Re: core dumps in auto_prewarm, tests succeed

Hi,

On 2024-01-23 22:00:01 +0300, Alexander Lakhin wrote:
> 23.01.2024 20:30, Andres Freund wrote:
> > I don't think that's viable and would cause more problems than it solves, it'd
> > make us think that we might have an old postgres process hanging around that
> > needs to be terminted before we can start up. And I simply don't see the point
> > - we already record whether we crashed in the control file, no?
>
> With an Assert injected in walsender.c (as in [1]) and test
> 012_subtransactions.pl modified to finish just after the first
> "$node_primary->stop;", I see:
> pg_controldata -D src/test/recovery/tmp_check/t_012_subtransactions_primary_data/pgdata/
> Database cluster state:               shut down
>
> But the assertion undoubtedly failed:
> grep TRAP src/test/recovery/tmp_check/log/*
> src/test/recovery/tmp_check/log/012_subtransactions_primary.log:TRAP: failed
> Assert("0"), File: "walsender.c", Line: 2688, PID: 142201

Yea, because it's after checkpointer has changed the state to "shutdowned". I
think we could add additional states, to be set by postmaster, instead of
checkpointer, for this purpose.

> As to the need to terminate a process, which is supposedly hanging around,
> I think, this situation doesn't differ in general from what we have after
> kill -9...

So? Making it more likely for postgres failing to restart successfully,
because the pid has been reused, is bad.

Greetings,

Andres Freund