yuqi1129 opened a new issue, #14151: URL: https://github.com/apache/gravitino/issues/14151
### What would you like to be improved? After #13049, all PR suites run as reusable workflows inside one **Fork CI** run in the contributor's fork. GitHub only allows "Re-run failed jobs" once the whole run has completed, so when one job fails early (e.g. a flaky BackendIT), the author has to wait for the slowest suite (about two hours) before retrying. Before #13049 each suite was its own workflow run and could be retried as soon as it failed. The upstream **Fork CI** check also stays "in progress" until then, so the failure is not visible on the PR. In addition, the scheduled status sweep spends at most 60 API calls per run. On apache/gravitino (15,000 requests per hour, about 20 recently updated PRs, and schedules firing every ~15 minutes) one sweep does not reach every recent PR, so a finished fork run can take 30+ minutes to show on the PR. ### How should we improve? - Add a job to `fork-ci.yml` that watches the run and cancels it as soon as any job fails or times out, so the author can re-run failed jobs right away (this also re-runs the cancelled jobs). - Report such a run on the upstream check as `failure` titled with the failed job, rather than `cancelled`. - Scale the scheduled sweep budget with the hourly API quota (at most about half the quota per hour, never below 60 calls per sweep). -- 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]
