Re: OOM killer and rocky linux

https://www.postgresql.org/docs/current/kernel-resources.html#LINUX-MEMORY-OVERCOMMIT
19.4. Managing Kernel Resources
postgresql.org

Hello!

Disabling overcommit is recommended.

Best regards

> kaido vaikla <kaido(dot)vaikla(at)gmail(dot)com> kirjoitti 9.11.2023 kello 14.57:
>
> Hi,
>
> The problem is probably somwhere between pg and linux.
> Seems like OOM killer on rocky linux behaves in a different way
> than on RHEL.
>
> For OOM killer invoking I used:
> set work_mem=50GB;
> select * from generate_series(1,100000000) order by random();
>
> pg runs as systemd service.
>
> Example 1:
> ==========
> pg 15.2
> o/s Red Hat Enterprise Linux Server release 7.9 (Maipo)
>
> pg log:
>
> 2023-07-05 12:56:28.848 EEST [:::119279:postmaster] LOG: server process (PID 120401) was terminated by signal 9: Killed
> 2023-07-05 12:56:28.848 EEST [:::119279:postmaster] LOG: terminating any other active server processes
> 2023-07-05 12:56:28.850 EEST [:::119279:postmaster] LOG: all server processes terminated; reinitializing
> 2023-07-05 12:56:28.885 EEST [:::120421:startup] LOG: database system was interrupted; last known up at 2023-07-05 12:55:34 EEST
> 2023-07-05 12:56:28.899 EEST [:::120421:startup] LOG: database system was not properly shut down; automatic recovery in progress
> 2023-07-05 12:56:28.904 EEST [:::120421:startup] LOG: redo starts at 1/D0002E8
> 2023-07-05 12:56:28.904 EEST [:::120421:startup] LOG: invalid record length at 1/D000320: wanted 24, got 0
> 2023-07-05 12:56:28.904 EEST [:::120421:startup] LOG: redo done at 1/D0002E8 system usage: CPU: user: 0.00 s, system: 0.00 s, elapsed: 0.00 s
> 2023-07-05 12:56:28.912 EEST [:::120422:checkpointer] LOG: checkpoint starting: end-of-recovery immediate wait
> 2023-07-05 12:56:28.918 EEST [:::120422:checkpointer] LOG: checkpoint complete: wrote 3 buffers (0.0%); 0 WAL file(s) added, 0 removed, 0 recycled; write=0.002 s, sync=0.002 s, total=0.013 s; sync files=2, longest=0.001 s, average=0.001 s; distance=0 kB, estimate=0 kB
>
>
> Instance is reinitializing
>
> Example 2:
> ==========
> pg 15.2
> o/s Rocky Linux release 9.2 (Blue Onyx)
>
> pg log and systemd log in chronological order (juuni == 06 ):
>
> juuni 30 14:19:11.491833 pgdb-forecast kernel: oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/system.slice/postgresql-15.service,task=postmaster,pid=2575533,uid=26
> juuni 30 14:19:11.491839 pgdb-forecast kernel: Out of memory: Killed process 2575533 (postmaster) total-vm:9716200kB, anon-rss:5113080kB, file-rss:0kB, shmem-rss:16828kB, UID:26 pgtables:10256kB oom_score_adj:0
> juuni 30 14:19:11.491846 pgdb-forecast kernel: oom_reaper: reaped process 2575533 (postmaster), now anon-rss:0kB, file-rss:0kB, shmem-rss:16828kB
> juuni 30 14:19:11.478410 pgdb-forecast systemd[1]: postgresql-15.service: A process of this unit has been killed by the OOM killer.
> 2023-06-30 14:19:11.505 EEST [:::2517202:postmaster] LOG: received fast shutdown request
> 2023-06-30 14:19:11.518 EEST [:::2517202:postmaster] LOG: aborting any active transactions
> 2023-06-30 14:19:11.564 EEST [:::2517202:postmaster] LOG: server process (PID 2575533) was terminated by signal 9: Killed
> 2023-06-30 14:19:11.564 EEST [:::2517202:postmaster] LOG: terminating any other active server processes
> 2023-06-30 14:19:11.633 EEST [:::2517202:postmaster] LOG: abnormal database system shutdown
> 2023-06-30 14:19:11.939 EEST [:::2517202:postmaster] LOG: database system is shut down
> juuni 30 14:19:11.940216 pgdb-forecast systemd[1]: postgresql-15.service: Main process exited, code=exited, status=1/FAILURE
> juuni 30 14:19:11.940255 pgdb-forecast systemd[1]: postgresql-15.service: Killing process 2517203 (postmaster) with signal SIGKILL.
> juuni 30 14:19:11.940700 pgdb-forecast systemd[1]: postgresql-15.service: Failed with result 'oom-kill'.
> juuni 30 14:19:11.940884 pgdb-forecast systemd[1]: postgresql-15.service: Consumed 46min 16.612s CPU time.
>
>
> Instance received fast shutdown request.
> And kernel sends kill twice: process 2575533, process 2517203.
>
> Googeling this problem gives me nothing, and i'm not sure,
> is it caused of pg or linux kernel.
>
> br
> Kaido
>
>
>
>
>