Raising to P1 because this wasn't acknowledged yesterday. Can someone please look into this?

I suspect that I can get my CL landed by messing with the Commit-Queue flag, but I think that leaving it as-is might offer debugging insight. If that's not the case, please let me know, and I'll press buttons until I get the CQ to notice my CL :)

I believe your CL has not made it to Gerrit Index, hence CQ can't possibly see it. Verifying...

And, btw +1 for leaving as it is -- debugging insight is indeed useful here.

According to CQ logs, CL 818451 was last processed ~2017-12-13 22:10:39.628 UTC, which corresponds to prior CQ Dry run. After that, Gerrit has never returned this CL when queried for CLs with CQ>0.

Yep, this is Gerrit index issue -- see http://b/67619969#comment11.

Gerrit team says there is no debugging value, so i'm poking your CL to make sure it's re-indexed.

gob-curl -X POST https://chromium-review.googlesource.com/a/changes/818451/index
and CQ is now processing your CL.

I'm picking this bug open.

Thank you very much for looking into this!

Gerrit team has reduced the changes of 40h delay repeating itself, but ~1 hour may still happen. Unfortunately, there is no fast way to fix this on either Gerrit or CQ side, but I'm well aware of how painful this is.

I think in early Q1  I'll be able to do something about it.

Oh, no worries!

I'm already used to babysitting my CLs, because of CQ flakes. I left this one open for so long because it exposed a new (to me) infra failure mode, and I figured it might have useful debugging info. If it's not a quick fix, I wouldn't worry about it too much!

Well, your bug report isn't the first filed, though first with a 40h delay.

http://shortn/_65JdkEQGDU shows 10 CLs which were "lost" for 10+ minutes after CQ vote in last 2 weeks (1 CL per workday), 7 of them were lost for >2 hours.

And sometimes CL is "unlost" because somebody commented on a CL, which ended up refreshing index, it's kinda funny that you can post "is CQ stuck" and that will "unstuck" CQ (it actually happened).

So, I think 1 CL per workday is an understatement.

For posterity, https://screenshot.googleplex.com/zsxzFfk6cAq is screenshot of the above graph.

Issue 794387 has been merged into this issue.

Things have improved somewhat, but we still had ~7 CLs "lost" for > 10 minutes, and 3 > 1 hour in last 4 weeks (http://shortn/_E4CncUxhHX, https://screenshot.googleplex.com/Yevd4h1syuY)