[alicefr]

Debugging TestMidRolloutImageChange test

[ptalgulk01]

Thanks for looking into this! The flake is real, but the CI logs from the two recent failures (Oct 16, Oct 19) from other PR.

• The poll caught the first node rebooting both times (78s/114s after the v1 patch, v2 switched 13–30ms later) — it can't miss it, since Rebooting lasts minutes and we poll every 2s.
• The failing "other" node never booted v1 (rollback entry = original image, events show Staging v1 → Staging v2). The revocation path works fine.
• What actually happened: after rebooting into v2, the node's daemon staged v2 again and rebooted again — same image, second reboot, with no controller involvement in between (controller logs confirm no new drain/approval). Boot count went 2→4.

So this looks like a real daemon bug, not a test race — likely fallout from the cancelled v1 stage (the first node, which staged v1 cleanly, never double-reboots).

[HarshwardhanPatil07]

some ci unit test seems to be failing here https://github.com/bootc-dev/bootc-operator/actions/runs/39783566535/job/119840734257?pr=251

[alicefr]

some ci unit test seems to be failing here https://github.com/bootc-dev/bootc-operator/actions/runs/39783566535/job/119840734257?pr=251

That's not important. It is a unit test that I added to try one scenarios, but it won't be merged if we don't confirm that's the real cause

[alicefr]

Trying to reproduce the flaky test by triggering multiple times the ci