Horton

Re: Spurious "apparent wraparound" via SimpleLruTruncate() rounding

Noah Misch <noah@leadboat.com>

View thread View on postgr.es
From: Noah Misch <noah@leadboat.com>
To: Andrey Borodin <x4mmm@yandex-team.ru>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Tomas Vondra <tomas.vondra@2ndquadrant.com>, pgsql-hackers@postgresql.org
Date: 2021-01-10T09:43:48Z
Lists: pgsql-hackers

Attachments

On Sun, Jan 10, 2021 at 11:44:14AM +0500, Andrey Borodin wrote:
> > 10 янв. 2021 г., в 03:15, Noah Misch <noah@leadboat.com> написал(а):
> > 
> > No; it deletes the most recent ~1B and leaves the older segments.  An
> > exception is multixact, as described in the commit message and the patch's
> > change to a comment in TruncateMultiXact().
> 
> Thanks for clarification.
> One more thing: retention point at 3/4 of overall space (half of wraparound) seems more or less random to me. Why not 5/8 or 9/16?

No reason for that exact value.  The purpose of that patch is to mitigate bugs
that cause the server to write data into a region of the SLRU that we permit
truncation to unlink.  If the patch instead tested "diff > INT_MIN * .99", the
new behavior would get little testing, because xidWarnLimit would start first.
Also, the new behavior wouldn't mitigate bugs that trespass >~20M XIDs into
unlink-eligible space.  If the patch tested "diff > INT_MIN * .01", more sites
would see disk consumption grow.  I think reasonable multipliers range from
0.5 (in the patch today) to 0.9, but it's a judgment call.

> Can you please send revised patches with fixes?

Attached.

Commits

  1. Prevent excess SimpleLruTruncate() deletion.

  2. Fix unlinking of SLRU segments.

  3. Defer flushing of SLRU files.

  4. Change XID and mxact limits to warn at 40M and stop at 3M.