Hi Wei-Chiu, Thanks for putting this dashboard together. I agree that CI reliability and turnaround time deserve continued attention. Substantial work is already underway, both merged and pending.
On reliability, we merged 12 CI test failure fixes in the past week, including one for a production SCM liveness bug exposed by an intermittent test timeout (HDDS-16132). Progress has also been made on test runtime and reducing duplicate CI work (Codex summarized this): - HDDS-16821: TestSstFileSetReader decreased from 132s to 4.8s locally. https://github.com/apache/ozone/pull/11469 - HDDS-16820: TestBlockOutputStreamCorrectness decreased from 71s to 7.1s in CI. https://github.com/apache/ozone/pull/11457 - HDDS-16099: TestStreamBlockInputStream decreased from 77s to 28s locally. https://github.com/apache/ozone/pull/11337 - HDDS-16655: reuse the shaded filesystem jar from the build job. The cited filesystem integration job decreased from about 27 to 21 minutes. https://github.com/apache/ozone/pull/11371 - HDDS-15749: run relevant tests for PRs changing only eligible JUnit test classes. https://github.com/apache/ozone/pull/10671 - HDDS-15883 and HDDS-16714: split compatibility acceptance tests by version and Kubernetes checks by environment. https://github.com/apache/ozone/pull/11419 https://github.com/apache/ozone/pull/11406 There are still 6 relevant PRs pending merge or review. For automation, I wonder if ASF Infra / GitHub could help (e.g. agent access in CI), since this isn't an issue isolated to Ozone. Right now I just forward any flaky CI runs I see in post-commit to Claude. Thanks, Siyao On Sun, Oct 11, 2026 at 12:53 PM Wei-Chiu Chuang <[email protected]> wrote: > Hi team, > > I (meaning, Cursor) built a dashboard to track Ozone community development > velocity,in order to understand the impact of AI. > https://jojochuang.github.io/ozonedevdashboard/ > > *The good:* > A significant increase in the number of PRs opened and merged, along with > an uptick in the number of contributors in the past few months. > [image: Screenshot 2026-10-11 at 12.46.39 PM.png] > [image: Screenshot 2026-10-11 at 12.46.33 PM.png] > > *The bad:* > > > Things that caught my attention: the CI pass rate has been deteriorating. > I'm not sure why but let's fix it. > [image: Screenshot 2026-10-11 at 12.43.14 PM.png] > > > Even with AI, our review speed remains slow and shows little > improvement. The median time to the first human review is around 1 day. > [image: Screenshot 2026-10-11 at 12.44.59 PM.png] > > Time to merge a PR after an approval is actually increasing. > [image: Screenshot 2026-10-11 at 12.45.40 PM.png] > > Takeaway? > > I think we should invest more time in CI. Fix the breakage and improve CI > build time. I also think we need more automation. Perhaps after Copilot > review, an agent could automatically address the issue. For trivial issues > like code style or code conflicts, coding agents should fix them so they > don't block review/merge. > > I want to put more effort into speeding up first reviews. But I can't do > that alone. Thoughts? >
