wombatu-kun opened a new issue, #20384: URL: https://github.com/apache/hudi/issues/20384
### Problem hudi-bot has been silent since 2026-10-06 09:15 UTC, so no PR has been built on Azure for 11 days. Its last action was the CI report on #20112 for build 17482. Since then it has pushed no `ci_<pr>_<sha>` branch to `apachehudi-ci/hudi-branch-ci` and posted no report, and the newest build in the `apachehudi` Azure project is still 17482. The pipeline itself is enabled; it simply has nothing to build. The `Azure CI: pending / In progress` status on PRs is not a running build. `azure_ci.js` sets it as the default when hudi-bot has no report for the head commit, and the five-minute cron in `scheduled_workflow.yml` keeps re-applying it. ### Impact (as of 2026-10-17) - 102 open PRs show `Azure CI` pending with no build link, 89 of them with a head commit made after the bot stopped. - 50 PRs have merged to master without an Azure run. Azure is the only CI that runs the Spark datasource suites on Spark 3.5 / Scala 2.12; GitHub Actions runs them on Spark 4.2 / Scala 2.13 only. - `Azure CI` is not a required check, so nothing is blocked, which is probably why this went unnoticed. ### What does not help - Pushing a new commit, since nothing is being mirrored. - `@hudi-bot run azure`, tried on #19790 and #19661 on 2026-10-07. ### What to check on the bot host `ci-bot` catches every exception and restarts itself after five minutes, so eleven days of silence means the JVM or its host is down, a call is hung, or every poll fails, for example on an expired or revoked GitHub token for `hudi-bot`. The Azure token alone would not explain it, since mirroring a new PR such as #20374 needs only GitHub. The previous gap, with no build queued between 2026-09-24 09:18 UTC and 2026-09-25 12:44 UTC, also started around 09:15-09:20 UTC, in case something on the host runs on a schedule. ### When restarting On startup the bot only looks at PRs updated in the last `--backlog` hours (default 24), plus PRs that still have a `ci_` branch in the mirror. Covering the whole outage needs a `--backlog` that reaches back to 2026-10-06 (about 270 hours today, roughly 100 builds); otherwise each stuck PR needs a new push or comment. cc @xushiyan @yihua @yanghua - I could not tell who runs the bot host, so I am pinging the people who set up the Azure pipelines, the `Azure CI` status check and the `apachehudi-ci` org. -- 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]
