Date: Tue, 2 Apr 2019 09:00:48 -0700
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: r.zharkov@postgrespro.ru, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #15727: PANIC: cannot abort transaction 295144144, it was
 already committed
Message-ID: <20190402160048.2rohtgdmwsbkhhw6@alap3.anarazel.de>
References: <15727-0be246e7d852d229@postgresql.org>
 <20190402152925.kwt4xm7sfnay3t7t@alap3.anarazel.de>
 <30650.1554219866@sss.pgh.pa.us>
 <20190402154720.skjy27z7h5xildfm@alap3.anarazel.de>
 <30912.1554220298@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <30912.1554220298@sss.pgh.pa.us>
Precedence: bulk

Hi,

On 2019-04-02 11:51:38 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> > On 2019-04-02 11:44:26 -0400, Tom Lane wrote:
> >> What that seems to indicate is that the "unexpected table_lock_tuple
> >> status" error was thrown during commit, which seems pretty odd.
> 
> > I suspect that's a deferred trigger. But that code obviously could throw
> > errors, so we gotta handle that correctly.
> 
> Deferred triggers execute before the real transaction commit, not during
> the critical section surrounding where we set the bit in pg_clog.
> But this error is seemingly getting thrown from within that critical
> section, otherwise we wouldn't have inconsistent xact status afterwards.

I don't think it was thrown during the critical section - we'd otherwise
see the ERROR about unexpected lock status converted to a PANIC (and the
"cannot abort transaction, already committed" is a PANIC itself).  I
suspect it's more likely that we corrupt the xact.c status somewhere -
but it's hard to say without knowing more about the workload.

Greetings,

Andres Freund