|From:||Andres Freund <andres(at)anarazel(dot)de>|
|To:||Michael Paquier <michael(at)paquier(dot)xyz>|
|Cc:||Ian Lawrence Barwick <barwick(at)gmail(dot)com>, Matthias van de Meent <boekewurm+postgres(at)gmail(dot)com>, Heikki Linnakangas <hlinnaka(at)iki(dot)fi>, David Zhang <david(dot)zhang(at)highgo(dot)ca>, Robert Haas <robertmhaas(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>|
|Subject:||Re: Non-replayable WAL records through overflows and >MaxAllocSize lengths|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
On 2022-12-02 14:22:55 +0900, Michael Paquier wrote:
> On Fri, Nov 04, 2022 at 09:52:39AM +0900, Ian Lawrence Barwick wrote:
> > CommitFest 2022-11 is currently underway, so if you are interested
> > in moving this patch forward, now would be a good time to update it.
> No replies after 4 weeks, so I have marked this entry as returned
> with feedback. I am still wondering what would be the best thing to
> do here..
IMO this a bugfix, I don't think we can just close the entry, even if Matthias
doesn't have time / energy to push it forward.
I think the big issue with the patch as it stands is that it will typically
cause PANICs on failure, because the record-too-large ERROR be a in a critical
section. That's still better than generating a record that can't be replayed,
but it's not good.
There's not all that many places with potentially huge records. I wonder if we
ought to modify at least the most prominent ones to prepare the record before
the critical section. I think the by far most prominent real-world case is
RecordTransactionCommit(). I think we could rename XactLogCommitRecord() to
XactBuildCommitRecord() build the commit record, then have the caller do
START_CRIT_SECTION(), set DELAY_CHKPT_START, and only then do the
That'd even have the nice side-effect of reducing the window in which
DELAY_CHKPT_START is set a bit.
|Next Message||Andres Freund||2022-12-02 17:09:13||Re: Non-replayable WAL records through overflows and >MaxAllocSize lengths|
|Previous Message||Andres Freund||2022-12-02 16:42:38||Re: Failed Assert in pgstat_assoc_relation|