Re: Linux SYSTEMD and Oracle proceses

Don't remember if it was oracle or postgres but something introduced me to their RemoveIPC=yes/no
setting in logind.conf(5)
This yanks the user's semaphores and shared memory if their last session logs out.
(Unless the user is considered 'system' user which is a vague concept based on uid - maybe compile time or maybe based on SYS_UID_MIN/MAX in login.defs(5))
There seems to be also an interesting KillUserProcesses= setting there and probably others
sprinkled around in various conf files and poorly documented and/or having interactions with well-known unix conventions. My pet peeve was systemd silently unmounting my archivelog destinations: https://lists.freedesktop.org/archives/systemd-devel/2017-June/039054.html

Andre

On Fri, Nov 26, 2021 at 1:08 AM Lyall Barbour <lyallbarbour_at_sanfranmail.com> wrote:
>
> Hello,
> Wanted to see if anyone else had these two issues with Oracle databases running on later versions of Linux implementing SYSTEMD configuration.
>
> 1) Oracle grid with Restart running the automated reboots, have the instance crash during server reboot (and probably ohasd crash) when the configuration is started under a user slice, rather then a system slice.
>
> 2) if a DBA manually starts an instance with sqlplus, and then their ssh tty is killed later on, the database crashes, at the instant the session is killed.
>
> I've had two SRs for the Grid Restart situation and our Unix Admins have a work around. But the Oracle Techs are not even able to understand the situation. It's really infuriating. The second situation is very new that we've noticed at our company over the past month.
>
> TIA
> Lyall Barbour
>
> --
> Sent from my Android phone with mail.com Mail. Please excuse my brevity.
> -- http://www.freelists.org/webpage/oracle-l

--
http://www.freelists.org/webpage/oracle-l