Watchdog hang on PVE reboot

[amosmes]

Hi,
I have a stand alone PVE server and today when I rebooted it I got the error:

watchdog: watchdog0: watchdog did not stop!

it was hang like that for a few minutes until I forcefully shutdown the server ... is it something to be worried about?
Is there a way around it ?

 

[Moayad]

Hi,

Did you check the syslog at that time? The syslog should gives you more info. if you see something interesting message in syslog please post them.

 

[amosmes]

Hi,
All I see is "systemd-shutdown: Watchdog running with a timeout of 10min." so I guess that if I wait 10 minutes the server would reboot.
But why is it 10 minutes ?

 

[gfngfn256]

systemd-shutdown: Watchdog running with a timeout of 10min

This is set in the configuration /etc/systemd/system.conf under RebootWatchdogSec=10min , see here for more details.

I don't believe this should be an issue on its own. More interesting would be what is causing this. If this was a one off situation, you can probably ignore.

 

[Moayad]

I would recommend waiting the full 10 minutes to allow the server to shut down or reboot naturally rather than forcefully shutting it down.

Did you enabled HA [0]?

Regarding 10 minute timeout, this is related to the watchdog's configuration in systemd. the 10 minute timeout is set to give the system sufficient time to cleanly shutdown all processes, services, and unmount filesystemss etc...


[0] https://pve.proxmox.com/pve-docs/chapter-ha-manager.html

 

[amosmes]

It is a standalone server so I didn't enable HA

 

[justinclift]

Hmmm, it'd be interesting to see what options the watchdog software was loaded with, just in case there's something unexpected happening.

Please run journalctl -b 0 | grep -i softdog, and paste the output here.

 

[amosmes]

Aug 12 12:40:27 pveserver kernel: softdog: initialized. soft_noboot=0 soft_margin=60 sec soft_panic=0 (nowayout=0)
Aug 12 12:40:27 pveserver kernel: softdog: soft_reboot_cmd=<not set> soft_active_on_boot=0

 

[justinclift]

Thanks. Nothing weird in that, so it's not a case of some strange option somehow mucking things up.