wombatu-kun opened a new pull request, #19441: URL: https://github.com/apache/hudi/pull/19441
### Describe the issue this Pull Request addresses `docker-java17-test` declares `cache: maven`, but `run_docker_java17.sh` builds inside a container and never mounts `$HOME/.m2`, so the host repo `actions/setup-java` snapshots is near-empty. All jobs share one cache key and the first to finish wins the write, so when this job misses on restore it seals that near-empty repo for the rest of the PR. Measured on run 30610414199: [job 91092305450](https://github.com/apache/hudi/actions/runs/30610414199/job/91092305450) logged `maven cache is not found`, then `Cache saved with the key: setup-java-Linux-x64-maven-233646b0...`. The resulting entry is 8,940,929 bytes against 1,213,136,182 bytes on `refs/heads/master`. On #19408 an identical entry was restored by a later job in the same run (`Cache Size: ~9 MB (8939585 B)`), which then re-downloaded ~1 GB from Maven Central. That burst is what gets a runner's shared egress IP rate limited. Originating failure: [job 91092304952](https://github.com/apache/hudi/actions/runs/30610414199/job/91092304952), `Build Project` dead before any test ran on `scala-maven-plugin:pom:3.4.6 ... status code: 403, reason phrase: Forbidden (403)`. Same defect #19244 fixed for `validate-source`. ### Summary and Changelog - Drop `cache: maven` from `docker-java17-test`, with a comment mirroring the existing one on `validate-source`. - Delete the dead `HUDI_VERSION=$(mvn help:evaluate ...)` in that job: `run_docker_java17.sh` is called with no arguments and never reads it (unlike `ci_run.sh`, where it is `$2`). The job is then left with no host-side Maven at all. Audited: this is the only live job with that shape. `test-common-and-other-modules` also builds in Docker but mounts `-v "$HOME/.m2:/root/.m2"`; every other `cache: maven` job in `bot.yml` runs a real host build; the same-shaped jobs in `maven_artifact_validation.yml` and `release_candidate_validation.yml` all carry `if: false`. ### Impact CI only. Jobs that previously restored ~9 MB and re-downloaded from Maven Central will restore ~1 GB instead. ### Risk Level low Confined to one job whose real work runs inside a container with its own local repository. `cache-read-only: true` would be more surgical, but it exists only on `actions/setup-java`'s `main`, not the released `v5` tag pinned here. Verifiable on this PR's run: the job still passes, its post step no longer prints `Cache saved with the key: setup-java-...`, and the entry under this PR's ref is ~1 GB rather than ~9 MB. ### Documentation Update none ### Contributor's checklist - [x] Read through [contributor's guide](https://hudi.apache.org/contribute/how-to-contribute) - [x] Enough context is provided in the sections above - [x] Adequate tests were added if applicable -- 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]
