CLA bot posts message after CLA is already signed

[nicksnyder]

See how CLA bot commented on #2135 multiple times. Chris signed the CLA Feb 2, 2019 13:44:10.

I wondered if this had to do with the pod restarting.

The most recent comment was Feb 27, 7:16 AM PT.

Describing the pod shows that it started 1.5 hours earlier:

State:          Running
      Started:      Wed, 27 Feb 2019 05:41:12 -0800
    Last State:     Terminated
      Reason:       Error
      Exit Code:    255
      Started:      Thu, 21 Feb 2019 14:07:26 -0800
      Finished:     Wed, 27 Feb 2019 05:40:51 -0800
Restart Count:  1

So restarting is a plausible theory if it takes 1.5 hours to resync everything.

@kevinburke would you expect the bot to re-post after being restarted?

[dadlerj]

@nicksnyder have you seen this reoccur since then?

[nicksnyder]

The last occurrence that I observed was March 18 on that same PR:
#2135 (comment)

[dadlerj]

Okay if I close for now until we see it again then?

[nicksnyder]

Sure

[nicksnyder]

We just fixed CLA bot and saw it post a duplicate message on #3835 (comment). The pod restarted between the messages so that might be the cause (i.e. we don't persist any state between restarts)

[eseliger]

Possibly fixed in sourcegraph/maintainerbot#1
awaiting result here: #8271
I realized that cla-bot has a super long warmup phase of ~5h (around 2s per issue * ~8300 issues/prs). So revisiting in a while...

[eseliger]

It looks like it's working. Closing for now

[eseliger]

Found a proof PR: #3835