I was doing some normal handler deploy/undeploy/editing/deploy cycles (slowly and carefully, not a stress test) and the function got stuck in Undeploying state.
The cluster is quite healthy. All buckets are online and UI works fine. I can even export the handler without trouble from UI after it's stuck in this state. See eventing.log
Then at 2019-07-19T18:09:31.330 I did "sudo killall eventing-producer eventing-consumer". The processes restarted, but it is still stuck undeploying. See eventing2.log
Out of desperation, I did "sudo killall -9 eventing-producer eventing-consumer" at 2019-07-19T18:15:58.577 and it restarted again (verified by observing PIDs of eventing-consumer and eventing-producer) change. It still is undeploying!
At this point, I'm giving up and attaching [^collectinfo-2019-07-19T125046email@example.com]
Note that I did curl 'http://click:8096/api/v1/functions/test' and it shows:
But REST API says:
|For Gerrit Dashboard: MB-35188|
|112314,2||MB-35188: Do not write to metakvappsettings path at the end of deployment||unstable||eventing||Status: MERGED||+2||+1|
|113011,3||MB-35188 MB-35462: Add retries around all metakv writes||unstable||eventing||Status: MERGED||+2||+1|