DanielLeens opened a new pull request, #12374: URL: https://github.com/apache/seatunnel/pull/12374
## Summary The all-modules `Run / unit-test` job is being cancelled at its 90-minute `timeout-minutes` budget even though nothing failed. Two occurrences in two days, on unrelated heads: | run | job | result | duration | | --- | --- | --- | --- | | `apache/seatunnel` dev push `35063682504` (2026-09-16) | `unit-test (11, windows-latest)` | cancelled | 90m | | #12352 fork run `35141464619` (2026-09-16) | `unit-test (11, ubuntu-latest)` | cancelled | 90m | In the #12352 case the job log shows Maven had already finished successfully before the runner was killed: ``` 2026-09-16T21:18:21Z [INFO] BUILD SUCCESS 2026-09-16T21:18:21Z [INFO] Total time: 01:28 h ``` The job started at `19:49:05Z` and was cancelled at `21:19:33Z` (90m28s): checkout, JDK setup and the 88-minute build left no room for the post-job steps. #12352 only validates SLS connector options; nothing in it can slow the whole reactor down. ## Root cause The job budget has not moved while the reactor keeps growing (several new connector modules merged per week). Durations of the same job on recent full-module runs: | run | 8 / ubuntu | 11 / ubuntu | 8 / windows | 11 / windows | | --- | --- | --- | --- | --- | | dev `35193601793` (09-17) | 77m | 70m | 56m | 50m | | dev `35063682504` (09-16) | 77m | 73m | failed early | **90m, cancelled** | | dev `34995029902` (09-15) | 77m | 66m | 64m | 60m | | #12352 `35141464619` | 83m | **90m, cancelled** | 65m | 59m | | #12351 `35123490714` | 85m | 79m | 58m | failed early | | #12338 `35045072216` | 70m | 74m | 57m | 62m | Successful ubuntu runs already sit at 70-85 minutes, i.e. within 5-20 minutes of the cap, so ordinary hosted-runner variance (roughly +/-10%) is enough to cross it. An exact-budget cancellation like this is not a flaky test and a rerun only re-rolls the dice; it will get more frequent as modules are added. ## Fix Raise `unit-test`'s `timeout-minutes` from 90 to 120 in `.github/workflows/backend.yml`, with a comment recording why. 120 matches the budget several single-connector IT jobs already have (`rocketmq-connector-it`, `mysql-cdc-connector-it`, `jdbc-connectors-it-part-*`, ...) and gives about 40% headroom over the slowest successful run observed, while still bounding a genuinely hung build. No test, module list, matrix entry or Maven argument changes; coverage is identical. ## Files - `.github/workflows/backend.yml` ## Test plan CI-configuration-only change. Verified by this PR's GitHub CI run of the workflow itself. 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- 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]
