nevzheng commented on issue #13515: URL: https://github.com/apache/gravitino/issues/13515#issuecomment-5881954602
**MySQL core DB integration tests now run in 7m05s, down from 26m25s — a 73.2% cut (3.73x) on CI-class hardware, with 520/520 still passing.** Point-in-time snapshot of where the `core` database-test stack (#13518 fixture reuse -> #13530 adaptive forks -> #13553 shared container + DB tuning) stands today. All numbers measured on 8vCPU/31GB hardware matching the CI runner class, `-PcoreDatabaseForks=2` where applicable. | Stage | MySQL | PostgreSQL | H2 | |:--|--:|--:|--:| | Baseline: 2 containers/fork (pre-sharing) | 26m 25s | 7m 10s | 1m 27s | | Shared container, untuned | 27m 53s | not separately measured | not separately measured | | Shared container, tuned v1 (4 innodb flags) | 22m 15s | — | — | | **Shared container, fully tuned (all flags + tmpfs) — current** | **7m 05s** | **2m 11s** | **1m 21s** | Deltas at current state vs. baseline: - **MySQL: -73.2% (3.73x)** - **PostgreSQL: -69.4% (3.27x)** - H2: -6.5% (1.07x) — the no-container control, confirming the win comes from container/DB tuning, not test-selection changes **Correctness held at every stage:** 520/520 tests pass, 0 deadlocks. PostgreSQL's 2 skips are pre-existing and appear identically in the baseline run. The untuned checkpoint (27m53s) is the important shape in that table: collapsing to one shared container alone was a wall-clock wash-to-small-loss. The speedup is the *tuning* on top of sharing, landed in `4b0626ba8`. **Go deeper:** full flag-by-flag breakdown and the durability-relaxation safety justification are in the PR comment on #13553: https://github.com/apache/gravitino/pull/13553#issuecomment-5881785368 Status: #13518, #13530, and #13553 are all still unmerged. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
