nevzheng commented on issue #13900: URL: https://github.com/apache/gravitino/issues/13900#issuecomment-6293741946
**Timing results, 2026-10-20 (local, preliminary)** Measured on an Apple-silicon Mac (OrbStack Docker), images already pulled, one Gradle invocation per module, `main` and the stack alternating. Medians of 5 rounds per side; same tests and pass counts on both sides. Apache CI numbers are not in yet. **Main against the stack (#14005–#14012)** | Module | Tests | main | stack | change | | --- | ---: | ---: | ---: | ---: | | idp-basic | 221 | 488 s | 102 s | −79 % | | hive-metastore-common | 52 | 394 s | 175 s | −56 % | | flink-1.20 | 439 | 426 s | 263 s | −38 % | | catalog-kafka | 25 | 34 s | 22 s | −35 % | | catalog-hive | 308 | 578 s | 395 s | −32 % | | catalog-fileset | 436 | 253 s | 237 s | −6 % | | spark-3.5 | 581 | 482 s | 452 s | −6 % | Modules that start few containers (fileset, Spark) gain little; most of their time is the tests themselves. **Disposable data (#14013), durable against disposable, 2–3 rounds** | Module | Containers | durable | disposable | change | | --- | --- | ---: | ---: | ---: | | idp-basic | MySQL, PostgreSQL | 77 s | 34 s | ≈ −55 % | | catalog-jdbc-mysql | MySQL | 55 s | 43 s | ≈ −22 % | | hive-metastore-common | Hive | 173 s | 171 s | ≈ 0 % | | catalog-hive | Hive | 391 s | 388 s | ≈ 0 % | **What this includes:** container start timing, the lifecycle and readiness fixes, the fixture framework and the caller migrations. The appendix of the design doc (#14005) has the per-fix numbers. **Not included yet:** Apache CI runners, cold image pulls, and the Hive readiness change just pushed to #14006 (it drops two duplicate Hive CLI checks per start). **Other opportunities that could be explored** (not committed to): - Caching or pre-pulling test images in CI, where every job pulls them fresh. - Wait strategies for Hive, Doris, Ranger, StarRocks and RustFS after the Testcontainers 2.x upgrade (#13958). - Disposable data for Doris and StarRocks, which needs image changes. - Measuring the same modules on Apache CI. - Sharding test classes across CI jobs, if a single job grows long enough for test time to dominate. - A shared per-test database reset outside `core`, if a module turns out to need one. -- 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]
