On Sat, Dec 14, 2019 at 02:39:12PM +0100, Tomas Vondra wrote:
>On Sat, Dec 14, 2019 at 05:01:05PM +1300, Thomas Munro wrote:
>>On Wed, Dec 11, 2019 at 6:32 AM Tomas Vondra
>><tomas.vondra@2ndquadrant.com> wrote:
>>>The annoying thing is that it's the workers that crash, and the leader
>>>failed to notice that, so it was waiting in WaitForParallelWorkersToExit
>>>forever. Not sure what the issue is.
>>
>>That's bad. Some edge case in
>>WaitForBackgroundWorkerShutdown()/GetBackgroundWorkerPid() that is
>>b0rked perhaps? Can you reproduce it, and tell us what the stack
>>looks like, maybe on a new thread?
>
>Sure. I'll try reproducing it and get this extra info. I was planning to
>do that, but then I lost the data for the report because of a silly
>mistake. Early next week, hopefully.
>
I've tried reproducing this, but so far no luck. I've only managed to do
two runs, though, because this particular machine only has space for the
1TB of temp files on spinning rust. I wonder if it's possible to
reproduce the issue with a smaller data set? I've tried using smaller
work_mem, but then it does not trigger the issue :-(
regards
--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services