Horton

Re: Race conditions in 019_replslot_limit.pl

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Masahiko Sawada <sawada.mshk@gmail.com>, Kyotaro Horiguchi <horikyota.ntt@gmail.com>, hlinnaka@iki.fi, pgsql-hackers@postgresql.org
Date: 2022-02-17T03:46:26Z
Lists: pgsql-hackers 
Hi,

On 2022-02-16 22:11:30 -0500, Tom Lane wrote:
> So (a) it broke around 48 hours ago, which is already a useful
> bit of info

Indeed. 2f6501fa3c54bbe4568e3bcccd9a60d26a46b5ee seems like the obvious commit
to blame.

We document before_shmem_exit hooks as
	/*
	 * Call before_shmem_exit callbacks.
	 *
	 * These should be things that need most of the system to still be up and
	 * working, such as cleanup of temp relations, which requires catalog
	 * access; or things that need to be completed because later cleanup steps
	 * depend on them, such as releasing lwlocks.
	 */
and several before_shmem_exit callbacks use lwlocks.

But right now I'm not seeing what prevents us from throwing a FATAL error
while holding an lwlock?


> , and (b) your animals seem far more susceptible than
> anyone else's.  Why do you suppose that is?

Flaviventris, serinus use the newest snapshot of gcc available in debian, one
with -O0, the other with O3.

desmoxytes, idiacanthus, komodoensis all have JIT forced for every query.

They all run on the same host - looking at stats it doesn't look crazily
overcommitted or such. But it might have more scheduler variance than most?


Greetings,

Andres Freund

Commits

  1. Revert 019_replslot_limit.pl related debugging aids.

  2. Don't fail for > 1 walsenders in 019_replslot_limit, add debug messages.

  3. Add retries for further investigation of 019_replslot_limit.pl failures.

  4. Add further debug info to help debug 019_replslot_limit.pl failures.

  5. Add temporary debug info to help debug 019_replslot_limit.pl failures.

  6. Move replication slot release to before_shmem_exit().

  7. Server-side fix for delayed NOTIFY and SIGTERM processing.