PS-11185 - Group replication test with docker containers by tplavcic · Pull Request #4 · Percona-QA/server-qa

tplavcic · 2026-06-10T11:41:06Z

This creates a pytest framework for running group replication tests with docker images. The purpose is to cover some scenarios which are used in K8S operators.
Docker images which are used can be customized with env variables SERVER_IMAGE, HAPROXY_IMAGE, ROUTER_IMAGE, XTRABACKUP_IMAGE and SYSBENCH_IMAGE.
The test run with both haproxy and mysql router.

It has helpers for following objects:

docker
group replication
xtrabackup
sysbench

Initial tests cover:

basic group replication sanity
failover with load
scaling test with load
backup/restore test with full and incremental backup

Add HAProxy as a second proxy in front of the Group Replication cluster alongside MySQL Router, and run the suite under both. - Generalize the proxy abstraction in GroupReplication: a `proxy` of router/haproxy/None drives rw_endpoint/ro_endpoint/exec_sql, verify (check_proxy=), and the renamed public wait_proxy_ready(). - HAProxy uses the Percona operator image (perconalab/percona-server- mysql-operator:main-haproxy): write frontend :3307, read frontend :3308. Since HAProxy is not SQL-aware and the forking external-check is unreliable under amd64-on-arm64 emulation, both backends use mysql-check for liveness and the framework pins the write target to the current primary via the runtime API (ready/maint), re-pinning on failover. - conftest: parametrize gr_cluster over both proxies with a @pytest.mark. proxies(...) marker for per-test selection; move create() inside the fixture try/finally so a failed bring-up tears down instead of leaking. - docker_helper.create gains a `platform` arg (image is amd64-only); _run and exec_mysql gain a `timeout` so a probe to a proxy with no live backend can't block forever. Full suite (test_basic + test_failover, both proxies) passes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Replace the implicit conftest parametrization (pytest_generate_tests + @pytest.mark.proxies marker) with an explicit @pytest.mark.parametrize("gr_cluster", [...], indirect=True) on each test, so the proxy set each test runs under is visible at the test itself. - conftest: drop pytest_configure (proxies marker) and pytest_generate_tests; keep PROXIES and the gr_cluster fixture (reads request.param, now supplied by each test's decorator). - test_basic / test_failover: import pytest and parametrize gr_cluster over ["router", "haproxy"]; narrowing to one proxy is just a shorter list. - README: document the explicit parametrize form (router-only/haproxy-only/both). - Add a scoped .gitignore for __pycache__/, *.pyc, .pytest_cache/. Verified: collection shows [router]/[haproxy] ids per test; test_basic passes under both proxies. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

verify() already reads membership/roles from replication_group_members; emit a [GR] log line listing the primary and secondary hosts (obeys GR_VERBOSE), so a run shows who holds which role at each verify — including the newly elected primary after failover. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Adds a self-contained pytest-based framework under test_scripts/ps/group-replication/ to spin up a Percona Server Group Replication cluster in containers (optionally fronted by MySQL Router or HAProxy) and run end-to-end scenarios similar to operator/K8s workflows.

Changes:

Introduces Docker CLI wrappers and a GroupReplication lifecycle helper (bootstrap, verify, failover, scale up/down, checksums).
Adds helper containers for workload (Sysbench) and physical backup/restore (XtraBackup).
Adds initial pytest suite covering sanity replication, primary failover, scaling under load, and full+incremental backup/restore, plus README/docs and pytest config.

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
test_scripts/ps/group-replication/docker_helper.py	Docker/Podman CLI wrapper for container/network/volume lifecycle and in-container mysql/mysqlsh execution.
test_scripts/ps/group-replication/group_replication_helper.py	Core GR cluster orchestration (create/destroy, proxy support, scale/failover helpers, verification and checksums).
test_scripts/ps/group-replication/sysbench_helper.py	Sysbench runner helper for load generation via ephemeral containers.
test_scripts/ps/group-replication/xtrabackup_helper.py	XtraBackup helper to run full/incremental backups and restore via ephemeral containers.
test_scripts/ps/group-replication/conftest.py	Pytest fixtures wiring cluster/proxy selection plus sysbench and xtrabackup setup/teardown.
test_scripts/ps/group-replication/test_basic.py	Basic replication sanity test (DDL/DML + checksum verification).
test_scripts/ps/group-replication/test_failover.py	Primary failover/recovery scenario under sysbench write workload.
test_scripts/ps/group-replication/test_scaling.py	Scale up/down scenario under sysbench workload with checksum validation.
test_scripts/ps/group-replication/test_backup_restore.py	Full+incremental backup/restore scenario and point-in-time checksum comparison.
test_scripts/ps/group-replication/pytest.ini	Local pytest configuration (discovery + live logging).
test_scripts/ps/group-replication/requirements.txt	Python dependencies for running the suite.
test_scripts/ps/group-replication/README.md	Usage docs, architecture notes, and how to run/debug scenarios locally.
test_scripts/ps/group-replication/.gitignore	Ignores Python/pytest local artifacts for this suite.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 3 comments.

… runs Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot