This harness compares:
- the community PGOed image built from this repository
- the upstream
percona/percona-serverimage
Covered workloads:
oltp_point_selectoltp_read_onlyoltp_read_write
docker- optional:
sysstatforiostatandpidstat
From the repository root:
bash docker/percona-server-8.4-pgoed/bench/run_head2head_sysbench.shThe script uses --network host and expects to bind 127.0.0.1:3306, so run it on a host without another MySQL instance listening there.
Default tunables:
OFFICIAL_IMAGE=percona/percona-server:8.4.8-8.1PGOED_IMAGE=ps-8.4.8-8-pgoed:latestSB_TABLES=16SB_TABLE_SIZE=500000SB_THREADS_LIST="1 4 8 16"SB_TIME=60SB_WARMUP_TIME=30SB_MYSQL_SSL=requiredSYSBENCH_IMAGE=perconalab/sysbench:latest
Output directory:
docker/percona-server-8.4-pgoed/bench/runs/<RUN_ID>/env.mdparams.envsysbench-version.txtofficial/pgoed/summary.jsonsummary.md
docker/percona-server-8.4-pgoed/bench/zz-benchmark.cnf
This config intentionally relaxes durability for throughput comparison. It is not a production recommendation.