The following bug has been logged on the website:
Bug reference: 17369
Logged by: yavuz selim sertoğlu
Email address: yavuzselimsertoglu@gmail.com
PostgreSQL version: 13.5
Operating system: Red Hat Enterprise Linux Server release 7.9 (Maipo
Description:
Hi,
This morning oom killer killed a process.
/var/log/messages
Jan 18 07:44:32 dbserver kernel: [ 28935] 26 28935 56024594 55931594
448675840 278 0 postgres
Jan 18 07:44:32 dbserver kernel:
oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0-2,global_oom,task_memcg=/,task=postgres,pid=28935,uid=26
Jan 18 07:44:32 dbserver kernel: Out of memory: Killed process 28935
(postgres) total-vm:224098376kB, anon-rss:210894304kB, file-rss:0kB,
shmem-rss:12832072kB, UID:26 pgtables:438160kB oom_score_adj:0
Jan 18 07:44:44 dbserver kernel: oom_reaper: reaped process 28935
(postgres), now anon-rss:0kB, file-rss:0kB, shmem-rss:12832072kB
postgresql.log
2022-01-18 07:44:45.492 +03:::[31678]LOG: server process (PID 28935) was
terminated by signal 9: Killed
The statement is like insert into select ...
The statement also created many tmp files and total size of it is about:
1.9TB (2064 files)
du -csh base/pgsql_tmp/
1.9T base/pgsql_tmp/
Last temp file created at 07.32 for this process
-rw-------. 1 postgres postgres 972M Jan 18 07:32 pgsql_tmp28935.4635
For another process, it is 07:44 (This process was not killed by Oom Killer,
it killed by PostgreSQL because of corrupted memory when Oom Killer killed
the other process)
-rw-------. 1 postgres postgres 278M Jan 18 07:44 pgsql_tmp28937.2326
PostgreSQL did many checkpoints.
2022-01-18 07:45:04.616 +03:::[11792]LOG: checkpoint starting:
end-of-recovery immediate
2022-01-18 07:55:12.165 +03:::[11811]LOG: checkpoint starting: time
2022-01-18 08:05:12.228 +03:::[11811]LOG: checkpoint starting: time
2022-01-18 08:15:12.229 +03:::[11811]LOG: checkpoint starting: time
...
It passed about 5 hours but tmp files didn't deleted. I searched online but
everyone say don't delete temp files manually, submit a bug report.
Is it a bug? Do I need to delete them manually? or Did I misconfigure
PostgreSQL?
Best regards,
Yavuz Selim Sertoğlu