On 01/24/18 20:29, Ed Greshko wrote:
On 01/24/18 20:03, Patrick O'Callaghan wrote:
> On rebooting today my system took an age to shutdown, for no apparent
> reason but possibly related to NFS mounts. Anyway, I looked at the
> journal and spotted this snippet:
>
> Jan 24 11:47:59 bree systemd[1]: Shutting down.
> Jan 24 11:47:59 bree systemd[1]: Hardware watchdog 'iTCO_wdt', version
> 0
> Jan 24 11:47:59 bree systemd[1]: Set hardware watchdog to 10min.
> Jan 24 11:47:59 bree kernel: watchdog: watchdog0: watchdog did not
> stop!
> Jan 24 11:47:59 bree systemd-shutdown[1]: Sending SIGTERM to remaining processes...
>
> So apparently systemd is setting a watchdog timer for 10 minutes, for
> some reason best known to itself. How can I change this to (say) 5
> seconds, which would be more than enough for my setup?
>
> Also, I don't know what 'watchdog did not stop!' is supposed to mean.
A few things. First, I believe some user-space processes periodically "kick"
/dev/watchdog and the kernel uses this to determine that not all the user processes
have exited on reboot. Basically trying to make sure files and such are closed
gracefully.
The 10 minute setting is a kernel parameter watchdog_thresh.
[egreshko@meimei kernel]$ cat /proc/sys/kernel/watchdog_thresh
10
If you want you should be able to set that to whatever you want with sysctl and
modifying the config file for it.
FORGET everything I've said about setting the value for the watchdog time out. My
assumption was clearly wrong (after a bit of checking...which I should have done
before responding)
Once again I've learned the value of the acronym ASS-U-ME. However, in this case
there is no U.
--
A motto of mine is: When in doubt, try it out