_server_ = std::make_unique<CoordinatorServer>(config)};
}
}
auto CoordinatorState::RegisterReplica(const CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus, CoordinatorClient *> {
const auto name_endpoint_status =
std::visit(memgraph::utils::Overloaded{[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
return RegisterMainReplicaCoordinatorStatus::NOT_COORDINATOR;
},
[&config](const CoordinatorData &coordinator_data) {
if (memgraph::coordination::CheckName(
coordinator_data.registered_replicas_, config)) {
return RegisterMainReplicaCoordinatorStatus::NAME_EXISTS;
}
return RegisterMainReplicaCoordinatorStatus::SUCCESS;
}},
data_);
auto CoordinatorState::RegisterReplicationInstance(CoordinatorClientConfig config)
-> RegisterInstanceCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
"Coordinator cannot register replica since variant holds wrong alternative");
if (name_endpoint_status != RegisterMainReplicaCoordinatorStatus::SUCCESS) {
return name_endpoint_status;
}
// Maybe no need to return client if you can start replica client here
return &std::get<CoordinatorData>(data_).registered_replicas_.emplace_back(config);
}
auto CoordinatorState::RegisterMain(const CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus, CoordinatorClient *> {
const auto endpoint_status = std::visit(
memgraph::utils::Overloaded{
[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
return RegisterMainReplicaCoordinatorStatus::NOT_COORDINATOR;
},
[](const CoordinatorData & /*coordinator_data*/) { return RegisterMainReplicaCoordinatorStatus::SUCCESS; }},
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);
}},
data_);
if (endpoint_status != RegisterMainReplicaCoordinatorStatus::SUCCESS) {
return endpoint_status;
}
auto &registered_main = std::get<CoordinatorData>(data_).registered_main_;
registered_main = std::make_unique<CoordinatorClient>(config);
return registered_main.get();
}
auto CoordinatorState::ShowReplicas() const -> std::vector<CoordinatorEntityInfo> {
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
"Can't call show replicas on data_, as variant holds wrong alternative");
std::vector<CoordinatorEntityInfo> result;
const auto &registered_replicas = std::get<CoordinatorData>(data_).registered_replicas_;
result.reserve(registered_replicas.size());
std::ranges::transform(registered_replicas, std::back_inserter(result), [](const auto &replica) {
return CoordinatorEntityInfo{replica.InstanceName(), replica.Endpoint()};
});
return result;
auto CoordinatorState::SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
"Coordinator cannot register replica since variant holds wrong alternative");
return std::visit(
memgraph::utils::Overloaded{[](const CoordinatorMainReplicaData & /*coordinator_main_replica_data*/) {
return SetInstanceToMainCoordinatorStatus::NOT_COORDINATOR;
},
[&instance_name](CoordinatorInstance &coordinator_instance) {
return coordinator_instance.SetReplicationInstanceToMain(instance_name);
}},
data_);
}
auto CoordinatorState::ShowMain() const -> std::optional<CoordinatorEntityInfo> {
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
"Can't call show main on data_, as variant holds wrong alternative");
const auto &registered_main = std::get<CoordinatorData>(data_).registered_main_;
if (registered_main) {
return CoordinatorEntityInfo{registered_main->InstanceName(), registered_main->Endpoint()};
}
return std::nullopt;
}
auto CoordinatorState::PingReplicas() const -> std::unordered_map<std::string_view, bool> {
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
"Can't call ping replicas on data_, as variant holds wrong alternative");
std::unordered_map<std::string_view, bool> result;
const auto &registered_replicas = std::get<CoordinatorData>(data_).registered_replicas_;
result.reserve(registered_replicas.size());
for (const CoordinatorClient &replica_client : registered_replicas) {
result.emplace(replica_client.InstanceName(), replica_client.DoHealthCheck());
}
return result;
}
auto CoordinatorState::PingMain() const -> std::optional<CoordinatorEntityHealthInfo> {
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_),
"Can't call show main on data_, as variant holds wrong alternative");
const auto &registered_main = std::get<CoordinatorData>(data_).registered_main_;
if (registered_main) {
return CoordinatorEntityHealthInfo{registered_main->InstanceName(), registered_main->DoHealthCheck()};
}
return std::nullopt;
}
auto CoordinatorState::DoFailover() -> DoFailoverStatus {
// 1. MAIN is already down, stop sending frequent checks
// 2. find new replica (coordinator)
// 3. make copy replica's client as potential new main client (coordinator)
// 4. send failover RPC to new main (coordinator and new main)
// 5. exchange old main to new main (coordinator)
// 6. remove replica which was promoted to main from all replicas -> this will shut down RPC frequent check client
// (coordinator)
// 7. for new main start frequent checks (coordinator)
MG_ASSERT(std::holds_alternative<CoordinatorData>(data_), "Cannot do failover since variant holds wrong alternative");
using ReplicationClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
// 1.
auto &current_main = std::get<CoordinatorData>(data_).registered_main_;
if (!current_main) {
return DoFailoverStatus::CLUSTER_UNINITIALIZED;
}
if (current_main->DoHealthCheck()) {
return DoFailoverStatus::MAIN_ALIVE;
}
current_main->StopFrequentCheck();
// 2.
// Get all replicas and find new main
auto &registered_replicas = std::get<CoordinatorData>(data_).registered_replicas_;
const auto chosen_replica = std::ranges::find_if(
registered_replicas, [](const CoordinatorClient &replica) { return replica.DoHealthCheck(); });
if (chosen_replica == registered_replicas.end()) {
return DoFailoverStatus::ALL_REPLICAS_DOWN;
}
std::vector<ReplicationClientInfo> repl_clients_info;
repl_clients_info.reserve(registered_replicas.size() - 1);
std::ranges::for_each(registered_replicas, [&chosen_replica, &repl_clients_info](const CoordinatorClient &replica) {
if (replica != *chosen_replica) {
repl_clients_info.emplace_back(replica.ReplicationClientInfo());
}
});
// 3.
// Set on coordinator data of new main
// allocate resources for new main, clear replication info on this replica as main
// set last response time
auto potential_new_main = std::make_unique<CoordinatorClient>(chosen_replica->Config());
potential_new_main->ReplicationClientInfo().reset();
potential_new_main->UpdateTimeCheck(chosen_replica->GetLastTimeResponse());
// 4.
if (!chosen_replica->SendPromoteReplicaToMainRpc(std::move(repl_clients_info))) {
spdlog::error("Sent RPC message, but exception was caught, aborting Failover");
// TODO: new status and rollback all changes that were done...
MG_ASSERT(false, "RPC message failed");
}
// 5.
current_main = std::move(potential_new_main);
// 6. remove old replica
// TODO: Stop pinging chosen_replica before failover.
// Check that it doesn't fail when you call StopFrequentCheck if it is already stopped
registered_replicas.erase(chosen_replica);
// 7.
current_main->StartFrequentCheck();
return DoFailoverStatus::SUCCESS;
auto CoordinatorState::ShowInstances() const -> std::vector<InstanceStatus> {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
"Can't call show instances on data_, as variant holds wrong alternative");
return std::get<CoordinatorInstance>(data_).ShowInstances();
}
auto CoordinatorState::GetCoordinatorServer() const -> CoordinatorServer & {
@@ -217,5 +81,13 @@ auto CoordinatorState::GetCoordinatorServer() const -> CoordinatorServer & {
"Cannot get coordinator server since variant holds wrong alternative");
return *std::get<CoordinatorMainReplicaData>(data_).coordinator_server_;
}
auto CoordinatorState::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
-> void {
MG_ASSERT(std::holds_alternative<CoordinatorInstance>(data_),
"Coordinator cannot register replica since variant holds wrong alternative");
return std::get<CoordinatorInstance>(data_).AddCoordinatorInstance(raft_server_id, raft_port, raft_address);
}
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,111 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "nuraft/coordinator_state_machine.hpp"
namespace memgraph::coordination {
auto CoordinatorStateMachine::EncodeRegisterReplicationInstance(const std::string &name) -> ptr<buffer> {
std::string str_log = name + "_replica";
ptr<buffer> log = buffer::alloc(sizeof(uint32_t) + str_log.size());
buffer_serializer bs(log);
bs.put_str(str_log);
return log;
}
auto CoordinatorStateMachine::DecodeRegisterReplicationInstance(buffer &data) -> std::string {
buffer_serializer bs(data);
return bs.get_str();
}
auto CoordinatorStateMachine::pre_commit(ulong const log_idx, buffer &data) -> ptr<buffer> {
buffer_serializer bs(data);
std::string str = bs.get_str();
spdlog::info("pre_commit {} : {}", log_idx, str);
return nullptr;
}
auto CoordinatorStateMachine::commit(ulong const log_idx, buffer &data) -> ptr<buffer> {
buffer_serializer bs(data);
std::string str = bs.get_str();
spdlog::info("commit {} : {}", log_idx, str);
last_committed_idx_ = log_idx;
return nullptr;
}
auto CoordinatorStateMachine::commit_config(ulong const log_idx, ptr<cluster_config> & /*new_conf*/) -> void {
last_committed_idx_ = log_idx;
}
auto CoordinatorStateMachine::rollback(ulong const log_idx, buffer &data) -> void {
buffer_serializer bs(data);
std::string str = bs.get_str();
spdlog::info("rollback {} : {}", log_idx, str);
}
auto CoordinatorStateMachine::read_logical_snp_obj(snapshot & /*snapshot*/, void *& /*user_snp_ctx*/, ulong /*obj_id*/,
ptr<buffer> &data_out, bool &is_last_obj) -> int {
// Put dummy data.
data_out = buffer::alloc(sizeof(int32));
buffer_serializer bs(data_out);
bs.put_i32(0);
is_last_obj = true;
return 0;
}
auto CoordinatorStateMachine::save_logical_snp_obj(snapshot &s, ulong &obj_id, buffer & /*data*/, bool /*is_first_obj*/,
bool /*is_last_obj*/) -> void {
spdlog::info("save snapshot {} term {} object ID", s.get_last_log_idx(), s.get_last_log_term(), obj_id);
// Request next object.
obj_id++;
}
auto CoordinatorStateMachine::apply_snapshot(snapshot &s) -> bool {
spdlog::info("apply snapshot {} term {}", s.get_last_log_idx(), s.get_last_log_term());
{
auto lock = std::lock_guard{last_snapshot_lock_};
ptr<buffer> snp_buf = s.serialize();
last_snapshot_ = snapshot::deserialize(*snp_buf);
}
return true;
}
auto CoordinatorStateMachine::free_user_snp_ctx(void *&user_snp_ctx) -> void {}
auto CoordinatorStateMachine::last_snapshot() -> ptr<snapshot> {
auto lock = std::lock_guard{last_snapshot_lock_};
return last_snapshot_;
}
auto CoordinatorStateMachine::last_commit_index() -> ulong { return last_committed_idx_; }
auto CoordinatorStateMachine::create_snapshot(snapshot &s, async_result<bool>::handler_type &when_done) -> void {
spdlog::info("create snapshot {} term {}", s.get_last_log_idx(), s.get_last_log_term());
// Clone snapshot from `s`.
{
auto lock = std::lock_guard{last_snapshot_lock_};
ptr<buffer> snp_buf = s.serialize();
last_snapshot_ = snapshot::deserialize(*snp_buf);
}
ptr<std::exception> except(nullptr);
bool ret = true;
when_done(ret, except);
}
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,68 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "nuraft/coordinator_state_manager.hpp"
namespace memgraph::coordination {
using nuraft::cluster_config;
using nuraft::cs_new;
using nuraft::srv_config;
using nuraft::srv_state;
using nuraft::state_mgr;
CoordinatorStateManager::CoordinatorStateManager(int srv_id, std::string const &endpoint)
: my_id_(srv_id), my_endpoint_(endpoint), cur_log_store_(cs_new<CoordinatorLogStore>()) {
my_srv_config_ = cs_new<srv_config>(srv_id, endpoint);
// Initial cluster config: contains only one server (myself).
cluster_config_ = cs_new<cluster_config>();
cluster_config_->get_servers().push_back(my_srv_config_);
}
auto CoordinatorStateManager::load_config() -> ptr<cluster_config> {
// Just return in-memory data in this example.
// May require reading from disk here, if it has been written to disk.
return cluster_config_;
}
auto CoordinatorStateManager::save_config(cluster_config const &config) -> void {
// Just keep in memory in this example.
// Need to write to disk here, if want to make it durable.
ptr<buffer> buf = config.serialize();
cluster_config_ = cluster_config::deserialize(*buf);
}
auto CoordinatorStateManager::save_state(srv_state const &state) -> void {
// Just keep in memory in this example.
// Need to write to disk here, if want to make it durable.
ptr<buffer> buf = state.serialize();
saved_state_ = srv_state::deserialize(*buf);
}
auto CoordinatorStateManager::read_state() -> ptr<srv_state> {
// Just return in-memory data in this example.
// May require reading from disk here, if it has been written to disk.
return saved_state_;
}
auto CoordinatorStateManager::load_log_store() -> ptr<log_store> { return cur_log_store_; }
auto CoordinatorStateManager::server_id() -> int32 { return my_id_; }
auto CoordinatorStateManager::system_exit(int const exit_code) -> void {}
auto CoordinatorStateManager::GetSrvConfig() const -> ptr<srv_config> { return my_srv_config_; }
} // namespace memgraph::coordination
#endif

View File

@@ -11,58 +11,67 @@
#pragma once
#include "utils/uuid.hpp"
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_config.hpp"
#include "rpc/client.hpp"
#include "utils/scheduler.hpp"
#include "utils/thread_pool.hpp"
#include <string_view>
namespace memgraph::coordination {
class CoordinatorInstance;
using HealthCheckCallback = std::function<void(CoordinatorInstance *, std::string_view)>;
using ReplicationClientsInfo = std::vector<ReplClientInfo>;
class CoordinatorClient {
public:
explicit CoordinatorClient(const CoordinatorClientConfig &config);
explicit CoordinatorClient(CoordinatorInstance *coord_instance, CoordinatorClientConfig config,
HealthCheckCallback succ_cb, HealthCheckCallback fail_cb);
~CoordinatorClient();
~CoordinatorClient() = default;
CoordinatorClient(CoordinatorClient &other) = delete;
CoordinatorClient &operator=(CoordinatorClient const &other) = delete;
CoordinatorClient(CoordinatorClient &) = delete;
CoordinatorClient &operator=(CoordinatorClient const &) = delete;
CoordinatorClient(CoordinatorClient &&) noexcept = delete;
CoordinatorClient &operator=(CoordinatorClient &&) noexcept = delete;
void StartFrequentCheck();
void StopFrequentCheck();
void PauseFrequentCheck();
void ResumeFrequentCheck();
auto DoHealthCheck() const -> bool;
auto SendPromoteReplicaToMainRpc(
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info) const -> bool;
auto InstanceName() const -> std::string;
auto SocketAddress() const -> std::string;
auto InstanceName() const -> std::string_view;
auto Endpoint() const -> io::network::Endpoint const &;
auto Config() const -> CoordinatorClientConfig const &;
auto ReplicationClientInfo() const -> CoordinatorClientConfig::ReplicationClientInfo const &;
auto ReplicationClientInfo() -> std::optional<CoordinatorClientConfig::ReplicationClientInfo> &;
void UpdateTimeCheck(const std::chrono::system_clock::time_point &last_checked_time);
auto GetLastTimeResponse() -> std::chrono::system_clock::time_point;
[[nodiscard]] auto DemoteToReplica() const -> bool;
auto SendPromoteReplicaToMainRpc(const utils::UUID &uuid, ReplicationClientsInfo replication_clients_info) const
-> bool;
auto SendSwapMainUUIDRpc(const utils::UUID &uuid) const -> bool;
auto ReplicationClientInfo() const -> ReplClientInfo;
auto SetCallbacks(HealthCheckCallback succ_cb, HealthCheckCallback fail_cb) -> void;
auto RpcClient() -> rpc::Client & { return rpc_client_; }
friend bool operator==(CoordinatorClient const &first, CoordinatorClient const &second) {
return first.config_ == second.config_;
}
private:
utils::ThreadPool thread_pool_{1};
utils::Scheduler replica_checker_;
utils::Scheduler instance_checker_;
// TODO: (andi) Pimpl?
communication::ClientContext rpc_context_;
mutable rpc::Client rpc_client_;
CoordinatorClientConfig config_;
std::atomic<std::chrono::system_clock::time_point> last_response_time_{};
static constexpr int alive_response_time_difference_sec_{5};
CoordinatorClientConfig config_;
CoordinatorInstance *coord_instance_;
HealthCheckCallback succ_cb_;
HealthCheckCallback fail_cb_;
};
} // namespace memgraph::coordination

View File

@@ -12,23 +12,11 @@
#pragma once
#ifdef MG_ENTERPRISE
#include "io/network/endpoint.hpp"
#include <string>
namespace memgraph::coordination {
struct CoordinatorEntityInfo {
std::string_view name;
const io::network::Endpoint &endpoint;
};
struct CoordinatorEntityHealthInfo {
std::string_view name;
bool alive;
struct CoordinatorClusterConfig {
static constexpr int alive_response_time_difference_sec_{5};
};
} // namespace memgraph::coordination
#endif

View File

@@ -25,16 +25,14 @@ namespace memgraph::coordination {
inline constexpr auto *kDefaultReplicationServerIp = "0.0.0.0";
struct CoordinatorClientConfig {
const std::string instance_name;
const std::string ip_address;
const uint16_t port{};
std::string instance_name;
std::string ip_address;
uint16_t port{};
std::chrono::seconds health_check_frequency_sec{1};
// Frequency with which coordinator pings main/replicas about it status
const 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 {
// Should be the same as CoordinatorClientConfig's instance_name
std::string instance_name;
replication_coordination_glue::ReplicationMode replication_mode{};
std::string replication_ip_address;
@@ -43,20 +41,22 @@ struct CoordinatorClientConfig {
friend bool operator==(ReplicationClientInfo const &, ReplicationClientInfo const &) = default;
};
std::optional<ReplicationClientInfo> replication_client_info;
ReplicationClientInfo replication_client_info;
struct SSL {
const std::string key_file;
const std::string cert_file;
std::string key_file;
std::string cert_file;
friend bool operator==(const SSL &, const SSL &) = default;
};
const std::optional<SSL> ssl;
std::optional<SSL> ssl;
friend bool operator==(CoordinatorClientConfig const &, CoordinatorClientConfig const &) = default;
};
using ReplClientInfo = CoordinatorClientConfig::ReplicationClientInfo;
struct CoordinatorServerConfig {
std::string ip_address;
uint16_t port{};

View File

@@ -16,16 +16,60 @@
#include "utils/exceptions.hpp"
namespace memgraph::coordination {
class CoordinatorFailoverException final : public utils::BasicException {
class CoordinatorRegisterInstanceException final : public utils::BasicException {
public:
explicit CoordinatorFailoverException(const std::string_view what) noexcept
: BasicException("Failover didn't complete successfully: " + std::string(what)) {}
explicit CoordinatorRegisterInstanceException(const std::string_view what) noexcept
: BasicException("Failed to create instance: " + std::string(what)) {}
template <class... Args>
explicit CoordinatorFailoverException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: CoordinatorFailoverException(fmt::format(fmt, std::forward<Args>(args)...)) {}
explicit CoordinatorRegisterInstanceException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: CoordinatorRegisterInstanceException(fmt::format(fmt, std::forward<Args>(args)...)) {}
SPECIALIZE_GET_EXCEPTION_NAME(CoordinatorFailoverException)
SPECIALIZE_GET_EXCEPTION_NAME(CoordinatorRegisterInstanceException)
};
class RaftServerStartException final : public utils::BasicException {
public:
explicit RaftServerStartException(std::string_view what) noexcept : BasicException(what) {}
template <class... Args>
explicit RaftServerStartException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: RaftServerStartException(fmt::format(fmt, std::forward<Args>(args)...)) {}
SPECIALIZE_GET_EXCEPTION_NAME(RaftServerStartException)
};
class RaftAddServerException final : public utils::BasicException {
public:
explicit RaftAddServerException(std::string_view what) noexcept : BasicException(what) {}
template <class... Args>
explicit RaftAddServerException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: RaftAddServerException(fmt::format(fmt, std::forward<Args>(args)...)) {}
SPECIALIZE_GET_EXCEPTION_NAME(RaftAddServerException)
};
class RaftBecomeLeaderException final : public utils::BasicException {
public:
explicit RaftBecomeLeaderException(std::string_view what) noexcept : BasicException(what) {}
template <class... Args>
explicit RaftBecomeLeaderException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: RaftBecomeLeaderException(fmt::format(fmt, std::forward<Args>(args)...)) {}
SPECIALIZE_GET_EXCEPTION_NAME(RaftBecomeLeaderException)
};
class RaftCouldNotFindEntryException final : public utils::BasicException {
public:
explicit RaftCouldNotFindEntryException(std::string_view what) noexcept : BasicException(what) {}
template <class... Args>
explicit RaftCouldNotFindEntryException(fmt::format_string<Args...> fmt, Args &&...args) noexcept
: RaftCouldNotFindEntryException(fmt::format(fmt, std::forward<Args>(args)...)) {}
SPECIALIZE_GET_EXCEPTION_NAME(RaftCouldNotFindEntryException)
};
} // namespace memgraph::coordination

View File

@@ -0,0 +1,40 @@
// 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);
};
} // namespace memgraph::dbms
#endif

View File

@@ -0,0 +1,57 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_server.hpp"
#include "coordination/instance_status.hpp"
#include "coordination/raft_instance.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
#include "coordination/replication_instance.hpp"
#include "utils/rw_lock.hpp"
#include "utils/thread_pool.hpp"
#include <list>
namespace memgraph::coordination {
class CoordinatorInstance {
public:
CoordinatorInstance();
[[nodiscard]] auto RegisterReplicationInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
[[nodiscard]] auto SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
auto ShowInstances() const -> std::vector<InstanceStatus>;
auto TryFailover() -> void;
auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
private:
auto ClusterHasAliveMain_() const -> bool;
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_;
RaftInstance self_;
};
} // namespace memgraph::coordination
#endif

View File

@@ -11,6 +11,7 @@
#pragma once
#include "utils/uuid.hpp"
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_config.hpp"
@@ -26,10 +27,13 @@ struct PromoteReplicaToMainReq {
static void Load(PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader);
static void Save(const PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder);
explicit PromoteReplicaToMainReq(std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info)
: replication_clients_info(std::move(replication_clients_info)) {}
explicit PromoteReplicaToMainReq(const utils::UUID &uuid,
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info)
: main_uuid_(uuid), replication_clients_info(std::move(replication_clients_info)) {}
PromoteReplicaToMainReq() = default;
// get uuid here
utils::UUID main_uuid_;
std::vector<CoordinatorClientConfig::ReplicationClientInfo> replication_clients_info;
};
@@ -48,19 +52,54 @@ struct PromoteReplicaToMainRes {
using PromoteReplicaToMainRpc = rpc::RequestResponse<PromoteReplicaToMainReq, PromoteReplicaToMainRes>;
struct DemoteMainToReplicaReq {
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);
explicit DemoteMainToReplicaReq(CoordinatorClientConfig::ReplicationClientInfo replication_client_info)
: replication_client_info(std::move(replication_client_info)) {}
DemoteMainToReplicaReq() = default;
CoordinatorClientConfig::ReplicationClientInfo replication_client_info;
};
struct DemoteMainToReplicaRes {
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);
explicit DemoteMainToReplicaRes(bool success) : success(success) {}
DemoteMainToReplicaRes() = default;
bool success;
};
using DemoteMainToReplicaRpc = rpc::RequestResponse<DemoteMainToReplicaReq, DemoteMainToReplicaRes>;
} // namespace memgraph::coordination
// SLK serialization declarations
namespace memgraph::slk {
// PromoteReplicaToMainRpc
void Save(const memgraph::coordination::PromoteReplicaToMainRes &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::PromoteReplicaToMainRes *self, memgraph::slk::Reader *reader);
void Save(const memgraph::coordination::PromoteReplicaToMainReq &self, memgraph::slk::Builder *builder);
void Load(memgraph::coordination::PromoteReplicaToMainReq *self, memgraph::slk::Reader *reader);
// 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);
} // namespace memgraph::slk
#endif

View File

@@ -13,81 +13,43 @@
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_client.hpp"
#include "coordination/coordinator_entity_info.hpp"
#include "coordination/coordinator_instance.hpp"
#include "coordination/coordinator_server.hpp"
#include "rpc/server.hpp"
#include "utils/result.hpp"
#include "utils/rw_spin_lock.hpp"
#include "utils/synchronized.hpp"
#include "coordination/instance_status.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
#include <list>
#include <variant>
namespace memgraph::coordination {
enum class RegisterMainReplicaCoordinatorStatus : uint8_t {
NAME_EXISTS,
END_POINT_EXISTS,
COULD_NOT_BE_PERSISTED,
NOT_COORDINATOR,
SUCCESS
};
enum class DoFailoverStatus : uint8_t { SUCCESS, ALL_REPLICAS_DOWN, MAIN_ALIVE, CLUSTER_UNINITIALIZED };
class CoordinatorState {
public:
CoordinatorState();
~CoordinatorState() = default;
CoordinatorState(const CoordinatorState &) = delete;
CoordinatorState &operator=(const CoordinatorState &) = delete;
CoordinatorState(CoordinatorState const &) = delete;
CoordinatorState &operator=(CoordinatorState const &) = delete;
CoordinatorState(CoordinatorState &&other) noexcept : data_(std::move(other.data_)) {}
CoordinatorState(CoordinatorState &&) noexcept = delete;
CoordinatorState &operator=(CoordinatorState &&) noexcept = delete;
CoordinatorState &operator=(CoordinatorState &&other) noexcept {
if (this == &other) {
return *this;
}
data_ = std::move(other.data_);
return *this;
}
[[nodiscard]] auto RegisterReplicationInstance(CoordinatorClientConfig config) -> RegisterInstanceCoordinatorStatus;
auto RegisterReplica(const CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus, CoordinatorClient *>;
[[nodiscard]] auto SetReplicationInstanceToMain(std::string instance_name) -> SetInstanceToMainCoordinatorStatus;
auto RegisterMain(const CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus, CoordinatorClient *>;
auto ShowInstances() const -> std::vector<InstanceStatus>;
auto ShowReplicas() const -> std::vector<CoordinatorEntityInfo>;
auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
auto PingReplicas() const -> std::unordered_map<std::string_view, bool>;
auto ShowMain() const -> std::optional<CoordinatorEntityInfo>;
auto PingMain() const -> std::optional<CoordinatorEntityHealthInfo>;
// The client code must check that the server exists before calling this method.
// NOTE: The client code must check that the server exists before calling this method.
auto GetCoordinatorServer() const -> CoordinatorServer &;
auto DoFailover() -> DoFailoverStatus;
private:
// TODO: Data is not thread safe
// Coordinator stores registered replicas and main
struct CoordinatorData {
std::list<CoordinatorClient> registered_replicas_;
std::unique_ptr<CoordinatorClient> registered_main_;
};
// Data which each main and replica stores
struct CoordinatorMainReplicaData {
std::unique_ptr<CoordinatorServer> coordinator_server_;
};
std::variant<CoordinatorData, CoordinatorMainReplicaData> data_;
std::variant<CoordinatorInstance, CoordinatorMainReplicaData> data_;
};
} // namespace memgraph::coordination

View File

@@ -0,0 +1,34 @@
// 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 "io/network/endpoint.hpp"
#include <string_view>
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 {
std::string instance_name;
std::string raft_socket_address;
std::string coord_socket_address;
std::string cluster_role;
bool is_alive;
};
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,73 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
#include <flags/replication.hpp>
#include <libnuraft/nuraft.hxx>
namespace memgraph::coordination {
using BecomeLeaderCb = std::function<void()>;
using BecomeFollowerCb = std::function<void()>;
using nuraft::buffer;
using nuraft::logger;
using nuraft::ptr;
using nuraft::raft_launcher;
using nuraft::raft_server;
using nuraft::srv_config;
using nuraft::state_machine;
using nuraft::state_mgr;
using raft_result = nuraft::cmd_result<ptr<buffer>>;
class RaftInstance {
public:
RaftInstance(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb);
RaftInstance(RaftInstance const &other) = delete;
RaftInstance &operator=(RaftInstance const &other) = delete;
RaftInstance(RaftInstance &&other) noexcept = delete;
RaftInstance &operator=(RaftInstance &&other) noexcept = delete;
~RaftInstance();
auto InstanceName() const -> std::string;
auto RaftSocketAddress() const -> std::string;
auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
auto GetAllCoordinators() const -> std::vector<ptr<srv_config>>;
auto RequestLeadership() -> bool;
auto IsLeader() const -> bool;
auto AppendRegisterReplicationInstance(std::string const &instance) -> ptr<raft_result>;
private:
ptr<state_machine> state_machine_;
ptr<state_mgr> state_manager_;
ptr<raft_server> raft_server_;
ptr<logger> logger_;
raft_launcher launcher_;
// TODO: (andi) I think variables below can be abstracted
uint32_t raft_server_id_;
uint32_t raft_port_;
std::string raft_address_;
BecomeLeaderCb become_leader_cb_;
BecomeFollowerCb become_follower_cb_;
};
} // namespace memgraph::coordination
#endif

View File

@@ -13,21 +13,28 @@
#ifdef MG_ENTERPRISE
#include "slk/serialization.hpp"
#include <cstdint>
namespace memgraph::dbms {
namespace memgraph::coordination {
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);
enum class RegisterInstanceCoordinatorStatus : uint8_t {
NAME_EXISTS,
ENDPOINT_EXISTS,
NOT_COORDINATOR,
RPC_FAILED,
NOT_LEADER,
RAFT_COULD_NOT_ACCEPT,
RAFT_COULD_NOT_APPEND,
SUCCESS
};
} // namespace memgraph::dbms
enum class SetInstanceToMainCoordinatorStatus : uint8_t {
NO_INSTANCE_WITH_NAME,
NOT_COORDINATOR,
SUCCESS,
COULD_NOT_PROMOTE_TO_MAIN,
SWAP_UUID_FAILED
};
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,86 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_client.hpp"
#include "coordination/coordinator_cluster_config.hpp"
#include "coordination/coordinator_exceptions.hpp"
#include "replication_coordination_glue/role.hpp"
#include <libnuraft/nuraft.hxx>
#include "utils/uuid.hpp"
namespace memgraph::coordination {
class CoordinatorInstance;
class ReplicationInstance {
public:
ReplicationInstance(CoordinatorInstance *peer, CoordinatorClientConfig config, HealthCheckCallback succ_cb,
HealthCheckCallback fail_cb);
ReplicationInstance(ReplicationInstance const &other) = delete;
ReplicationInstance &operator=(ReplicationInstance const &other) = delete;
ReplicationInstance(ReplicationInstance &&other) noexcept = delete;
ReplicationInstance &operator=(ReplicationInstance &&other) noexcept = delete;
~ReplicationInstance() = default;
auto OnSuccessPing() -> void;
auto OnFailPing() -> bool;
auto IsAlive() const -> bool;
auto InstanceName() const -> std::string;
auto SocketAddress() const -> std::string;
auto IsReplica() const -> bool;
auto IsMain() const -> bool;
auto PromoteToMain(utils::UUID uuid, ReplicationClientsInfo repl_clients_info, HealthCheckCallback main_succ_cb,
HealthCheckCallback main_fail_cb) -> bool;
auto DemoteToReplica(HealthCheckCallback replica_succ_cb, HealthCheckCallback replica_fail_cb) -> bool;
auto StartFrequentCheck() -> void;
auto StopFrequentCheck() -> void;
auto PauseFrequentCheck() -> void;
auto ResumeFrequentCheck() -> void;
auto ReplicationClientInfo() const -> ReplClientInfo;
auto SendSwapAndUpdateUUID(const utils::UUID &main_uuid) -> bool;
auto GetClient() -> CoordinatorClient &;
void SetNewMainUUID(const std::optional<utils::UUID> &main_uuid = std::nullopt);
auto GetMainUUID() -> const std::optional<utils::UUID> &;
private:
CoordinatorClient client_;
replication_coordination_glue::ReplicationRole replication_role_;
std::chrono::system_clock::time_point last_response_time_{};
bool is_alive_{false};
// for replica this is main uuid of current main
// for "main" main this same as in CoordinatorData
// it is set to nullopt when replica is down
// TLDR; when replica is down and comes back up we reset uuid of main replica is listening to
// so we need to send swap uuid again
std::optional<utils::UUID> main_uuid_;
friend bool operator==(ReplicationInstance const &first, ReplicationInstance const &second) {
return first.client_ == second.client_ && first.replication_role_ == second.replication_role_;
}
};
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,70 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
#include <libnuraft/nuraft.hxx>
namespace memgraph::coordination {
using nuraft::buffer;
using nuraft::int32;
using nuraft::int64;
using nuraft::log_entry;
using nuraft::log_store;
using nuraft::ptr;
using nuraft::raft_server;
class CoordinatorLogStore : public log_store {
public:
CoordinatorLogStore();
CoordinatorLogStore(CoordinatorLogStore const &) = delete;
CoordinatorLogStore &operator=(CoordinatorLogStore const &) = delete;
CoordinatorLogStore(CoordinatorLogStore &&) = delete;
CoordinatorLogStore &operator=(CoordinatorLogStore &&) = delete;
~CoordinatorLogStore() override;
ulong next_slot() const override;
ulong start_index() const override;
ptr<log_entry> last_entry() const override;
ulong append(ptr<log_entry> &entry) override;
void write_at(ulong index, ptr<log_entry> &entry) override;
ptr<std::vector<ptr<log_entry>>> log_entries(ulong start, ulong end) override;
ptr<log_entry> entry_at(ulong index) override;
ulong term_at(ulong index) override;
ptr<buffer> pack(ulong index, int32 cnt) override;
void apply_pack(ulong index, buffer &pack) override;
bool compact(ulong last_log_index) override;
bool flush() override;
private:
auto FindOrDefault_(ulong index) const -> ptr<log_entry>;
std::map<ulong, ptr<log_entry>> logs_;
mutable std::mutex logs_lock_;
std::atomic<ulong> start_idx_;
};
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,76 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
#include <spdlog/spdlog.h>
#include <libnuraft/nuraft.hxx>
namespace memgraph::coordination {
using nuraft::async_result;
using nuraft::buffer;
using nuraft::buffer_serializer;
using nuraft::cluster_config;
using nuraft::int32;
using nuraft::ptr;
using nuraft::snapshot;
using nuraft::state_machine;
class CoordinatorStateMachine : public state_machine {
public:
CoordinatorStateMachine() = default;
CoordinatorStateMachine(CoordinatorStateMachine const &) = delete;
CoordinatorStateMachine &operator=(CoordinatorStateMachine const &) = delete;
CoordinatorStateMachine(CoordinatorStateMachine &&) = delete;
CoordinatorStateMachine &operator=(CoordinatorStateMachine &&) = delete;
~CoordinatorStateMachine() override {}
static auto EncodeRegisterReplicationInstance(const std::string &name) -> ptr<buffer>;
static auto DecodeRegisterReplicationInstance(buffer &data) -> std::string;
auto pre_commit(ulong log_idx, buffer &data) -> ptr<buffer> override;
auto commit(ulong log_idx, buffer &data) -> ptr<buffer> override;
auto commit_config(ulong log_idx, ptr<cluster_config> & /*new_conf*/) -> void override;
auto rollback(ulong log_idx, buffer &data) -> void override;
auto read_logical_snp_obj(snapshot & /*snapshot*/, void *& /*user_snp_ctx*/, ulong /*obj_id*/, ptr<buffer> &data_out,
bool &is_last_obj) -> int override;
auto save_logical_snp_obj(snapshot &s, ulong &obj_id, buffer & /*data*/, bool /*is_first_obj*/, bool /*is_last_obj*/)
-> void override;
auto apply_snapshot(snapshot &s) -> bool override;
auto free_user_snp_ctx(void *&user_snp_ctx) -> void override;
auto last_snapshot() -> ptr<snapshot> override;
auto last_commit_index() -> ulong override;
auto create_snapshot(snapshot &s, async_result<bool>::handler_type &when_done) -> void override;
private:
std::atomic<uint64_t> last_committed_idx_{0};
ptr<snapshot> last_snapshot_;
std::mutex last_snapshot_lock_;
};
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,66 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#ifdef MG_ENTERPRISE
#include "nuraft/coordinator_log_store.hpp"
#include <spdlog/spdlog.h>
#include <libnuraft/nuraft.hxx>
namespace memgraph::coordination {
using nuraft::cluster_config;
using nuraft::cs_new;
using nuraft::srv_config;
using nuraft::srv_state;
using nuraft::state_mgr;
class CoordinatorStateManager : public state_mgr {
public:
explicit CoordinatorStateManager(int srv_id, std::string const &endpoint);
CoordinatorStateManager(CoordinatorStateManager const &) = delete;
CoordinatorStateManager &operator=(CoordinatorStateManager const &) = delete;
CoordinatorStateManager(CoordinatorStateManager &&) = delete;
CoordinatorStateManager &operator=(CoordinatorStateManager &&) = delete;
~CoordinatorStateManager() override = default;
auto load_config() -> ptr<cluster_config> override;
auto save_config(cluster_config const &config) -> void override;
auto save_state(srv_state const &state) -> void override;
auto read_state() -> ptr<srv_state> override;
auto load_log_store() -> ptr<log_store> override;
auto server_id() -> int32 override;
auto system_exit(int exit_code) -> void override;
auto GetSrvConfig() const -> ptr<srv_config>;
private:
int my_id_;
std::string my_endpoint_;
ptr<CoordinatorLogStore> cur_log_store_;
ptr<srv_config> my_srv_config_;
ptr<cluster_config> cluster_config_;
ptr<srv_state> saved_state_;
};
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,126 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "coordination/raft_instance.hpp"
#include "coordination/coordinator_exceptions.hpp"
#include "nuraft/coordinator_state_machine.hpp"
#include "nuraft/coordinator_state_manager.hpp"
#include "utils/counter.hpp"
namespace memgraph::coordination {
using nuraft::asio_service;
using nuraft::cb_func;
using nuraft::CbReturnCode;
using nuraft::cmd_result;
using nuraft::cs_new;
using nuraft::ptr;
using nuraft::raft_params;
using nuraft::raft_server;
using nuraft::srv_config;
using raft_result = cmd_result<ptr<buffer>>;
RaftInstance::RaftInstance(BecomeLeaderCb become_leader_cb, BecomeFollowerCb become_follower_cb)
: raft_server_id_(FLAGS_raft_server_id),
raft_port_(FLAGS_raft_server_port),
raft_address_("127.0.0.1"),
become_leader_cb_(std::move(become_leader_cb)),
become_follower_cb_(std::move(become_follower_cb)) {
auto raft_endpoint = raft_address_ + ":" + std::to_string(raft_port_);
state_manager_ = cs_new<CoordinatorStateManager>(raft_server_id_, raft_endpoint);
state_machine_ = cs_new<CoordinatorStateMachine>();
logger_ = nullptr;
// TODO: (andi) Maybe params file
asio_service::options asio_opts;
asio_opts.thread_pool_size_ = 1; // TODO: (andi) Improve this
raft_params params;
params.heart_beat_interval_ = 100;
params.election_timeout_lower_bound_ = 200;
params.election_timeout_upper_bound_ = 400;
// 5 logs are preserved before the last snapshot
params.reserved_log_items_ = 5;
// Create snapshot for every 5 log appends
params.snapshot_distance_ = 5;
params.client_req_timeout_ = 3000;
params.return_method_ = raft_params::blocking;
raft_server::init_options init_opts;
init_opts.raft_callback_ = [this](cb_func::Type event_type, cb_func::Param *param) -> nuraft::CbReturnCode {
if (event_type == cb_func::BecomeLeader) {
spdlog::info("Node {} became leader", param->leaderId);
become_leader_cb_();
} else if (event_type == cb_func::BecomeFollower) {
spdlog::info("Node {} became follower", param->myId);
become_follower_cb_();
}
return CbReturnCode::Ok;
};
raft_server_ = launcher_.init(state_machine_, state_manager_, logger_, static_cast<int>(raft_port_), asio_opts,
params, init_opts);
if (!raft_server_) {
throw RaftServerStartException("Failed to launch raft server on {}", raft_endpoint);
}
auto maybe_stop = utils::ResettableCounter<20>();
while (!raft_server_->is_initialized() && !maybe_stop()) {
std::this_thread::sleep_for(std::chrono::milliseconds(250));
}
if (!raft_server_->is_initialized()) {
throw RaftServerStartException("Failed to initialize raft server on {}", raft_endpoint);
}
spdlog::info("Raft server started on {}", raft_endpoint);
}
RaftInstance::~RaftInstance() { launcher_.shutdown(); }
auto RaftInstance::InstanceName() const -> std::string { return "coordinator_" + std::to_string(raft_server_id_); }
auto RaftInstance::RaftSocketAddress() const -> std::string { return raft_address_ + ":" + std::to_string(raft_port_); }
auto RaftInstance::AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address)
-> void {
auto const endpoint = raft_address + ":" + std::to_string(raft_port);
srv_config const srv_config_to_add(static_cast<int>(raft_server_id), endpoint);
if (!raft_server_->add_srv(srv_config_to_add)->get_accepted()) {
throw RaftAddServerException("Failed to add server {} to the cluster", endpoint);
}
spdlog::info("Request to add server {} to the cluster accepted", endpoint);
}
auto RaftInstance::GetAllCoordinators() const -> std::vector<ptr<srv_config>> {
std::vector<ptr<srv_config>> all_srv_configs;
raft_server_->get_srv_config_all(all_srv_configs);
return all_srv_configs;
}
auto RaftInstance::IsLeader() const -> bool { return raft_server_->is_leader(); }
auto RaftInstance::RequestLeadership() -> bool {
return raft_server_->is_leader() || raft_server_->request_leadership();
}
auto RaftInstance::AppendRegisterReplicationInstance(std::string const &instance) -> ptr<raft_result> {
auto new_log = CoordinatorStateMachine::EncodeRegisterReplicationInstance(instance);
return raft_server_->append_entries({new_log});
}
} // namespace memgraph::coordination
#endif

View File

@@ -0,0 +1,100 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "coordination/replication_instance.hpp"
#include "replication_coordination_glue/handler.hpp"
namespace memgraph::coordination {
ReplicationInstance::ReplicationInstance(CoordinatorInstance *peer, CoordinatorClientConfig config,
HealthCheckCallback succ_cb, HealthCheckCallback fail_cb)
: client_(peer, std::move(config), std::move(succ_cb), std::move(fail_cb)),
replication_role_(replication_coordination_glue::ReplicationRole::REPLICA) {
if (!client_.DemoteToReplica()) {
throw CoordinatorRegisterInstanceException("Failed to demote instance {} to replica", client_.InstanceName());
}
client_.StartFrequentCheck();
}
auto ReplicationInstance::OnSuccessPing() -> void {
last_response_time_ = std::chrono::system_clock::now();
is_alive_ = true;
}
auto ReplicationInstance::OnFailPing() -> bool {
is_alive_ =
std::chrono::duration_cast<std::chrono::seconds>(std::chrono::system_clock::now() - last_response_time_).count() <
CoordinatorClusterConfig::alive_response_time_difference_sec_;
return is_alive_;
}
auto ReplicationInstance::InstanceName() const -> std::string { return client_.InstanceName(); }
auto ReplicationInstance::SocketAddress() const -> std::string { return client_.SocketAddress(); }
auto ReplicationInstance::IsAlive() const -> bool { return is_alive_; }
auto ReplicationInstance::IsReplica() const -> bool {
return replication_role_ == replication_coordination_glue::ReplicationRole::REPLICA;
}
auto ReplicationInstance::IsMain() const -> bool {
return replication_role_ == replication_coordination_glue::ReplicationRole::MAIN;
}
auto ReplicationInstance::PromoteToMain(utils::UUID uuid, ReplicationClientsInfo repl_clients_info,
HealthCheckCallback main_succ_cb, HealthCheckCallback main_fail_cb) -> bool {
if (!client_.SendPromoteReplicaToMainRpc(uuid, std::move(repl_clients_info))) {
return false;
}
replication_role_ = replication_coordination_glue::ReplicationRole::MAIN;
client_.SetCallbacks(std::move(main_succ_cb), std::move(main_fail_cb));
return true;
}
auto ReplicationInstance::DemoteToReplica(HealthCheckCallback replica_succ_cb, HealthCheckCallback replica_fail_cb)
-> bool {
if (!client_.DemoteToReplica()) {
return false;
}
replication_role_ = replication_coordination_glue::ReplicationRole::REPLICA;
client_.SetCallbacks(std::move(replica_succ_cb), std::move(replica_fail_cb));
return true;
}
auto ReplicationInstance::StartFrequentCheck() -> void { client_.StartFrequentCheck(); }
auto ReplicationInstance::StopFrequentCheck() -> void { client_.StopFrequentCheck(); }
auto ReplicationInstance::PauseFrequentCheck() -> void { client_.PauseFrequentCheck(); }
auto ReplicationInstance::ResumeFrequentCheck() -> void { client_.ResumeFrequentCheck(); }
auto ReplicationInstance::ReplicationClientInfo() const -> CoordinatorClientConfig::ReplicationClientInfo {
return client_.ReplicationClientInfo();
}
auto ReplicationInstance::GetClient() -> CoordinatorClient & { return client_; }
void ReplicationInstance::SetNewMainUUID(const std::optional<utils::UUID> &main_uuid) { main_uuid_ = main_uuid; }
auto ReplicationInstance::GetMainUUID() -> const std::optional<utils::UUID> & { return main_uuid_; }
auto ReplicationInstance::SendSwapAndUpdateUUID(const utils::UUID &main_uuid) -> bool {
if (!replication_coordination_glue::SendSwapMainUUIDRpc(client_.RpcClient(), main_uuid)) {
return false;
}
SetNewMainUUID(main_uuid_);
return true;
}
} // namespace memgraph::coordination
#endif

View File

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

View File

@@ -13,85 +13,30 @@
#include "dbms/coordinator_handler.hpp"
#include "dbms/dbms_handler.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
namespace memgraph::dbms {
CoordinatorHandler::CoordinatorHandler(DbmsHandler &dbms_handler) : dbms_handler_(dbms_handler) {}
CoordinatorHandler::CoordinatorHandler(coordination::CoordinatorState &coordinator_state)
: coordinator_state_(coordinator_state) {}
auto CoordinatorHandler::RegisterReplicaOnCoordinator(const memgraph::coordination::CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus> {
auto instance_client = dbms_handler_.CoordinatorState().RegisterReplica(config);
using repl_status = memgraph::coordination::RegisterMainReplicaCoordinatorStatus;
using dbms_status = memgraph::dbms::RegisterMainReplicaCoordinatorStatus;
if (instance_client.HasError()) {
switch (instance_client.GetError()) {
case memgraph::coordination::RegisterMainReplicaCoordinatorStatus::NOT_COORDINATOR:
MG_ASSERT(false, "Only coordinator instance can register main and replica!");
return {};
case repl_status::NAME_EXISTS:
return dbms_status::NAME_EXISTS;
case repl_status::END_POINT_EXISTS:
return dbms_status::END_POINT_EXISTS;
case repl_status::COULD_NOT_BE_PERSISTED:
return dbms_status::COULD_NOT_BE_PERSISTED;
case repl_status::SUCCESS:
break;
}
}
instance_client.GetValue()->StartFrequentCheck();
return {};
auto CoordinatorHandler::RegisterReplicationInstance(memgraph::coordination::CoordinatorClientConfig config)
-> coordination::RegisterInstanceCoordinatorStatus {
return coordinator_state_.RegisterReplicationInstance(config);
}
auto CoordinatorHandler::RegisterMainOnCoordinator(const memgraph::coordination::CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus> {
auto instance_client = dbms_handler_.CoordinatorState().RegisterMain(config);
if (instance_client.HasError()) switch (instance_client.GetError()) {
case memgraph::coordination::RegisterMainReplicaCoordinatorStatus::NOT_COORDINATOR:
MG_ASSERT(false, "Only coordinator instance can register main and replica!");
case memgraph::coordination::RegisterMainReplicaCoordinatorStatus::NAME_EXISTS:
return memgraph::dbms::RegisterMainReplicaCoordinatorStatus::NAME_EXISTS;
case memgraph::coordination::RegisterMainReplicaCoordinatorStatus::END_POINT_EXISTS:
return memgraph::dbms::RegisterMainReplicaCoordinatorStatus::END_POINT_EXISTS;
case memgraph::coordination::RegisterMainReplicaCoordinatorStatus::COULD_NOT_BE_PERSISTED:
return memgraph::dbms::RegisterMainReplicaCoordinatorStatus::COULD_NOT_BE_PERSISTED;
case memgraph::coordination::RegisterMainReplicaCoordinatorStatus::SUCCESS:
break;
}
instance_client.GetValue()->StartFrequentCheck();
return {};
auto CoordinatorHandler::SetReplicationInstanceToMain(std::string instance_name)
-> coordination::SetInstanceToMainCoordinatorStatus {
return coordinator_state_.SetReplicationInstanceToMain(std::move(instance_name));
}
auto CoordinatorHandler::ShowReplicasOnCoordinator() const -> std::vector<coordination::CoordinatorEntityInfo> {
return dbms_handler_.CoordinatorState().ShowReplicas();
auto CoordinatorHandler::ShowInstances() const -> std::vector<coordination::InstanceStatus> {
return coordinator_state_.ShowInstances();
}
auto CoordinatorHandler::PingReplicasOnCoordinator() const -> std::unordered_map<std::string_view, bool> {
return dbms_handler_.CoordinatorState().PingReplicas();
}
auto CoordinatorHandler::ShowMainOnCoordinator() const -> std::optional<coordination::CoordinatorEntityInfo> {
return dbms_handler_.CoordinatorState().ShowMain();
}
auto CoordinatorHandler::PingMainOnCoordinator() const -> std::optional<coordination::CoordinatorEntityHealthInfo> {
return dbms_handler_.CoordinatorState().PingMain();
}
auto CoordinatorHandler::DoFailover() const -> DoFailoverStatus {
auto status = dbms_handler_.CoordinatorState().DoFailover();
switch (status) {
case memgraph::coordination::DoFailoverStatus::ALL_REPLICAS_DOWN:
return memgraph::dbms::DoFailoverStatus::ALL_REPLICAS_DOWN;
case memgraph::coordination::DoFailoverStatus::SUCCESS:
return memgraph::dbms::DoFailoverStatus::SUCCESS;
case memgraph::coordination::DoFailoverStatus::MAIN_ALIVE:
return memgraph::dbms::DoFailoverStatus::MAIN_ALIVE;
case memgraph::coordination::DoFailoverStatus::CLUSTER_UNINITIALIZED:
return memgraph::dbms::DoFailoverStatus::CLUSTER_UNINITIALIZED;
}
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

View File

@@ -13,54 +13,32 @@
#ifdef MG_ENTERPRISE
#include "utils/result.hpp"
#include "coordination/coordinator_config.hpp"
#include "coordination/coordinator_state.hpp"
#include "coordination/instance_status.hpp"
#include "coordination/register_main_replica_coordinator_status.hpp"
#include <cstdint>
#include <optional>
#include <vector>
namespace memgraph::coordination {
struct CoordinatorEntityInfo;
struct CoordinatorEntityHealthInfo;
struct CoordinatorClientConfig;
} // namespace memgraph::coordination
namespace memgraph::dbms {
enum class RegisterMainReplicaCoordinatorStatus : uint8_t {
NAME_EXISTS,
END_POINT_EXISTS,
COULD_NOT_BE_PERSISTED,
NOT_COORDINATOR,
SUCCESS
};
enum class DoFailoverStatus : uint8_t { SUCCESS, ALL_REPLICAS_DOWN, MAIN_ALIVE, CLUSTER_UNINITIALIZED };
class DbmsHandler;
class CoordinatorHandler {
public:
explicit CoordinatorHandler(DbmsHandler &dbms_handler);
explicit CoordinatorHandler(coordination::CoordinatorState &coordinator_state);
auto RegisterReplicaOnCoordinator(const memgraph::coordination::CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus>;
auto RegisterReplicationInstance(coordination::CoordinatorClientConfig config)
-> coordination::RegisterInstanceCoordinatorStatus;
auto RegisterMainOnCoordinator(const memgraph::coordination::CoordinatorClientConfig &config)
-> utils::BasicResult<RegisterMainReplicaCoordinatorStatus>;
auto SetReplicationInstanceToMain(std::string instance_name) -> coordination::SetInstanceToMainCoordinatorStatus;
auto ShowReplicasOnCoordinator() const -> std::vector<memgraph::coordination::CoordinatorEntityInfo>;
auto ShowInstances() const -> std::vector<coordination::InstanceStatus>;
auto ShowMainOnCoordinator() const -> std::optional<memgraph::coordination::CoordinatorEntityInfo>;
auto PingReplicasOnCoordinator() const -> std::unordered_map<std::string_view, bool>;
auto PingMainOnCoordinator() const -> std::optional<memgraph::coordination::CoordinatorEntityHealthInfo>;
auto DoFailover() const -> DoFailoverStatus;
auto AddCoordinatorInstance(uint32_t raft_server_id, uint32_t raft_port, std::string raft_address) -> void;
private:
DbmsHandler &dbms_handler_;
coordination::CoordinatorState &coordinator_state_;
};
} // namespace memgraph::dbms

View File

@@ -1,106 +0,0 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#ifdef MG_ENTERPRISE
#include "dbms/coordinator_handlers.hpp"
#include "dbms/utils.hpp"
#include "coordination/coordinator_exceptions.hpp"
#include "coordination/coordinator_rpc.hpp"
#include "dbms/dbms_handler.hpp"
#include "dbms/replication_client.hpp"
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 from coordinator server");
CoordinatorHandlers::PromoteReplicaToMainHandler(dbms_handler, req_reader, 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;
}
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);
std::vector<replication::ReplicationClientConfig> clients_config;
clients_config.reserve(req.replication_clients_info.size());
std::ranges::transform(req.replication_clients_info, std::back_inserter(clients_config),
[](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,
};
});
std::ranges::for_each(clients_config, [&dbms_handler, &repl_state, &res_builder](const auto &config) {
auto instance_client = repl_state.RegisterReplica(config);
if (instance_client.HasError()) {
switch (instance_client.GetError()) {
case memgraph::replication::RegisterReplicaError::NOT_MAIN:
spdlog::error("Failover must be performed to main!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
case memgraph::replication::RegisterReplicaError::NAME_EXISTS:
spdlog::error("Replica with the same name already exists!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
case memgraph::replication::RegisterReplicaError::END_POINT_EXISTS:
spdlog::error("Replica with the same endpoint already exists!");
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
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;
}
}
auto &instance_client_ref = *instance_client.GetValue();
const bool all_clients_good = memgraph::dbms::RegisterAllDatabasesClients(dbms_handler, instance_client_ref);
if (!all_clients_good) {
spdlog::error("Failed to register all databases to the REPLICA \"{}\"", config.name);
slk::Save(coordination::PromoteReplicaToMainRes{false}, res_builder);
return;
}
StartReplicaClient(dbms_handler, instance_client_ref);
});
slk::Save(coordination::PromoteReplicaToMainRes{true}, res_builder);
}
} // namespace memgraph::dbms
#endif

View File

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

View File

@@ -110,7 +110,7 @@ class Database {
* @param force_directory Use the configured directory, do not try to decipher the multi-db version
* @return DatabaseInfo
*/
DatabaseInfo GetInfo(bool force_directory, replication::ReplicationRole replication_role) const {
DatabaseInfo GetInfo(bool force_directory, replication_coordination_glue::ReplicationRole replication_role) const {
DatabaseInfo info;
info.storage_info = storage_->GetInfo(force_directory, replication_role);
info.triggers = trigger_store_.GetTriggerInfo().size();

View File

@@ -11,29 +11,75 @@
#include "dbms/dbms_handler.hpp"
#include "dbms/coordinator_handlers.hpp"
#include "flags/replication.hpp"
#include <cstdint>
#include <filesystem>
#include "dbms/constants.hpp"
#include "dbms/global.hpp"
#include "dbms/replication_client.hpp"
#include "spdlog/spdlog.h"
#include "system/include/system/system.hpp"
#include "utils/exceptions.hpp"
#include "utils/logging.hpp"
#include "utils/uuid.hpp"
namespace memgraph::dbms {
#ifdef MG_ENTERPRISE
namespace {
constexpr std::string_view kDBPrefix = "database:"; // Key prefix for database durability
constexpr std::string_view kLastCommitedSystemTsKey = "last_commited_system_ts"; // Key for timestamp durability
constexpr std::string_view kDBPrefix = "database:"; // Key prefix for database durability
std::string RegisterReplicaErrorToString(query::RegisterReplicaError error) {
switch (error) {
using enum query::RegisterReplicaError;
case NAME_EXISTS:
return "NAME_EXISTS";
case ENDPOINT_EXISTS:
return "ENDPOINT_EXISTS";
case CONNECTION_FAILED:
return "CONNECTION_FAILED";
case COULD_NOT_BE_PERSISTED:
return "COULD_NOT_BE_PERSISTED";
case ERROR_ACCEPTING_MAIN:
return "ERROR_ACCEPTING_MAIN";
}
}
// Per storage
// NOTE Storage will connect to all replicas. Future work might change this
void RestoreReplication(replication::RoleMainData &mainData, DatabaseAccess db_acc) {
spdlog::info("Restoring replication role.");
// Each individual client has already been restored and started. Here we just go through each database and start its
// client
for (auto &instance_client : mainData.registered_replicas_) {
spdlog::info("Replica {} restoration started for {}.", instance_client.name_, db_acc->name());
const auto &ret = db_acc->storage()->repl_storage_state_.replication_clients_.WithLock(
[&, db_acc](auto &storage_clients) mutable -> utils::BasicResult<query::RegisterReplicaError> {
auto client = std::make_unique<storage::ReplicationStorageClient>(instance_client, mainData.uuid_);
auto *storage = db_acc->storage();
client->Start(storage, std::move(db_acc));
// After start the storage <-> replica state should be READY or RECOVERING (if correctly started)
// MAYBE_BEHIND isn't a statement of the current state, this is the default value
// Failed to start due to branching of MAIN and REPLICA
if (client->State() == storage::replication::ReplicaState::MAYBE_BEHIND) {
spdlog::warn("Connection failed when registering replica {}. Replica will still be registered.",
instance_client.name_);
}
storage_clients.push_back(std::move(client));
return {};
});
if (ret.HasError()) {
MG_ASSERT(query::RegisterReplicaError::CONNECTION_FAILED != ret.GetError());
LOG_FATAL("Failure when restoring replica {}: {}.", instance_client.name_,
RegisterReplicaErrorToString(ret.GetError()));
}
spdlog::info("Replica {} restored for {}.", instance_client.name_, db_acc->name());
}
spdlog::info("Replication role restored to MAIN.");
}
} // namespace
#ifdef MG_ENTERPRISE
struct Durability {
enum class DurabilityVersion : uint8_t {
V0 = 0,
@@ -112,11 +158,9 @@ struct Durability {
}
};
DbmsHandler::DbmsHandler(
storage::Config config,
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
bool recovery_on_startup)
: default_config_{std::move(config)}, repl_state_{ReplicationStateRootPath(default_config_)} {
DbmsHandler::DbmsHandler(storage::Config config, memgraph::system::System &system,
replication::ReplicationState &repl_state, auth::SynchedAuth &auth, bool recovery_on_startup)
: default_config_{std::move(config)}, auth_{auth}, repl_state_{repl_state}, system_{&system} {
// TODO: Decouple storage config from dbms config
// TODO: Save individual db configs inside the kvstore and restore from there
@@ -150,19 +194,13 @@ DbmsHandler::DbmsHandler(
const auto uuid = json.at("uuid").get<utils::UUID>();
const auto rel_dir = json.at("rel_dir").get<std::filesystem::path>();
spdlog::info("Restoring database {} at {}.", name, rel_dir);
auto new_db = New_(name, uuid, rel_dir);
auto new_db = New_(name, uuid, nullptr, rel_dir);
MG_ASSERT(!new_db.HasError(), "Failed while creating database {}.", name);
directories.emplace(rel_dir.filename());
spdlog::info("Database {} restored.", name);
}
// Read the last timestamp
auto lcst = durability_->Get(kLastCommitedSystemTsKey);
if (lcst) {
last_commited_system_timestamp_ = std::stoul(*lcst);
system_timestamp_ = last_commited_system_timestamp_;
}
} else { // Clear databases from the durability list and auth
auto locked_auth = auth->Lock();
auto locked_auth = auth_.Lock();
auto it = durability_->begin(std::string{kDBPrefix});
auto end = durability_->end(std::string{kDBPrefix});
for (; it != end; ++it) {
@@ -172,8 +210,6 @@ DbmsHandler::DbmsHandler(
locked_auth->DeleteDatabase(name);
durability_->Delete(key);
}
// Delete the last timestamp
durability_->Delete(kLastCommitedSystemTsKey);
}
/*
@@ -198,45 +234,31 @@ DbmsHandler::DbmsHandler(
*/
// Setup the default DB
SetupDefault_();
/*
* REPLICATION RECOVERY AND STARTUP
*/
// Startup replication state (if recovered at startup)
auto replica = [this](replication::RoleReplicaData const &data) { return StartRpcServer(*this, data); };
// Replication recovery and frequent check start
auto main = [this](replication::RoleMainData &data) {
for (auto &client : data.registered_replicas_) {
SystemRestore(client);
}
ForEach([this](DatabaseAccess db) { RecoverReplication(db); });
for (auto &client : data.registered_replicas_) {
StartReplicaClient(*this, client);
}
return true;
};
// Startup proccess for main/replica
MG_ASSERT(std::visit(memgraph::utils::Overloaded{replica, main}, repl_state_.ReplicationData()),
"Replica recovery failure!");
// Warning
if (default_config_.durability.snapshot_wal_mode == storage::Config::Durability::SnapshotWalMode::DISABLED &&
repl_state_.IsMain()) {
spdlog::warn(
"The instance has the MAIN replication role, but durability logs and snapshots are disabled. Please "
"consider "
"enabling durability by using --storage-snapshot-interval-sec and --storage-wal-enabled flags because "
"without write-ahead logs this instance is not replicating any data.");
}
// MAIN or REPLICA instance
if (FLAGS_coordinator_server_port) {
CoordinatorHandlers::Register(*this);
MG_ASSERT(coordinator_state_.GetCoordinatorServer().Start(), "Failed to start coordinator server!");
}
}
DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name) {
struct DropDatabase : memgraph::system::ISystemAction {
explicit DropDatabase(utils::UUID uuid) : uuid_{uuid} {}
void DoDurability() override { /* Done during DBMS execution */
}
bool DoReplication(replication::ReplicationClient &client, const utils::UUID &main_uuid,
replication::ReplicationEpoch const &epoch,
memgraph::system::Transaction const &txn) const override {
auto check_response = [](const storage::replication::DropDatabaseRes &response) {
return response.result != storage::replication::DropDatabaseRes::Result::FAILURE;
};
return client.SteamAndFinalizeDelta<storage::replication::DropDatabaseRpc>(
check_response, main_uuid, std::string(epoch.id()), txn.last_committed_system_timestamp(), txn.timestamp(),
uuid_);
}
void PostReplication(replication::RoleMainData &mainData) const override {}
private:
utils::UUID uuid_;
};
DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name, system::Transaction *transaction) {
std::lock_guard<LockT> wr(lock_);
if (db_name == kDefaultDB) {
// MSG cannot delete the default db
@@ -273,9 +295,10 @@ DbmsHandler::DeleteResult DbmsHandler::TryDelete(std::string_view db_name) {
// Success
// Save delta
if (system_transaction_) {
system_transaction_->delta.emplace(SystemTransaction::Delta::drop_database, uuid);
if (transaction) {
transaction->AddAction<DropDatabase>(uuid);
}
return {};
}
@@ -296,18 +319,50 @@ DbmsHandler::DeleteResult DbmsHandler::Delete(utils::UUID uuid) {
return Delete_(db_name);
}
DbmsHandler::NewResultT DbmsHandler::New_(storage::Config storage_config) {
struct CreateDatabase : memgraph::system::ISystemAction {
explicit CreateDatabase(storage::SalientConfig config, DatabaseAccess db_acc)
: config_{std::move(config)}, db_acc(db_acc) {}
void DoDurability() override {
// Done during dbms execution
}
bool DoReplication(replication::ReplicationClient &client, const utils::UUID &main_uuid,
replication::ReplicationEpoch const &epoch,
memgraph::system::Transaction const &txn) const override {
auto check_response = [](const storage::replication::CreateDatabaseRes &response) {
return response.result != storage::replication::CreateDatabaseRes::Result::FAILURE;
};
return client.SteamAndFinalizeDelta<storage::replication::CreateDatabaseRpc>(
check_response, main_uuid, std::string(epoch.id()), txn.last_committed_system_timestamp(), txn.timestamp(),
config_);
}
void PostReplication(replication::RoleMainData &mainData) const override {
// Sync database with REPLICAs
// NOTE: The function bellow is used to create ReplicationStorageClient, so it must be called on a new storage
// We don't need to have it here, since the function won't fail even if the replication client fails to
// connect We will just have everything ready, for recovery at some point.
dbms::DbmsHandler::RecoverStorageReplication(db_acc, mainData);
}
private:
storage::SalientConfig config_;
DatabaseAccess db_acc;
};
DbmsHandler::NewResultT DbmsHandler::New_(storage::Config storage_config, system::Transaction *txn) {
auto new_db = db_handler_.New(storage_config, repl_state_);
if (new_db.HasValue()) { // Success
// Save delta
if (system_transaction_) {
system_transaction_->delta.emplace(SystemTransaction::Delta::create_database, storage_config.salient);
}
UpdateDurability(storage_config);
return new_db.GetValue();
if (txn) {
txn->AddAction<CreateDatabase>(storage_config.salient, new_db.GetValue());
}
}
return new_db.GetError();
return new_db;
}
DbmsHandler::DeleteResult DbmsHandler::Delete_(std::string_view db_name) {
@@ -361,89 +416,16 @@ void DbmsHandler::UpdateDurability(const storage::Config &config, std::optional<
durability_->Put(key, val);
}
AllSyncReplicaStatus DbmsHandler::Commit() {
if (system_transaction_ == std::nullopt || system_transaction_->delta == std::nullopt)
return AllSyncReplicaStatus::AllCommitsConfirmed; // Nothing to commit
const auto &delta = *system_transaction_->delta;
auto sync_status = AllSyncReplicaStatus::AllCommitsConfirmed;
// TODO Create a system client that can handle all of this automatically
switch (delta.action) {
using enum SystemTransaction::Delta::Action;
case CREATE_DATABASE: {
// Replication
auto main_handler = [&](memgraph::replication::RoleMainData &main_data) {
// TODO: data race issue? registered_replicas_ access not protected
// This is sync in any case, as this is the startup
for (auto &client : main_data.registered_replicas_) {
bool completed = SteamAndFinalizeDelta<storage::replication::CreateDatabaseRpc>(
client,
[](const storage::replication::CreateDatabaseRes &response) {
return response.result != storage::replication::CreateDatabaseRes::Result::FAILURE;
},
std::string(main_data.epoch_.id()), last_commited_system_timestamp_,
system_transaction_->system_timestamp, delta.config);
// TODO: reduce duplicate code
if (!completed && client.mode_ == replication_coordination_glue::ReplicationMode::SYNC) {
sync_status = AllSyncReplicaStatus::SomeCommitsUnconfirmed;
}
}
// Sync database with REPLICAs
RecoverReplication(Get_(delta.config.name));
};
auto replica_handler = [](memgraph::replication::RoleReplicaData &) { /* Nothing to do */ };
std::visit(utils::Overloaded{main_handler, replica_handler}, repl_state_.ReplicationData());
} break;
case DROP_DATABASE: {
// Replication
auto main_handler = [&](memgraph::replication::RoleMainData &main_data) {
// TODO: data race issue? registered_replicas_ access not protected
// This is sync in any case, as this is the startup
for (auto &client : main_data.registered_replicas_) {
bool completed = SteamAndFinalizeDelta<storage::replication::DropDatabaseRpc>(
client,
[](const storage::replication::DropDatabaseRes &response) {
return response.result != storage::replication::DropDatabaseRes::Result::FAILURE;
},
std::string(main_data.epoch_.id()), last_commited_system_timestamp_,
system_transaction_->system_timestamp, delta.uuid);
// TODO: reduce duplicate code
if (!completed && client.mode_ == replication_coordination_glue::ReplicationMode::SYNC) {
sync_status = AllSyncReplicaStatus::SomeCommitsUnconfirmed;
}
}
};
auto replica_handler = [](memgraph::replication::RoleReplicaData &) { /* Nothing to do */ };
std::visit(utils::Overloaded{main_handler, replica_handler}, repl_state_.ReplicationData());
} break;
}
durability_->Put(kLastCommitedSystemTsKey, std::to_string(system_transaction_->system_timestamp));
last_commited_system_timestamp_ = system_transaction_->system_timestamp;
ResetSystemTransaction();
return sync_status;
}
#else // not MG_ENTERPRISE
AllSyncReplicaStatus DbmsHandler::Commit() {
if (system_transaction_ == std::nullopt || system_transaction_->delta == std::nullopt) {
return AllSyncReplicaStatus::AllCommitsConfirmed; // Nothing to commit
}
const auto &delta = *system_transaction_->delta;
switch (delta.action) {
using enum SystemTransaction::Delta::Action;
case CREATE_DATABASE:
case DROP_DATABASE:
/* Community edition doesn't support multi-tenant replication */
break;
}
last_commited_system_timestamp_ = system_transaction_->system_timestamp;
ResetSystemTransaction();
return AllSyncReplicaStatus::AllCommitsConfirmed;
}
#endif
void DbmsHandler::RecoverStorageReplication(DatabaseAccess db_acc, replication::RoleMainData &role_main_data) {
if (allow_mt_repl || db_acc->name() == dbms::kDefaultDB) {
// Handle global replication state
spdlog::info("Replication configuration will be stored and will be automatically restored in case of a crash.");
// RECOVER REPLICA CONNECTIONS
memgraph::dbms::RestoreReplication(role_main_data, db_acc);
} else if (!role_main_data.registered_replicas_.empty()) {
spdlog::warn("Multi-tenant replication is currently not supported!");
}
}
} // namespace memgraph::dbms

View File

@@ -25,24 +25,25 @@
#include "constants.hpp"
#include "dbms/database.hpp"
#include "dbms/inmemory/replication_handlers.hpp"
#include "dbms/replication_handler.hpp"
#include "dbms/rpc.hpp"
#include "kvstore/kvstore.hpp"
#include "license/license.hpp"
#include "replication/replication_client.hpp"
#include "replication_coordination_glue/handler.hpp"
#include "storage/v2/config.hpp"
#include "storage/v2/replication/enums.hpp"
#include "storage/v2/replication/rpc.hpp"
#include "storage/v2/transaction.hpp"
#include "system/system.hpp"
#include "utils/thread_pool.hpp"
#ifdef MG_ENTERPRISE
#include "coordination/coordinator_state.hpp"
#include "dbms/database_handler.hpp"
#endif
#include "dbms/transaction.hpp"
#include "global.hpp"
#include "query/config.hpp"
#include "query/interpreter_context.hpp"
#include "spdlog/spdlog.h"
#include "storage/v2/isolation_level.hpp"
#include "system/system.hpp"
#include "utils/logging.hpp"
#include "utils/result.hpp"
#include "utils/rw_lock.hpp"
@@ -51,11 +52,6 @@
namespace memgraph::dbms {
enum class AllSyncReplicaStatus {
AllCommitsConfirmed,
SomeCommitsUnconfirmed,
};
struct Statistics {
uint64_t num_vertex; //!< Sum of vertexes in every database
uint64_t num_edges; //!< Sum of edges in every database
@@ -111,8 +107,8 @@ class DbmsHandler {
* @param auth pointer to the global authenticator
* @param recovery_on_startup restore databases (and its content) and authentication data
*/
DbmsHandler(storage::Config config,
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
DbmsHandler(storage::Config config, memgraph::system::System &system, replication::ReplicationState &repl_state,
auth::SynchedAuth &auth,
bool recovery_on_startup); // TODO If more arguments are added use a config struct
#else
/**
@@ -120,15 +116,14 @@ class DbmsHandler {
*
* @param configs storage configuration
*/
DbmsHandler(storage::Config config)
: repl_state_{ReplicationStateRootPath(config)},
DbmsHandler(storage::Config config, memgraph::system::System &system, replication::ReplicationState &repl_state)
: repl_state_{repl_state},
system_{&system},
db_gatekeeper_{[&] {
config.salient.name = kDefaultDB;
return std::move(config);
}(),
repl_state_} {
RecoverReplication(Get());
}
repl_state_} {}
#endif
#ifdef MG_ENTERPRISE
@@ -138,10 +133,10 @@ class DbmsHandler {
* @param name name of the database
* @return NewResultT context on success, error on failure
*/
NewResultT New(const std::string &name) {
NewResultT New(const std::string &name, system::Transaction *txn = nullptr) {
std::lock_guard<LockT> wr(lock_);
const auto uuid = utils::UUID{};
return New_(name, uuid);
return New_(name, uuid, txn);
}
/**
@@ -234,7 +229,7 @@ class DbmsHandler {
* @param db_name database name
* @return DeleteResult error on failure
*/
DeleteResult TryDelete(std::string_view db_name);
DeleteResult TryDelete(std::string_view db_name, system::Transaction *transaction = nullptr);
/**
* @brief Delete or defer deletion of database.
@@ -274,7 +269,7 @@ class DbmsHandler {
bool IsReplica() const { return repl_state_.IsReplica(); }
#ifdef MG_ENTERPRISE
coordination::CoordinatorState &CoordinatorState() { return coordinator_state_; }
// coordination::CoordinatorState &CoordinatorState() { return coordinator_state_; }
#endif
/**
@@ -282,8 +277,7 @@ class DbmsHandler {
*
* @return Statistics
*/
Statistics Stats() {
auto const replication_role = repl_state_.GetRole();
Statistics Stats(memgraph::replication_coordination_glue::ReplicationRole replication_role) {
Statistics stats{};
// TODO: Handle overflow?
#ifdef MG_ENTERPRISE
@@ -319,8 +313,7 @@ class DbmsHandler {
*
* @return std::vector<DatabaseInfo>
*/
std::vector<DatabaseInfo> Info() {
auto const replication_role = repl_state_.GetRole();
std::vector<DatabaseInfo> Info(memgraph::replication_coordination_glue::ReplicationRole replication_role) {
std::vector<DatabaseInfo> res;
#ifdef MG_ENTERPRISE
std::shared_lock<LockT> rd(lock_);
@@ -407,98 +400,17 @@ class DbmsHandler {
}
}
void NewSystemTransaction() {
DMG_ASSERT(!system_transaction_, "Already running a system transaction");
system_transaction_.emplace(++system_timestamp_);
}
void ResetSystemTransaction() { system_transaction_.reset(); }
//! \tparam RPC An rpc::RequestResponse
//! \tparam Args the args type
//! \param client the client to use for rpc communication
//! \param check predicate to check response is ok
//! \param args arguments to forward to the rpc request
//! \return If replica stream is completed or enqueued
template <typename RPC, typename... Args>
bool SteamAndFinalizeDelta(auto &client, auto &&check, Args &&...args) {
try {
auto stream = client.rpc_client_.template Stream<RPC>(std::forward<Args>(args)...);
auto task = [&client, check = std::forward<decltype(check)>(check), stream = std::move(stream)]() mutable {
if (stream.IsDefunct()) {
client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
return false;
}
try {
if (check(stream.AwaitResponse())) {
return true;
}
} catch (memgraph::rpc::GenericRpcFailedException const &e) {
// swallow error, fallthrough to error handling
}
// This replica needs SYSTEM recovery
client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
return false;
};
if (client.mode_ == memgraph::replication_coordination_glue::ReplicationMode::ASYNC) {
client.thread_pool_.AddTask([task = utils::CopyMovableFunctionWrapper{std::move(task)}]() mutable { task(); });
return true;
}
return task();
} catch (memgraph::rpc::GenericRpcFailedException const &e) {
// This replica needs SYSTEM recovery
client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
return false;
}
};
AllSyncReplicaStatus Commit();
auto LastCommitedTS() const -> uint64_t { return last_commited_system_timestamp_; }
void SetLastCommitedTS(uint64_t new_ts) { last_commited_system_timestamp_.store(new_ts); }
static void RecoverStorageReplication(DatabaseAccess db_acc, replication::RoleMainData &role_main_data);
auto default_config() const -> storage::Config const & {
#ifdef MG_ENTERPRISE
// When being called by intepreter no need to gain lock, it should already be under a system transaction
// But concurrently the FrequentCheck is running and will need to lock before reading last_commited_system_timestamp_
template <bool REQUIRE_LOCK = false>
void SystemRestore(replication::ReplicationClient &client) {
// Check if system is up to date
if (client.state_.WithLock(
[](auto &state) { return state == memgraph::replication::ReplicationClient::State::READY; }))
return;
// Try to recover...
{
auto [database_configs, last_commited_system_timestamp] = std::invoke([&] {
auto sys_guard =
std::unique_lock{system_lock_, std::defer_lock}; // ensure no other system transaction in progress
if constexpr (REQUIRE_LOCK) {
sys_guard.lock();
}
auto configs = std::vector<storage::SalientConfig>{};
ForEach([&configs](DatabaseAccess acc) { configs.emplace_back(acc->config().salient); });
return std::pair{configs, last_commited_system_timestamp_.load()};
});
try {
auto stream = client.rpc_client_.Stream<storage::replication::SystemRecoveryRpc>(last_commited_system_timestamp,
std::move(database_configs));
const auto response = stream.AwaitResponse();
if (response.result == storage::replication::SystemRecoveryRes::Result::FAILURE) {
client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
return;
}
} catch (memgraph::rpc::GenericRpcFailedException const &e) {
client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::BEHIND; });
return;
}
}
// Successfully recovered
client.state_.WithLock([](auto &state) { state = memgraph::replication::ReplicationClient::State::READY; });
}
return default_config_;
#else
const auto acc = db_gatekeeper_.access();
MG_ASSERT(acc, "Failed to get default database!");
return acc->get()->config();
#endif
}
private:
#ifdef MG_ENTERPRISE
@@ -524,7 +436,8 @@ class DbmsHandler {
* @param uuid undelying RocksDB directory
* @return NewResultT context on success, error on failure
*/
NewResultT New_(std::string_view name, utils::UUID uuid, std::optional<std::filesystem::path> rel_dir = {}) {
NewResultT New_(std::string_view name, utils::UUID uuid, system::Transaction *txn = nullptr,
std::optional<std::filesystem::path> rel_dir = {}) {
auto config_copy = default_config_;
config_copy.salient.name = name;
config_copy.salient.uuid = uuid;
@@ -535,7 +448,7 @@ class DbmsHandler {
storage::UpdatePaths(config_copy,
default_config_.durability.storage_directory / kMultiTenantDir / std::string{uuid});
}
return New_(std::move(config_copy));
return New_(std::move(config_copy), txn);
}
/**
@@ -544,11 +457,11 @@ class DbmsHandler {
* @param config configuration to be used
* @return NewResultT context on success, error on failure
*/
NewResultT New_(const storage::SalientConfig &config) {
NewResultT New_(const storage::SalientConfig &config, system::Transaction *txn = nullptr) {
auto config_copy = default_config_;
config_copy.salient = config; // name, uuid, mode, etc
UpdatePaths(config_copy, config_copy.durability.storage_directory / kMultiTenantDir / std::string{config.uuid});
return New_(std::move(config_copy));
return New_(std::move(config_copy), txn);
}
/**
@@ -557,7 +470,7 @@ class DbmsHandler {
* @param storage_config storage configuration
* @return NewResultT context on success, error on failure
*/
NewResultT New_(storage::Config storage_config);
DbmsHandler::NewResultT New_(storage::Config storage_config, system::Transaction *txn = nullptr);
// TODO: new overload of Delete_ with DatabaseAccess
DeleteResult Delete_(std::string_view db_name);
@@ -572,7 +485,8 @@ class DbmsHandler {
Get(kDefaultDB);
} catch (const UnknownDatabaseException &) {
// No default DB restored, create it
MG_ASSERT(New_(kDefaultDB, {/* random UUID */}, ".").HasValue(), "Failed while creating the default database");
MG_ASSERT(New_(kDefaultDB, {/* random UUID */}, nullptr, ".").HasValue(),
"Failed while creating the default database");
}
// For back-compatibility...
@@ -659,35 +573,24 @@ class DbmsHandler {
}
#endif
void RecoverReplication(DatabaseAccess db_acc) {
if (allow_mt_repl || db_acc->name() == dbms::kDefaultDB) {
// Handle global replication state
spdlog::info("Replication configuration will be stored and will be automatically restored in case of a crash.");
// RECOVER REPLICA CONNECTIONS
memgraph::dbms::RestoreReplication(repl_state_, std::move(db_acc));
} else if (const ::memgraph::replication::RoleMainData *data =
std::get_if<::memgraph::replication::RoleMainData>(&repl_state_.ReplicationData());
data && !data->registered_replicas_.empty()) {
spdlog::warn("Multi-tenant replication is currently not supported!");
}
}
#ifdef MG_ENTERPRISE
mutable LockT lock_{utils::RWLock::Priority::READ}; //!< protective lock
storage::Config default_config_; //!< Storage configuration used when creating new databases
DatabaseHandler db_handler_; //!< multi-tenancy storage handler
std::unique_ptr<kvstore::KVStore> durability_; //!< list of active dbs (pointer so we can postpone its creation)
coordination::CoordinatorState coordinator_state_; //!< Replication coordinator
// TODO: move to be common
std::unique_ptr<kvstore::KVStore> durability_; //!< list of active dbs (pointer so we can postpone its creation)
auth::SynchedAuth &auth_; //!< Synchronized auth::Auth
#endif
// TODO: Make an api
public:
utils::ResourceLock system_lock_{}; //!> Ensure exclusive access for system queries
private:
std::optional<SystemTransaction> system_transaction_; //!< Current system transaction (only one at a time)
uint64_t system_timestamp_{storage::kTimestampInitialId}; //!< System timestamp
std::atomic_uint64_t last_commited_system_timestamp_{
storage::kTimestampInitialId}; //!< Last commited system timestamp
replication::ReplicationState repl_state_; //!< Global replication state
// NOTE: atm the only reason this exists here, is because we pass it into the construction of New Database's
// Database only uses it as a convience to make the correct Access without out needing to be told the
// current replication role. TODO: make Database Access explicit about the role and remove this from
// dbms stuff
replication::ReplicationState &repl_state_; //!< Ref to global replication state
public:
// TODO fix to be non public/remove from dbms....maybe
system::System *system_;
#ifndef MG_ENTERPRISE
mutable utils::Gatekeeper<Database> db_gatekeeper_; //!< Single databases gatekeeper
#endif

View File

@@ -23,7 +23,7 @@
#include "storage/v2/inmemory/storage.hpp"
#include "storage/v2/inmemory/unique_constraints.hpp"
using memgraph::replication::ReplicationRole;
using memgraph::replication_coordination_glue::ReplicationRole;
using memgraph::storage::Delta;
using memgraph::storage::EdgeAccessor;
using memgraph::storage::EdgeRef;
@@ -76,47 +76,84 @@ std::optional<DatabaseAccess> GetDatabaseAccessor(dbms::DbmsHandler *dbms_handle
return std::nullopt;
}
}
void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
std::string_view rpc_req) {
spdlog::error("Received {} with main_id: {} != current_main_uuid: {}", rpc_req, std::string(main_req_id),
current_main_uuid.has_value() ? std::string(current_main_uuid.value()) : "");
}
} // namespace
void InMemoryReplicationHandlers::Register(dbms::DbmsHandler *dbms_handler, replication::ReplicationServer &server) {
server.rpc_server_.Register<storage::replication::HeartbeatRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received HeartbeatRpc");
InMemoryReplicationHandlers::HeartbeatHandler(dbms_handler, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::AppendDeltasRpc>(
[dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received AppendDeltasRpc");
InMemoryReplicationHandlers::AppendDeltasHandler(dbms_handler, req_reader, res_builder);
void InMemoryReplicationHandlers::Register(dbms::DbmsHandler *dbms_handler, replication::RoleReplicaData &data) {
auto &server = *data.server;
server.rpc_server_.Register<storage::replication::HeartbeatRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received HeartbeatRpc");
InMemoryReplicationHandlers::HeartbeatHandler(dbms_handler, data.uuid_, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::AppendDeltasRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received AppendDeltasRpc");
InMemoryReplicationHandlers::AppendDeltasHandler(dbms_handler, data.uuid_, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::SnapshotRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received SnapshotRpc");
InMemoryReplicationHandlers::SnapshotHandler(dbms_handler, data.uuid_, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::WalFilesRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received WalFilesRpc");
InMemoryReplicationHandlers::WalFilesHandler(dbms_handler, data.uuid_, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::CurrentWalRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received CurrentWalRpc");
InMemoryReplicationHandlers::CurrentWalHandler(dbms_handler, data.uuid_, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::TimestampRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received TimestampRpc");
InMemoryReplicationHandlers::TimestampHandler(dbms_handler, data.uuid_, req_reader, res_builder);
});
server.rpc_server_.Register<replication_coordination_glue::SwapMainUUIDRpc>(
[&data, dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received SwapMainUUIDHandler");
InMemoryReplicationHandlers::SwapMainUUIDHandler(dbms_handler, data, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::SnapshotRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received SnapshotRpc");
InMemoryReplicationHandlers::SnapshotHandler(dbms_handler, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::WalFilesRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received WalFilesRpc");
InMemoryReplicationHandlers::WalFilesHandler(dbms_handler, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::CurrentWalRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received CurrentWalRpc");
InMemoryReplicationHandlers::CurrentWalHandler(dbms_handler, req_reader, res_builder);
});
server.rpc_server_.Register<storage::replication::TimestampRpc>([dbms_handler](auto *req_reader, auto *res_builder) {
spdlog::debug("Received TimestampRpc");
InMemoryReplicationHandlers::TimestampHandler(dbms_handler, req_reader, res_builder);
});
}
void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder) {
void InMemoryReplicationHandlers::SwapMainUUIDHandler(dbms::DbmsHandler *dbms_handler,
replication::RoleReplicaData &role_replica_data,
slk::Reader *req_reader, slk::Builder *res_builder) {
if (!dbms_handler->IsReplica()) {
spdlog::error("Setting main uuid must be performed on replica.");
slk::Save(replication_coordination_glue::SwapMainUUIDRes{false}, res_builder);
return;
}
replication_coordination_glue::SwapMainUUIDReq req;
slk::Load(&req, req_reader);
spdlog::info(fmt::format("Set replica data UUID to main uuid {}", std::string(req.uuid)));
dbms_handler->ReplicationState().TryPersistRoleReplica(role_replica_data.config, req.uuid);
role_replica_data.uuid_ = req.uuid;
slk::Save(replication_coordination_glue::SwapMainUUIDRes{true}, res_builder);
}
void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
storage::replication::HeartbeatReq req;
slk::Load(&req, req_reader);
auto const db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
if (!db_acc) {
if (!current_main_uuid.has_value() || req.main_uuid != *current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::HeartbeatReq::kType.name);
storage::replication::HeartbeatRes res{false, 0, ""};
slk::Save(res, res_builder);
return;
}
// TODO: this handler is agnostic of InMemory, move to be reused by on-disk
auto const *storage = db_acc->get()->storage();
storage::replication::HeartbeatRes res{true, storage->repl_storage_state_.last_commit_timestamp_.load(),
@@ -124,10 +161,19 @@ void InMemoryReplicationHandlers::HeartbeatHandler(dbms::DbmsHandler *dbms_handl
slk::Save(res, res_builder);
}
void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder) {
void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
storage::replication::AppendDeltasReq req;
slk::Load(&req, req_reader);
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::AppendDeltasReq::kType.name);
storage::replication::AppendDeltasRes res{false, 0};
slk::Save(res, res_builder);
return;
}
auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
if (!db_acc) {
storage::replication::AppendDeltasRes res{false, 0};
@@ -187,8 +233,9 @@ void InMemoryReplicationHandlers::AppendDeltasHandler(dbms::DbmsHandler *dbms_ha
spdlog::debug("Replication recovery from append deltas finished, replica is now up to date!");
}
void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder) {
void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
storage::replication::SnapshotReq req;
slk::Load(&req, req_reader);
auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -197,6 +244,12 @@ void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handle
slk::Save(res, res_builder);
return;
}
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::SnapshotReq::kType.name);
storage::replication::SnapshotRes res{false, 0};
slk::Save(res, res_builder);
return;
}
storage::replication::Decoder decoder(req_reader);
@@ -270,8 +323,9 @@ void InMemoryReplicationHandlers::SnapshotHandler(dbms::DbmsHandler *dbms_handle
spdlog::debug("Replication recovery from snapshot finished!");
}
void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder) {
void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
storage::replication::WalFilesReq req;
slk::Load(&req, req_reader);
auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -280,6 +334,12 @@ void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handle
slk::Save(res, res_builder);
return;
}
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::WalFilesReq::kType.name);
storage::replication::WalFilesRes res{false, 0};
slk::Save(res, res_builder);
return;
}
const auto wal_file_number = req.file_number;
spdlog::debug("Received WAL files: {}", wal_file_number);
@@ -298,8 +358,9 @@ void InMemoryReplicationHandlers::WalFilesHandler(dbms::DbmsHandler *dbms_handle
spdlog::debug("Replication recovery from WAL files ended successfully, replica is now up to date!");
}
void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder) {
void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
storage::replication::CurrentWalReq req;
slk::Load(&req, req_reader);
auto db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -309,6 +370,13 @@ void InMemoryReplicationHandlers::CurrentWalHandler(dbms::DbmsHandler *dbms_hand
return;
}
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::CurrentWalReq::kType.name);
storage::replication::CurrentWalRes res{false, 0};
slk::Save(res, res_builder);
return;
}
storage::replication::Decoder decoder(req_reader);
auto *storage = static_cast<storage::InMemoryStorage *>(db_acc->get()->storage());
@@ -370,8 +438,9 @@ void InMemoryReplicationHandlers::LoadWal(storage::InMemoryStorage *storage, sto
}
}
void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handler, slk::Reader *req_reader,
slk::Builder *res_builder) {
void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handler,
const std::optional<utils::UUID> &current_main_uuid,
slk::Reader *req_reader, slk::Builder *res_builder) {
storage::replication::TimestampReq req;
slk::Load(&req, req_reader);
auto const db_acc = GetDatabaseAccessor(dbms_handler, req.uuid);
@@ -381,12 +450,20 @@ void InMemoryReplicationHandlers::TimestampHandler(dbms::DbmsHandler *dbms_handl
return;
}
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, storage::replication::TimestampReq::kType.name);
storage::replication::CurrentWalRes res{false, 0};
slk::Save(res, res_builder);
return;
}
// TODO: this handler is agnostic of InMemory, move to be reused by on-disk
auto const *storage = db_acc->get()->storage();
storage::replication::TimestampRes res{true, storage->repl_storage_state_.last_commit_timestamp_.load()};
slk::Save(res, res_builder);
}
/////// AF how does this work, does it get all deltas at once or what?
uint64_t InMemoryReplicationHandlers::ReadAndApplyDelta(storage::InMemoryStorage *storage,
storage::durability::BaseDecoder *decoder,
const uint64_t version) {

View File

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

View File

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

View File

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

View File

@@ -1,400 +0,0 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "dbms/replication_handler.hpp"
#include <algorithm>
#include "dbms/constants.hpp"
#include "dbms/dbms_handler.hpp"
#include "dbms/global.hpp"
#include "dbms/inmemory/replication_handlers.hpp"
#include "dbms/replication_client.hpp"
#include "dbms/utils.hpp"
#include "replication/messages.hpp"
#include "replication/state.hpp"
#include "spdlog/spdlog.h"
#include "storage/v2/config.hpp"
#include "storage/v2/replication/rpc.hpp"
#include "utils/on_scope_exit.hpp"
using memgraph::replication::RoleMainData;
using memgraph::replication::RoleReplicaData;
namespace memgraph::dbms {
namespace {
std::string RegisterReplicaErrorToString(RegisterReplicaError error) {
switch (error) {
using enum RegisterReplicaError;
case NAME_EXISTS:
return "NAME_EXISTS";
case END_POINT_EXISTS:
return "END_POINT_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 instance_client = dbms_handler_.ReplicationState().RegisterReplica(config);
if (instance_client.HasError()) {
switch (instance_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::END_POINT_EXISTS:
return memgraph::dbms::RegisterReplicaError::END_POINT_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(*instance_client.GetValue());
#endif
const auto dbms_error = memgraph::dbms::HandleErrorOnReplicaClient(instance_client);
if (dbms_error.has_value()) {
return *dbms_error;
}
auto &instance_client_ptr = instance_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.GetValue());
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::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

@@ -1,82 +0,0 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#include "dbms/database.hpp"
#include "replication/role.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, END_POINT_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::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

@@ -0,0 +1,205 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#include "dbms/replication_handlers.hpp"
#include "dbms/database.hpp"
#include "dbms/dbms_handler.hpp"
#include "storage/v2/storage.hpp"
#include "system/state.hpp"
namespace memgraph::dbms {
#ifdef MG_ENTERPRISE
void CreateDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, DbmsHandler &dbms_handler,
slk::Reader *req_reader, slk::Builder *res_builder) {
using memgraph::storage::replication::CreateDatabaseRes;
CreateDatabaseRes res(CreateDatabaseRes::Result::FAILURE);
// Ignore if no license
if (!license::global_license_checker.IsEnterpriseValidFast()) {
spdlog::error("Handling CreateDatabase, an enterprise RPC message, without license.");
memgraph::slk::Save(res, res_builder);
return;
}
memgraph::storage::replication::CreateDatabaseReq req;
memgraph::slk::Load(&req, req_reader);
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, memgraph::storage::replication::CreateDatabaseReq::kType.name);
memgraph::slk::Save(res, res_builder);
return;
}
// Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
// of the set of databases. Hence no history exists to maintain regarding epoch change.
// If MAIN has changed we need to check this new group_timestamp is consistent with
// what we have so far.
if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
spdlog::debug("CreateDatabaseHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
system_state_access.LastCommitedTS());
memgraph::slk::Save(res, res_builder);
return;
}
try {
// Create new
auto new_db = dbms_handler.Update(req.config);
if (new_db.HasValue()) {
// Successfully create db
system_state_access.SetLastCommitedTS(req.new_group_timestamp);
res = CreateDatabaseRes(CreateDatabaseRes::Result::SUCCESS);
spdlog::debug("CreateDatabaseHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
}
} catch (...) {
// Failure
}
memgraph::slk::Save(res, res_builder);
}
void DropDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, DbmsHandler &dbms_handler,
slk::Reader *req_reader, slk::Builder *res_builder) {
using memgraph::storage::replication::DropDatabaseRes;
DropDatabaseRes res(DropDatabaseRes::Result::FAILURE);
// Ignore if no license
if (!license::global_license_checker.IsEnterpriseValidFast()) {
spdlog::error("Handling DropDatabase, an enterprise RPC message, without license.");
memgraph::slk::Save(res, res_builder);
return;
}
memgraph::storage::replication::DropDatabaseReq req;
memgraph::slk::Load(&req, req_reader);
if (!current_main_uuid.has_value() || req.main_uuid != current_main_uuid) [[unlikely]] {
LogWrongMain(current_main_uuid, req.main_uuid, memgraph::storage::replication::DropDatabaseReq::kType.name);
memgraph::slk::Save(res, res_builder);
return;
}
// Note: No need to check epoch, recovery mechanism is done by a full uptodate snapshot
// of the set of databases. Hence no history exists to maintain regarding epoch change.
// If MAIN has changed we need to check this new group_timestamp is consistent with
// what we have so far.
if (req.expected_group_timestamp != system_state_access.LastCommitedTS()) {
spdlog::debug("DropDatabaseHandler: bad expected timestamp {},{}", req.expected_group_timestamp,
system_state_access.LastCommitedTS());
memgraph::slk::Save(res, res_builder);
return;
}
try {
// NOTE: Single communication channel can exist at a time, no other database can be deleted/created at the moment.
auto new_db = dbms_handler.Delete(req.uuid);
if (new_db.HasError()) {
if (new_db.GetError() == DeleteError::NON_EXISTENT) {
// Nothing to drop
system_state_access.SetLastCommitedTS(req.new_group_timestamp);
res = DropDatabaseRes(DropDatabaseRes::Result::NO_NEED);
}
} else {
// Successfully drop db
system_state_access.SetLastCommitedTS(req.new_group_timestamp);
res = DropDatabaseRes(DropDatabaseRes::Result::SUCCESS);
spdlog::debug("DropDatabaseHandler: SUCCESS updated LCTS to {}", req.new_group_timestamp);
}
} catch (...) {
// Failure
}
memgraph::slk::Save(res, res_builder);
}
bool SystemRecoveryHandler(DbmsHandler &dbms_handler, const std::vector<storage::SalientConfig> &database_configs) {
/*
* NO LICENSE
*/
if (!license::global_license_checker.IsEnterpriseValidFast()) {
spdlog::error("Handling SystemRecovery, an enterprise RPC message, without license.");
for (const auto &config : database_configs) {
// Only handle default DB
if (config.name != kDefaultDB) continue;
try {
if (dbms_handler.Update(config).HasError()) {
return false;
}
} catch (const UnknownDatabaseException &) {
return false;
}
}
return true;
}
/*
* MULTI-TENANCY
*/
// Get all current dbs
auto old = dbms_handler.All();
// Check/create the incoming dbs
for (const auto &config : database_configs) {
// Missing db
try {
if (dbms_handler.Update(config).HasError()) {
spdlog::debug("SystemRecoveryHandler: Failed to update database \"{}\".", config.name);
return false;
}
} catch (const UnknownDatabaseException &) {
spdlog::debug("SystemRecoveryHandler: UnknownDatabaseException");
return false;
}
const auto it = std::find(old.begin(), old.end(), config.name);
if (it != old.end()) old.erase(it);
}
// Delete all the leftover old dbs
for (const auto &remove_db : old) {
const auto del = dbms_handler.Delete(remove_db);
if (del.HasError()) {
// Some errors are not terminal
if (del.GetError() == DeleteError::DEFAULT_DB || del.GetError() == DeleteError::NON_EXISTENT) {
spdlog::debug("SystemRecoveryHandler: Dropped database \"{}\".", remove_db);
continue;
}
spdlog::debug("SystemRecoveryHandler: Failed to drop database \"{}\".", remove_db);
return false;
}
}
/*
* SUCCESS
*/
return true;
}
void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
dbms::DbmsHandler &dbms_handler) {
// NOTE: Register even without license as the user could add a license at run-time
data.server->rpc_server_.Register<storage::replication::CreateDatabaseRpc>(
[&data, system_state_access, &dbms_handler](auto *req_reader, auto *res_builder) mutable {
spdlog::debug("Received CreateDatabaseRpc");
CreateDatabaseHandler(system_state_access, data.uuid_, dbms_handler, req_reader, res_builder);
});
data.server->rpc_server_.Register<storage::replication::DropDatabaseRpc>(
[&data, system_state_access, &dbms_handler](auto *req_reader, auto *res_builder) mutable {
spdlog::debug("Received DropDatabaseRpc");
DropDatabaseHandler(system_state_access, data.uuid_, dbms_handler, req_reader, res_builder);
});
}
#endif
} // namespace memgraph::dbms

View File

@@ -0,0 +1,42 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#include "dbms/dbms_handler.hpp"
#include "replication/state.hpp"
#include "slk/streams.hpp"
#include "system/state.hpp"
namespace memgraph::dbms {
#ifdef MG_ENTERPRISE
inline void LogWrongMain(const std::optional<utils::UUID> &current_main_uuid, const utils::UUID &main_req_id,
std::string_view rpc_req) {
spdlog::error("Received {} with main_id: {} != current_main_uuid: {}", rpc_req, std::string(main_req_id),
current_main_uuid.has_value() ? std::string(current_main_uuid.value()) : "");
}
// RPC handlers
void CreateDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, DbmsHandler &dbms_handler,
slk::Reader *req_reader, slk::Builder *res_builder);
void DropDatabaseHandler(memgraph::system::ReplicaHandlerAccessToState &system_state_access,
const std::optional<utils::UUID> &current_main_uuid, DbmsHandler &dbms_handler,
slk::Reader *req_reader, slk::Builder *res_builder);
bool SystemRecoveryHandler(DbmsHandler &dbms_handler, const std::vector<storage::SalientConfig> &database_configs);
// RPC registration
void Register(replication::RoleReplicaData const &data, system::ReplicaHandlerAccessToState &system_state_access,
dbms::DbmsHandler &dbms_handler);
#endif
} // namespace memgraph::dbms

118
src/dbms/rpc.cpp Normal file
View File

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

122
src/dbms/rpc.hpp Normal file
View File

@@ -0,0 +1,122 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#include <cstdint>
#include <string>
#include <utility>
#include "rpc/messages.hpp"
#include "slk/streams.hpp"
#include "storage/v2/config.hpp"
#include "utils/uuid.hpp"
namespace memgraph::storage::replication {
struct CreateDatabaseReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(CreateDatabaseReq *self, memgraph::slk::Reader *reader);
static void Save(const CreateDatabaseReq &self, memgraph::slk::Builder *builder);
CreateDatabaseReq() = default;
CreateDatabaseReq(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_group_timestamp,
uint64_t new_group_timestamp, storage::SalientConfig config)
: main_uuid(main_uuid),
epoch_id(std::move(epoch_id)),
expected_group_timestamp{expected_group_timestamp},
new_group_timestamp(new_group_timestamp),
config(std::move(config)) {}
utils::UUID main_uuid;
std::string epoch_id;
uint64_t expected_group_timestamp;
uint64_t new_group_timestamp;
storage::SalientConfig config;
};
struct CreateDatabaseRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
static void Load(CreateDatabaseRes *self, memgraph::slk::Reader *reader);
static void Save(const CreateDatabaseRes &self, memgraph::slk::Builder *builder);
CreateDatabaseRes() = default;
explicit CreateDatabaseRes(Result res) : result(res) {}
Result result;
};
using CreateDatabaseRpc = rpc::RequestResponse<CreateDatabaseReq, CreateDatabaseRes>;
struct DropDatabaseReq {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
static void Load(DropDatabaseReq *self, memgraph::slk::Reader *reader);
static void Save(const DropDatabaseReq &self, memgraph::slk::Builder *builder);
DropDatabaseReq() = default;
DropDatabaseReq(const utils::UUID &main_uuid, std::string epoch_id, uint64_t expected_group_timestamp,
uint64_t new_group_timestamp, const utils::UUID &uuid)
: main_uuid(main_uuid),
epoch_id(std::move(epoch_id)),
expected_group_timestamp{expected_group_timestamp},
new_group_timestamp(new_group_timestamp),
uuid(uuid) {}
utils::UUID main_uuid;
std::string epoch_id;
uint64_t expected_group_timestamp;
uint64_t new_group_timestamp;
utils::UUID uuid;
};
struct DropDatabaseRes {
static const utils::TypeInfo kType;
static const utils::TypeInfo &GetTypeInfo() { return kType; }
enum class Result : uint8_t { SUCCESS, NO_NEED, FAILURE, /* Leave at end */ N };
static void Load(DropDatabaseRes *self, memgraph::slk::Reader *reader);
static void Save(const DropDatabaseRes &self, memgraph::slk::Builder *builder);
DropDatabaseRes() = default;
explicit DropDatabaseRes(Result res) : result(res) {}
Result result;
};
using DropDatabaseRpc = rpc::RequestResponse<DropDatabaseReq, DropDatabaseRes>;
} // namespace memgraph::storage::replication
// SLK serialization declarations
namespace memgraph::slk {
void Save(const memgraph::storage::replication::CreateDatabaseReq &self, memgraph::slk::Builder *builder);
void Load(memgraph::storage::replication::CreateDatabaseReq *self, memgraph::slk::Reader *reader);
void Save(const memgraph::storage::replication::CreateDatabaseRes &self, memgraph::slk::Builder *builder);
void Load(memgraph::storage::replication::CreateDatabaseRes *self, memgraph::slk::Reader *reader);
void Save(const memgraph::storage::replication::DropDatabaseReq &self, memgraph::slk::Builder *builder);
void Load(memgraph::storage::replication::DropDatabaseReq *self, memgraph::slk::Reader *reader);
void Save(const memgraph::storage::replication::DropDatabaseRes &self, memgraph::slk::Builder *builder);
void Load(memgraph::storage::replication::DropDatabaseRes *self, memgraph::slk::Reader *reader);
} // namespace memgraph::slk

View File

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

View File

@@ -1,100 +0,0 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
#include "dbms/dbms_handler.hpp"
#include "dbms/replication_handler.hpp"
#include "replication/include/replication/state.hpp"
#include "utils/result.hpp"
namespace memgraph::dbms {
inline bool DoReplicaToMainPromotion(dbms::DbmsHandler &dbms_handler) {
// 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 (!dbms_handler.ReplicationState().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 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;
}
storage_clients.push_back(std::move(client));
return true;
});
});
return all_clients_good;
}
inline std::optional<RegisterReplicaError> HandleErrorOnReplicaClient(
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::END_POINT_EXISTS:
return dbms::RegisterReplicaError::END_POINT_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

@@ -195,3 +195,9 @@ DEFINE_HIDDEN_string(organization_name, "", "Organization name.");
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_string(auth_user_or_role_name_regex, memgraph::glue::kDefaultUserRoleRegex.data(),
"Set to the regular expression that each user or role name must fulfill.");
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_bool(auth_password_permit_null, true, "Set to false to disable null passwords.");
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DEFINE_string(auth_password_strength_regex, memgraph::glue::kDefaultPasswordRegex.data(),
"The regular expression that should be used to match the entire "
"entered password to ensure its strength.");

View File

@@ -118,3 +118,7 @@ DECLARE_string(license_key);
DECLARE_string(organization_name);
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_string(auth_user_or_role_name_regex);
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_bool(auth_password_permit_null);
// NOLINTNEXTLINE(cppcoreguidelines-avoid-non-const-global-variables)
DECLARE_string(auth_password_strength_regex);

View File

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

View File

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

View File

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

View File

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

View File

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

View File

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

View File

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

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -13,4 +13,5 @@
namespace memgraph::glue {
inline constexpr std::string_view kDefaultUserRoleRegex = "[a-zA-Z0-9_.+-@]+";
static constexpr std::string_view kDefaultPasswordRegex = ".+";
} // namespace memgraph::glue

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -210,16 +210,25 @@ std::vector<std::vector<memgraph::query::TypedValue>> ShowFineGrainedUserPrivile
if (!memgraph::license::global_license_checker.IsEnterpriseValidFast()) {
return {};
}
const auto &label_permissions = user->GetFineGrainedAccessLabelPermissions();
const auto &edge_type_permissions = user->GetFineGrainedAccessEdgeTypePermissions();
auto all_fine_grained_permissions =
GetFineGrainedPermissionForPrivilegeForUserOrRole(label_permissions, "LABEL", "USER");
auto edge_type_fine_grained_permissions =
GetFineGrainedPermissionForPrivilegeForUserOrRole(edge_type_permissions, "EDGE_TYPE", "USER");
auto all_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
user->GetUserFineGrainedAccessLabelPermissions(), "LABEL", "USER");
auto all_role_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
user->GetRoleFineGrainedAccessLabelPermissions(), "LABEL", "ROLE");
all_fine_grained_permissions.insert(all_fine_grained_permissions.end(),
std::make_move_iterator(all_role_fine_grained_permissions.begin()),
std::make_move_iterator(all_role_fine_grained_permissions.end()));
all_fine_grained_permissions.insert(all_fine_grained_permissions.end(), edge_type_fine_grained_permissions.begin(),
edge_type_fine_grained_permissions.end());
auto edge_type_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
user->GetUserFineGrainedAccessEdgeTypePermissions(), "EDGE_TYPE", "USER");
auto role_edge_type_fine_grained_permissions = GetFineGrainedPermissionForPrivilegeForUserOrRole(
user->GetRoleFineGrainedAccessEdgeTypePermissions(), "EDGE_TYPE", "ROLE");
all_fine_grained_permissions.insert(all_fine_grained_permissions.end(),
std::make_move_iterator(edge_type_fine_grained_permissions.begin()),
std::make_move_iterator(edge_type_fine_grained_permissions.end()));
all_fine_grained_permissions.insert(all_fine_grained_permissions.end(),
std::make_move_iterator(role_edge_type_fine_grained_permissions.begin()),
std::make_move_iterator(role_edge_type_fine_grained_permissions.end()));
return ConstructFineGrainedPrivilegesResult(all_fine_grained_permissions);
}
@@ -233,9 +242,9 @@ std::vector<std::vector<memgraph::query::TypedValue>> ShowFineGrainedRolePrivile
const auto &edge_type_permissions = role->GetFineGrainedAccessEdgeTypePermissions();
auto all_fine_grained_permissions =
GetFineGrainedPermissionForPrivilegeForUserOrRole(label_permissions, "LABEL", "USER");
GetFineGrainedPermissionForPrivilegeForUserOrRole(label_permissions, "LABEL", "ROLE");
auto edge_type_fine_grained_permissions =
GetFineGrainedPermissionForPrivilegeForUserOrRole(edge_type_permissions, "EDGE_TYPE", "USER");
GetFineGrainedPermissionForPrivilegeForUserOrRole(edge_type_permissions, "EDGE_TYPE", "ROLE");
all_fine_grained_permissions.insert(all_fine_grained_permissions.end(), edge_type_fine_grained_permissions.begin(),
edge_type_fine_grained_permissions.end());
@@ -248,31 +257,15 @@ std::vector<std::vector<memgraph::query::TypedValue>> ShowFineGrainedRolePrivile
namespace memgraph::glue {
AuthQueryHandler::AuthQueryHandler(
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
std::string name_regex_string)
: auth_(auth), name_regex_string_(std::move(name_regex_string)), name_regex_(name_regex_string_) {}
AuthQueryHandler::AuthQueryHandler(memgraph::auth::SynchedAuth *auth) : auth_(auth) {}
bool AuthQueryHandler::CreateUser(const std::string &username, const std::optional<std::string> &password) {
if (name_regex_string_ != kDefaultUserRoleRegex) {
if (const auto license_check_result =
memgraph::license::global_license_checker.IsEnterpriseValid(memgraph::utils::global_settings);
license_check_result.HasError()) {
throw memgraph::auth::AuthException(
"Custom user/role regex is a Memgraph Enterprise feature. Please set the config "
"(\"--auth-user-or-role-name-regex\") to its default value (\"{}\") or remove the flag.\n{}",
kDefaultUserRoleRegex,
memgraph::license::LicenseCheckErrorToString(license_check_result.GetError(), "user/role regex"));
}
}
if (!std::regex_match(username, name_regex_)) {
throw query::QueryRuntimeException("Invalid user name.");
}
bool AuthQueryHandler::CreateUser(const std::string &username, const std::optional<std::string> &password,
system::Transaction *system_tx) {
try {
const auto [first_user, user_added] = std::invoke([&, this] {
auto locked_auth = auth_->Lock();
const auto first_user = !locked_auth->HasUsers();
const auto user_added = locked_auth->AddUser(username, password).has_value();
const auto user_added = locked_auth->AddUser(username, password, system_tx).has_value();
return std::make_pair(first_user, user_added);
});
@@ -291,10 +284,11 @@ bool AuthQueryHandler::CreateUser(const std::string &username, const std::option
}
}
#endif
);
,
system_tx);
#ifdef MG_ENTERPRISE
GrantDatabaseToUser(auth::kAllDatabases, username);
SetMainDatabase(dbms::kDefaultDB, username);
GrantDatabaseToUser(auth::kAllDatabases, username, system_tx);
SetMainDatabase(dbms::kDefaultDB, username, system_tx);
#endif
}
@@ -304,73 +298,61 @@ bool AuthQueryHandler::CreateUser(const std::string &username, const std::option
}
}
bool AuthQueryHandler::DropUser(const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
bool AuthQueryHandler::DropUser(const std::string &username, system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
if (!user) return false;
return locked_auth->RemoveUser(username);
return locked_auth->RemoveUser(username, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
void AuthQueryHandler::SetPassword(const std::string &username, const std::optional<std::string> &password) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
void AuthQueryHandler::SetPassword(const std::string &username, const std::optional<std::string> &password,
system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
if (!user) {
throw memgraph::query::QueryRuntimeException("User '{}' doesn't exist.", username);
}
user->UpdatePassword(password);
locked_auth->SaveUser(*user);
locked_auth->UpdatePassword(*user, password);
locked_auth->SaveUser(*user, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
bool AuthQueryHandler::CreateRole(const std::string &rolename) {
if (!std::regex_match(rolename, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid role name.");
}
bool AuthQueryHandler::CreateRole(const std::string &rolename, system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
return locked_auth->AddRole(rolename).has_value();
return locked_auth->AddRole(rolename, system_tx).has_value();
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
#ifdef MG_ENTERPRISE
bool AuthQueryHandler::RevokeDatabaseFromUser(const std::string &db, const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
bool AuthQueryHandler::RevokeDatabaseFromUser(const std::string &db_name, const std::string &username,
system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
if (!user) return false;
return locked_auth->RevokeDatabaseFromUser(db, username);
return locked_auth->RevokeDatabaseFromUser(db_name, username, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
bool AuthQueryHandler::GrantDatabaseToUser(const std::string &db, const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
bool AuthQueryHandler::GrantDatabaseToUser(const std::string &db_name, const std::string &username,
system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
if (!user) return false;
return locked_auth->GrantDatabaseToUser(db, username);
return locked_auth->GrantDatabaseToUser(db_name, username, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
@@ -378,9 +360,6 @@ bool AuthQueryHandler::GrantDatabaseToUser(const std::string &db, const std::str
std::vector<std::vector<memgraph::query::TypedValue>> AuthQueryHandler::GetDatabasePrivileges(
const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user or role name.");
}
try {
auto locked_auth = auth_->ReadLock();
auto user = locked_auth->GetUser(username);
@@ -393,33 +372,28 @@ std::vector<std::vector<memgraph::query::TypedValue>> AuthQueryHandler::GetDatab
}
}
bool AuthQueryHandler::SetMainDatabase(std::string_view db, const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
bool AuthQueryHandler::SetMainDatabase(std::string_view db_name, const std::string &username,
system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
if (!user) return false;
return locked_auth->SetMainDatabase(db, username);
return locked_auth->SetMainDatabase(db_name, username, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
void AuthQueryHandler::DeleteDatabase(std::string_view db) {
void AuthQueryHandler::DeleteDatabase(std::string_view db_name, system::Transaction *system_tx) {
try {
auth_->Lock()->DeleteDatabase(std::string(db));
auth_->Lock()->DeleteDatabase(std::string(db_name), system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
#endif
bool AuthQueryHandler::DropRole(const std::string &rolename) {
if (!std::regex_match(rolename, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid role name.");
}
bool AuthQueryHandler::DropRole(const std::string &rolename, system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto role = locked_auth->GetRole(rolename);
@@ -428,7 +402,7 @@ bool AuthQueryHandler::DropRole(const std::string &rolename) {
return false;
};
return locked_auth->RemoveRole(rolename);
return locked_auth->RemoveRole(rolename, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
@@ -465,9 +439,6 @@ std::vector<memgraph::query::TypedValue> AuthQueryHandler::GetRolenames() {
}
std::optional<std::string> AuthQueryHandler::GetRolenameForUser(const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
try {
auto locked_auth = auth_->ReadLock();
auto user = locked_auth->GetUser(username);
@@ -485,9 +456,6 @@ std::optional<std::string> AuthQueryHandler::GetRolenameForUser(const std::strin
}
std::vector<memgraph::query::TypedValue> AuthQueryHandler::GetUsernamesForRole(const std::string &rolename) {
if (!std::regex_match(rolename, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid role name.");
}
try {
auto locked_auth = auth_->ReadLock();
auto role = locked_auth->GetRole(rolename);
@@ -506,13 +474,8 @@ std::vector<memgraph::query::TypedValue> AuthQueryHandler::GetUsernamesForRole(c
}
}
void AuthQueryHandler::SetRole(const std::string &username, const std::string &rolename) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
if (!std::regex_match(rolename, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid role name.");
}
void AuthQueryHandler::SetRole(const std::string &username, const std::string &rolename,
system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
@@ -528,16 +491,13 @@ void AuthQueryHandler::SetRole(const std::string &username, const std::string &r
current_role->rolename());
}
user->SetRole(*role);
locked_auth->SaveUser(*user);
locked_auth->SaveUser(*user, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
void AuthQueryHandler::ClearRole(const std::string &username) {
if (!std::regex_match(username, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user name.");
}
void AuthQueryHandler::ClearRole(const std::string &username, system::Transaction *system_tx) {
try {
auto locked_auth = auth_->Lock();
auto user = locked_auth->GetUser(username);
@@ -545,16 +505,13 @@ void AuthQueryHandler::ClearRole(const std::string &username) {
throw memgraph::query::QueryRuntimeException("User '{}' doesn't exist .", username);
}
user->ClearRole();
locked_auth->SaveUser(*user);
locked_auth->SaveUser(*user, system_tx);
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());
}
}
std::vector<std::vector<memgraph::query::TypedValue>> AuthQueryHandler::GetPrivileges(const std::string &user_or_role) {
if (!std::regex_match(user_or_role, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user or role name.");
}
try {
auto locked_auth = auth_->ReadLock();
std::vector<std::vector<memgraph::query::TypedValue>> grants;
@@ -602,7 +559,8 @@ void AuthQueryHandler::GrantPrivilege(
const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
&edge_type_privileges
#endif
) {
,
system::Transaction *system_tx) {
EditPermissions(
user_or_role, privileges,
#ifdef MG_ENTERPRISE
@@ -625,11 +583,13 @@ void AuthQueryHandler::GrantPrivilege(
}
}
#endif
);
,
system_tx);
} // namespace memgraph::glue
void AuthQueryHandler::DenyPrivilege(const std::string &user_or_role,
const std::vector<memgraph::query::AuthQuery::Privilege> &privileges) {
const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,
system::Transaction *system_tx) {
EditPermissions(
user_or_role, privileges,
#ifdef MG_ENTERPRISE
@@ -645,7 +605,8 @@ void AuthQueryHandler::DenyPrivilege(const std::string &user_or_role,
,
[](auto &fine_grained_permissions, const auto &privilege_collection) {}
#endif
);
,
system_tx);
}
void AuthQueryHandler::RevokePrivilege(
@@ -657,7 +618,8 @@ void AuthQueryHandler::RevokePrivilege(
const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
&edge_type_privileges
#endif
) {
,
system::Transaction *system_tx) {
EditPermissions(
user_or_role, privileges,
#ifdef MG_ENTERPRISE
@@ -679,7 +641,8 @@ void AuthQueryHandler::RevokePrivilege(
}
}
#endif
);
,
system_tx);
} // namespace memgraph::glue
template <class TEditPermissionsFun
@@ -703,10 +666,8 @@ void AuthQueryHandler::EditPermissions(
,
const TEditFineGrainedPermissionsFun &edit_fine_grained_permissions_fun
#endif
) {
if (!std::regex_match(user_or_role, name_regex_)) {
throw memgraph::query::QueryRuntimeException("Invalid user or role name.");
}
,
system::Transaction *system_tx) {
try {
std::vector<memgraph::auth::Permission> permissions;
permissions.reserve(privileges.size());
@@ -735,7 +696,7 @@ void AuthQueryHandler::EditPermissions(
}
}
#endif
locked_auth->SaveUser(*user);
locked_auth->SaveUser(*user, system_tx);
} else {
for (const auto &permission : permissions) {
edit_permissions_fun(role->permissions(), permission);
@@ -751,7 +712,7 @@ void AuthQueryHandler::EditPermissions(
}
}
#endif
locked_auth->SaveRole(*role);
locked_auth->SaveRole(*role, system_tx);
}
} catch (const memgraph::auth::AuthException &e) {
throw memgraph::query::QueryRuntimeException(e.what());

View File

@@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -23,35 +23,36 @@
namespace memgraph::glue {
class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth_;
std::string name_regex_string_;
std::regex name_regex_;
memgraph::auth::SynchedAuth *auth_;
public:
AuthQueryHandler(memgraph::utils::Synchronized<memgraph::auth::Auth, memgraph::utils::WritePrioritizedRWLock> *auth,
std::string name_regex_string);
explicit AuthQueryHandler(memgraph::auth::SynchedAuth *auth);
bool CreateUser(const std::string &username, const std::optional<std::string> &password) override;
bool CreateUser(const std::string &username, const std::optional<std::string> &password,
system::Transaction *system_tx) override;
bool DropUser(const std::string &username) override;
bool DropUser(const std::string &username, system::Transaction *system_tx) override;
void SetPassword(const std::string &username, const std::optional<std::string> &password) override;
void SetPassword(const std::string &username, const std::optional<std::string> &password,
system::Transaction *system_tx) override;
#ifdef MG_ENTERPRISE
bool RevokeDatabaseFromUser(const std::string &db, const std::string &username) override;
bool RevokeDatabaseFromUser(const std::string &db_name, const std::string &username,
system::Transaction *system_tx) override;
bool GrantDatabaseToUser(const std::string &db, const std::string &username) override;
bool GrantDatabaseToUser(const std::string &db_name, const std::string &username,
system::Transaction *system_tx) override;
std::vector<std::vector<memgraph::query::TypedValue>> GetDatabasePrivileges(const std::string &username) override;
bool SetMainDatabase(std::string_view db, const std::string &username) override;
bool SetMainDatabase(std::string_view db_name, const std::string &username, system::Transaction *system_tx) override;
void DeleteDatabase(std::string_view db) override;
void DeleteDatabase(std::string_view db_name, system::Transaction *system_tx) override;
#endif
bool CreateRole(const std::string &rolename) override;
bool CreateRole(const std::string &rolename, system::Transaction *system_tx) override;
bool DropRole(const std::string &rolename) override;
bool DropRole(const std::string &rolename, system::Transaction *system_tx) override;
std::vector<memgraph::query::TypedValue> GetUsernames() override;
@@ -61,9 +62,9 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
std::vector<memgraph::query::TypedValue> GetUsernamesForRole(const std::string &rolename) override;
void SetRole(const std::string &username, const std::string &rolename) override;
void SetRole(const std::string &username, const std::string &rolename, system::Transaction *system_tx) override;
void ClearRole(const std::string &username) override;
void ClearRole(const std::string &username, system::Transaction *system_tx) override;
std::vector<std::vector<memgraph::query::TypedValue>> GetPrivileges(const std::string &user_or_role) override;
@@ -77,10 +78,12 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
&edge_type_privileges
#endif
) override;
,
system::Transaction *system_tx) override;
void DenyPrivilege(const std::string &user_or_role,
const std::vector<memgraph::query::AuthQuery::Privilege> &privileges) override;
const std::vector<memgraph::query::AuthQuery::Privilege> &privileges,
system::Transaction *system_tx) override;
void RevokePrivilege(
const std::string &user_or_role, const std::vector<memgraph::query::AuthQuery::Privilege> &privileges
@@ -91,7 +94,8 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
const std::vector<std::unordered_map<memgraph::query::AuthQuery::FineGrainedPrivilege, std::vector<std::string>>>
&edge_type_privileges
#endif
) override;
,
system::Transaction *system_tx) override;
private:
template <class TEditPermissionsFun
@@ -115,6 +119,7 @@ class AuthQueryHandler final : public memgraph::query::AuthQueryHandler {
,
const TEditFineGrainedPermissionsFun &edit_fine_grained_permissions_fun
#endif
);
,
system::Transaction *system_tx);
};
} // namespace memgraph::glue

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 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@@ -61,8 +61,8 @@ struct Endpoint {
* it into an ip address and a port number; even if a default port is given,
* it won't be used, as we expect that it is given in the address string.
*/
static std::optional<std::pair<std::string, uint16_t>> ParseSocketOrIpAddress(const std::string &address,
std::optional<uint16_t> default_port);
static std::optional<std::pair<std::string, uint16_t>> ParseSocketOrIpAddress(
const std::string &address, std::optional<uint16_t> default_port = {});
/**
* Tries to parse given string as either socket address or hostname.

View File

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

View File

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

View File

@@ -32,7 +32,7 @@
#include "utils/timer.hpp"
#include "version.hpp"
using memgraph::replication::ReplicationRole;
using memgraph::replication_coordination_glue::ReplicationRole;
bool ValidateControlCharacter(const char *flagname, const std::string &value) {
if (value.empty()) {

View File

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

View File

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

View File

@@ -3034,7 +3034,7 @@ class ReplicationQuery : public memgraph::query::Query {
enum class SyncMode { SYNC, ASYNC };
enum class ReplicaState { READY, REPLICATING, RECOVERY, MAYBE_BEHIND };
enum class ReplicaState { READY, REPLICATING, RECOVERY, MAYBE_BEHIND, UNREACHABLE };
ReplicationQuery() = default;
@@ -3071,14 +3071,7 @@ class CoordinatorQuery : public memgraph::query::Query {
static const utils::TypeInfo kType;
const utils::TypeInfo &GetTypeInfo() const override { return kType; }
enum class Action {
REGISTER_MAIN_COORDINATOR_SERVER,
REGISTER_REPLICA_COORDINATOR_SERVER,
SHOW_REPLICATION_CLUSTER,
DO_FAILOVER
};
enum class ReplicationRole { MAIN, REPLICA };
enum class Action { REGISTER_INSTANCE, SET_INSTANCE_TO_MAIN, SHOW_INSTANCES, ADD_COORDINATOR_INSTANCE };
enum class SyncMode { SYNC, ASYNC };
@@ -3087,21 +3080,24 @@ class CoordinatorQuery : public memgraph::query::Query {
DEFVISITABLE(QueryVisitor<void>);
memgraph::query::CoordinatorQuery::Action action_;
memgraph::query::CoordinatorQuery::ReplicationRole role_;
std::string instance_name_;
memgraph::query::Expression *socket_address_{nullptr};
memgraph::query::Expression *replication_socket_address_{nullptr};
memgraph::query::Expression *coordinator_socket_address_{nullptr};
memgraph::query::Expression *raft_socket_address_{nullptr};
memgraph::query::Expression *raft_server_id_{nullptr};
memgraph::query::CoordinatorQuery::SyncMode sync_mode_;
CoordinatorQuery *Clone(AstStorage *storage) const override {
auto *object = storage->Create<CoordinatorQuery>();
object->action_ = action_;
object->role_ = role_;
object->instance_name_ = instance_name_;
object->socket_address_ = socket_address_ ? socket_address_->Clone(storage) : nullptr;
object->replication_socket_address_ =
replication_socket_address_ ? replication_socket_address_->Clone(storage) : nullptr;
object->sync_mode_ = sync_mode_;
object->coordinator_socket_address_ =
coordinator_socket_address_ ? coordinator_socket_address_->Clone(storage) : nullptr;
object->raft_socket_address_ = raft_socket_address_ ? raft_socket_address_->Clone(storage) : nullptr;
object->raft_server_id_ = raft_server_id_ ? raft_server_id_->Clone(storage) : nullptr;
return object;
}

View File

@@ -374,60 +374,53 @@ antlrcpp::Any CypherMainVisitor::visitRegisterReplica(MemgraphCypher::RegisterRe
return replication_query;
}
// License check is done in the interpreter.
antlrcpp::Any CypherMainVisitor::visitRegisterCoordinatorServer(MemgraphCypher::RegisterCoordinatorServerContext *ctx) {
MG_ASSERT(ctx->children.size() == 1, "RegisterCoordinatorServerQuery should have exactly one child!");
auto *coordinator_query = std::any_cast<CoordinatorQuery *>(ctx->children[0]->accept(this));
query_ = coordinator_query;
return coordinator_query;
}
// License check is done in the interpreter
antlrcpp::Any CypherMainVisitor::visitShowReplicationCluster(MemgraphCypher::ShowReplicationClusterContext * /*ctx*/) {
antlrcpp::Any CypherMainVisitor::visitRegisterInstanceOnCoordinator(
MemgraphCypher::RegisterInstanceOnCoordinatorContext *ctx) {
auto *coordinator_query = storage_->Create<CoordinatorQuery>();
coordinator_query->action_ = CoordinatorQuery::Action::SHOW_REPLICATION_CLUSTER;
return coordinator_query;
}
// License check is done in the interpreter
antlrcpp::Any CypherMainVisitor::visitRegisterReplicaCoordinatorServer(
MemgraphCypher::RegisterReplicaCoordinatorServerContext *ctx) {
auto *coordinator_query = storage_->Create<CoordinatorQuery>();
if (!ctx->socketAddress()->literal()->StringLiteral()) {
throw SemanticException("Socket address should be a string literal!");
if (!ctx->replicationSocketAddress()->literal()->StringLiteral()) {
throw SemanticException("Replication socket address should be a string literal!");
}
if (!ctx->coordinatorSocketAddress()->literal()->StringLiteral()) {
throw SemanticException("Coordinator socket address should be a string literal!");
}
coordinator_query->action_ = CoordinatorQuery::Action::REGISTER_REPLICA_COORDINATOR_SERVER;
coordinator_query->role_ = CoordinatorQuery::ReplicationRole::REPLICA;
coordinator_query->socket_address_ = std::any_cast<Expression *>(ctx->socketAddress()->accept(this));
coordinator_query->action_ = CoordinatorQuery::Action::REGISTER_INSTANCE;
coordinator_query->replication_socket_address_ =
std::any_cast<Expression *>(ctx->replicationSocketAddress()->accept(this));
coordinator_query->coordinator_socket_address_ =
std::any_cast<Expression *>(ctx->coordinatorSocketAddress()->accept(this));
coordinator_query->instance_name_ = std::any_cast<std::string>(ctx->instanceName()->symbolicName()->accept(this));
if (ctx->SYNC()) {
coordinator_query->sync_mode_ = memgraph::query::CoordinatorQuery::SyncMode::SYNC;
} else if (ctx->ASYNC()) {
if (ctx->ASYNC()) {
coordinator_query->sync_mode_ = memgraph::query::CoordinatorQuery::SyncMode::ASYNC;
} else {
coordinator_query->sync_mode_ = memgraph::query::CoordinatorQuery::SyncMode::SYNC;
}
return coordinator_query;
}
// License check is done in the interpreter
antlrcpp::Any CypherMainVisitor::visitRegisterMainCoordinatorServer(
MemgraphCypher::RegisterMainCoordinatorServerContext *ctx) {
if (!ctx->coordinatorSocketAddress()->literal()->StringLiteral()) {
throw SemanticException("Coordinator socket address should be a string literal!");
}
antlrcpp::Any CypherMainVisitor::visitAddCoordinatorInstance(MemgraphCypher::AddCoordinatorInstanceContext *ctx) {
auto *coordinator_query = storage_->Create<CoordinatorQuery>();
coordinator_query->action_ = CoordinatorQuery::Action::REGISTER_MAIN_COORDINATOR_SERVER;
coordinator_query->role_ = CoordinatorQuery::ReplicationRole::MAIN;
coordinator_query->coordinator_socket_address_ =
std::any_cast<Expression *>(ctx->coordinatorSocketAddress()->accept(this));
coordinator_query->instance_name_ = std::any_cast<std::string>(ctx->instanceName()->symbolicName()->accept(this));
if (!ctx->raftSocketAddress()->literal()->StringLiteral()) {
throw SemanticException("Raft socket address should be a string literal!");
}
if (!ctx->raftServerId()->literal()->numberLiteral()) {
throw SemanticException("Raft server id should be a number literal!");
}
coordinator_query->action_ = CoordinatorQuery::Action::ADD_COORDINATOR_INSTANCE;
coordinator_query->raft_socket_address_ = std::any_cast<Expression *>(ctx->raftSocketAddress()->accept(this));
coordinator_query->raft_server_id_ = std::any_cast<Expression *>(ctx->raftServerId()->accept(this));
return coordinator_query;
}
// License check is done in the interpreter
antlrcpp::Any CypherMainVisitor::visitShowInstances(MemgraphCypher::ShowInstancesContext * /*ctx*/) {
auto *coordinator_query = storage_->Create<CoordinatorQuery>();
coordinator_query->action_ = CoordinatorQuery::Action::SHOW_INSTANCES;
return coordinator_query;
}
@@ -445,9 +438,10 @@ antlrcpp::Any CypherMainVisitor::visitShowReplicas(MemgraphCypher::ShowReplicasC
}
// License check is done in the interpreter
antlrcpp::Any CypherMainVisitor::visitDoFailover(MemgraphCypher::DoFailoverContext * /*ctx*/) {
antlrcpp::Any CypherMainVisitor::visitSetInstanceToMain(MemgraphCypher::SetInstanceToMainContext *ctx) {
auto *coordinator_query = storage_->Create<CoordinatorQuery>();
coordinator_query->action_ = CoordinatorQuery::Action::DO_FAILOVER;
coordinator_query->action_ = CoordinatorQuery::Action::SET_INSTANCE_TO_MAIN;
coordinator_query->instance_name_ = std::any_cast<std::string>(ctx->instanceName()->symbolicName()->accept(this));
query_ = coordinator_query;
return coordinator_query;
}

View File

@@ -241,28 +241,22 @@ class CypherMainVisitor : public antlropencypher::MemgraphCypherBaseVisitor {
/**
* @return CoordinatorQuery*
*/
antlrcpp::Any visitRegisterCoordinatorServer(MemgraphCypher::RegisterCoordinatorServerContext *ctx) override;
antlrcpp::Any visitRegisterInstanceOnCoordinator(MemgraphCypher::RegisterInstanceOnCoordinatorContext *ctx) override;
/**
* @return CoordinatorQuery*
*/
antlrcpp::Any visitRegisterMainCoordinatorServer(MemgraphCypher::RegisterMainCoordinatorServerContext *ctx) override;
antlrcpp::Any visitSetInstanceToMain(MemgraphCypher::SetInstanceToMainContext *ctx) override;
/**
* @return CoordinatorQuery*
*/
antlrcpp::Any visitRegisterReplicaCoordinatorServer(
MemgraphCypher::RegisterReplicaCoordinatorServerContext *ctx) override;
antlrcpp::Any visitAddCoordinatorInstance(MemgraphCypher::AddCoordinatorInstanceContext *ctx) override;
/**
* @return CoordinatorQuery*
*/
antlrcpp::Any visitShowReplicationCluster(MemgraphCypher::ShowReplicationClusterContext *ctx) override;
/**
* @return CoordinatorQuery*
*/
antlrcpp::Any visitDoFailover(MemgraphCypher::DoFailoverContext *ctx) override;
antlrcpp::Any visitShowInstances(MemgraphCypher::ShowInstancesContext *ctx) override;
/**
* @return LockPathQuery*

View File

@@ -102,6 +102,7 @@ FILTER : F I L T E R ;
IN : I N ;
INDEX : I N D E X ;
INFO : I N F O ;
INSTANCE : I N S T A N C E ;
IS : I S ;
KB : K B ;
KEY : K E Y ;
@@ -122,6 +123,7 @@ PROCEDURE : P R O C E D U R E ;
PROFILE : P R O F I L E ;
QUERY : Q U E R Y ;
REDUCE : R E D U C E ;
REGISTER : R E G I S T E R;
REMOVE : R E M O V E ;
RETURN : R E T U R N ;
SET : S E T ;

View File

@@ -20,6 +20,7 @@ options { tokenVocab=MemgraphCypherLexer; }
import Cypher ;
memgraphCypherKeyword : cypherKeyword
| ADD
| ACTIVE
| AFTER
| ALTER
@@ -63,6 +64,8 @@ memgraphCypherKeyword : cypherKeyword
| GRANT
| HEADER
| IDENTIFIED
| INSTANCE
| INSTANCES
| NODE_LABELS
| NULLIF
| IMPORT
@@ -186,9 +189,10 @@ replicationQuery : setReplicationRole
| showReplicas
;
coordinatorQuery : registerCoordinatorServer
| showReplicationCluster
| doFailover
coordinatorQuery : registerInstanceOnCoordinator
| setInstanceToMain
| showInstances
| addCoordinatorInstance
;
triggerQuery : createTrigger
@@ -252,8 +256,6 @@ transactionQueueQuery : showTransactions
showTransactions : SHOW TRANSACTIONS ;
doFailover : DO FAILOVER ;
terminateTransactions : TERMINATE TRANSACTIONS transactionIdList;
loadCsv : LOAD CSV FROM csvFile ( WITH | NO ) HEADER
@@ -375,22 +377,26 @@ setReplicationRole : SET REPLICATION ROLE TO ( MAIN | REPLICA )
showReplicationRole : SHOW REPLICATION ROLE ;
showReplicationCluster : SHOW REPLICATION CLUSTER ;
showInstances : SHOW INSTANCES ;
instanceName : symbolicName ;
socketAddress : literal ;
coordinatorSocketAddress : literal ;
replicationSocketAddress : literal ;
raftSocketAddress : literal ;
registerReplica : REGISTER REPLICA instanceName ( SYNC | ASYNC )
TO socketAddress ;
registerReplicaCoordinatorServer: REGISTER REPLICA instanceName ( ASYNC | SYNC ) TO socketAddress WITH COORDINATOR SERVER ON coordinatorSocketAddress ;
registerInstanceOnCoordinator : REGISTER INSTANCE instanceName ON coordinatorSocketAddress ( AS ASYNC ) ? WITH replicationSocketAddress ;
registerMainCoordinatorServer: REGISTER MAIN instanceName WITH COORDINATOR SERVER ON coordinatorSocketAddress ;
setInstanceToMain : SET INSTANCE instanceName TO MAIN ;
registerCoordinatorServer : registerMainCoordinatorServer | registerReplicaCoordinatorServer ;
raftServerId : literal ;
addCoordinatorInstance : ADD COORDINATOR raftServerId ON raftSocketAddress ;
dropReplica : DROP REPLICA instanceName ;

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