Re: lockup in parallel hash join on dikkop (freebsd 14.0-current)

On Thu, Aug 31, 2023 at 2:32 PM Thomas Munro <thomas(dot)munro(at)gmail(dot)com> wrote:
> On Thu, Aug 31, 2023 at 12:16 AM Tomas Vondra
> <tomas(dot)vondra(at)enterprisedb(dot)com> wrote:
> > I have another case of this on dikkop (on 11 again). Is there anything
> > else we'd want to try? Or maybe someone would want access to the machine
> > and do some investigation directly?

Hmm. No conclusion tonight but I think it's weird. We have:

bsd(at)generic:/mnt/data/buildfarm $ ps x -O wchan | grep 52663
52663 select - Is 0:07.40 postgres: bsd regression [local] SELECT (postgres)
52731 select - Is 0:00.09 postgres: parallel worker for PID 52663
(postgres)
52732 select - Is 0:00.06 postgres: parallel worker for PID 52663
(postgres)
81525 piperd 0 S+ 0:00.01 grep 52663

wchan=select means sleeping in poll()/select().

bsd(at)generic:/mnt/data/buildfarm $ procstat -i 52732 | grep USR1
52732 postgres USR1 P-C
bsd(at)generic:/mnt/data/buildfarm $ procstat -j 52732 | grep USR1
52732 100121 postgres USR1 --

We have a signal that is pending and not blocked, so I don't
immediately know why poll() hasn't returned control.