On Wed, Nov 16, 2016 at 09:33:44AM -0500, Justin Stephenson wrote:
On 11/16/2016 06:19 AM, richard.y.collins(a)aib.ie wrote:
> We are still seeing random intermittent stoppages of the SSSD service.
>
> Following Justin's advice I setup an stap script to catch what was killing sssd
and it's related processes.
>
> It turns out sssd is killing itself. See stap output below. Would there be any
reason for this?
>
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd (pid:13831) by sssd uid:0
I found it a bit strange that sssd sends a signal to itself here. It
almost looks like a graceful shutdown...
I agree with Justin that sssd debug logs would provide a bit more
insight here as well.
Knowing what version you run might help, too.
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_sudo
(pid:13835) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_sudo (pid:13835) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_pam (pid:13834) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_pam (pid:13834) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_nss (pid:13833) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_nss (pid:13833) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_be (pid:13832) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd_be (pid:13832) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to sssd (pid:13831) by sssd uid:0
> [Wed Nov 16 10:39:34 2016] SIGTERM was sent to oddjobd (pid:10391) by oddjobd uid:0
> [Wed Nov 16 10:39:35 2016] SIGTERM was sent to oddjobd (pid:22422) by oddjobd uid:0
> _______________________________________________
> sssd-users mailing list -- sssd-users(a)lists.fedorahosted.org
> To unsubscribe send an email to sssd-users-leave(a)lists.fedorahosted.org
>
The SSSD debug logs should give some indication of what's happening around
the Nov 16 10:39:34 timeframe.
The primary SSSD service sends heartbeat pings to other SSSD services, if
there is no response from 3 pings then SSSD will attempt to send a SIGTERM
to the service.
Note:
https://fedorahosted.org/sssd/wiki/DevelTips#WhenIdebuganSSSDprocessinade...
The 'timeout' value in sssd.conf configures the time interval between pings
which defaults to 10 seconds but can be increased(it can be added to each
section of sssd.conf).
For example:
[sssd]
timeout = 60
[nss]
timeout = 60
[pam]
timeout = 60
[sudo]
timeout = 60
[domain/MYDOMAIN]
timeout = 60
Kind regards,
Justin Stephenson
This is correct and useful information but only valid up to and
including sssd-1.13. In 1.14, we switched to 'watchdog' which means the
services are no longer watched using ping-pongs from the monitor, but
have a built-in timer that resets periodically. If the timer doesn't
reset within the 'timeout' interval, the service kills itself.
I amended the DevelTips page to include this information.