Horton

Re: Logical WAL sender unresponsive during decoding commit

Amit Kapila <amit.kapila16@gmail.com>

View thread View on postgr.es
From: Amit Kapila <amit.kapila16@gmail.com>
To: Masahiko Sawada <sawada.mshk@gmail.com>
Cc: Andrey Borodin <x4mmm@yandex-team.ru>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2022-08-16T05:31:15Z
Lists: pgsql-hackers 
On Tue, Aug 16, 2022 at 10:56 AM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>
> On Tue, Aug 16, 2022 at 2:08 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> >
> > On Tue, Aug 16, 2022 at 9:28 AM Andrey Borodin <x4mmm@yandex-team.ru> wrote:
> > >
> > > Hi hackers!
> > >
> > > Some time ago I've seen a hanging logical replication that was trying to send transaction commit after doing table pg_repack.
> > > I understand that those things do not mix well. Yet walsender was ignoring pg_terminate_backend() and I think this worth fixing.
> > > Can we add CHECK_FOR_INTERRUPTS(); somewhere in this backtrace?
> > >
> >
> > I think if we want to do this in this code path then it may be it is
> > better to add it in ReorderBufferProcessTXN where we are looping to
> > process each change.
>
> +1
>
> The same issue is recently reported[1] on -bugs and I proposed the
> patch that adds CHECK_FOR_INTERRUPTS() to the loop in
> ReorderBufferProcessTXN(). I think it should be backpatched.
>

I agree that it is better to backpatch this as well. Would you like to
verify if your patch works for all branches or if it need some tweaks?

-- 
With Regards,
Amit Kapila.

Commits

  1. Add CHECK_FOR_INTERRUPTS while restoring changes during decoding.

  2. Add CHECK_FOR_INTERRUPTS while decoding changes.