Horton

stress test for parallel workers

Justin Pryzby <pryzby@telsasoft.com>

View thread View on postgr.es
From: Justin Pryzby <pryzby@telsasoft.com>
To: pgsql-hackers@postgresql.org
Cc: Thomas Munro <thomas.munro@enterprisedb.com>
Date: 2019-07-23T16:27:03Z
Lists: pgsql-hackers 
Does anyone have a stress test for parallel workers ?

On a customer's new VM, I got this several times while (trying to) migrate their DB:

< 2019-07-23 10:33:51.552 CDT postgres >FATAL:  postmaster exited during a parallel transaction
< 2019-07-23 10:33:51.552 CDT postgres >STATEMENT:  CREATE UNIQUE INDEX unused0_huawei_umts_nodeb_locell_201907_unique_idx ON child.unused0_huawei_umts_nodeb_locell_201907 USING btree ...

There's nothing in dmesg nor in postgres logs.

At first I thought it's maybe because of a network disconnection, then I
thought it's because we ran out of space (wal), then they had a faulty blade.
After that, I'd tried and failed to reproduce it a number of times, but it's
just recurred during what was intended to be their final restore.  I've set
max_parallel_workers_per_gather=0, but I'm planning to try to diagnose an issue
in another instance.  Ideally a minimal test, since I'm apparently going to
have to run under gdb to see how it's dying, or even what process is failing.

DMI: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/21/2015
CentOS release 6.9 (Final)
Linux alextelsasrv01 2.6.32-754.17.1.el6.x86_64 #1 SMP Tue Jul 2 12:42:48 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
version | PostgreSQL 11.4 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23), 64-bit

Justin

Commits

  1. In the postmaster, rely on the signal infrastructure to block signals.

  2. Paper over regression failures in infinite_recurse() on PPC64 Linux.

  3. Hack pg_ctl to report postmaster's exit status.

  4. Re-order some regression test scripts for more parallelism.