Quote from: bytiges on February 09, 2018, 12:49:24 PM

Quote from: papampi on February 09, 2018, 09:56:23 AM

Thanks mate
I know how it works, and I can load watchdog manually but couldn't get it to auto start on system boot.
Also no idea which modules to remove from blacklist .

watchdog install fails because /etc/init.d/gpu by th00bear does not complain to lsb requirements. You need to add the following to the header (from line 3 onwards):

Code:

### BEGIN INIT INFO
# Provides: gpu
# Short-Description: th00ber gpu
# Required-Start:
# Required-Stop:
# Should-Start:
# Should-Stop:
# Default-Start: 2 3 4 5
# Default-Stop: 0 1 6
### END INIT INFO

Also there's an unbalanced quote in /lib/systemd/system/watchdog.service. The execstartpre line should have a single quote at the end ('):

Code:

ExecStartPre=/bin/sh -c '[ -z "${watchdog_module}" ] || [ "${watchdog_module}" = "none" ] || /sbin/modprobe $watchdog_module'

after that apt-get install watchdog (or apt-get -f install to fix it).

then

#systemctl start watchdog

then check that it's running with

#systemctl status watchdog

and finally enable at boot with

#systemctl enable watchdog

and don't forget to fine-tune with /etc/watchdog.conf
also a hardware driver needs to be specified in /etc/default/watchdog.
The drivers are in /lib/modules/$(uname -r)/kernel/drivers/watchdog
Usually (proper) mainboard manuals have the type of watchdog installed. On intel hardware, usually iTCO_wdt works.
You can use the softdog.ko module, but it's not as effective

After all this, I tried rebooting my nvoc rig and watchdog start is stuck in starting state. If I stop and start it by hand on the console, it will start but it won't start by itself. Perhaps there's an issue related to nvoc or not (my other ubuntu boxes start normally).

I tried looking for a fix for about an hour now, and still nogo. :/

EDIT: RE: Blacklisted modules:
There's no need to remove blacklisted modules. /etc/default/watchdog contains the module name to load and watchdog start will load it regardless of blacklist. This is something that cannot be automated because there's not reliable way to know if a hardware watchdog module is present. Some modules load and assume they work, creating the /dev/watchdog* devices. Other will check and fail. It's a mess in general.

I was able to install/update watchdog with removing /etc/inint.d/gpu and then added it back, but your solution is definitely better to fix that script.
Then it got stuck at start at boot time too. Fixing the watchdog.service did not fix it either, may be we add it to rc.local so it start at boot?

Also I read there should be a watchdog in /dev, but after manual start it does not show up in there too.

Edit:
remove/purged then installed watchdog again, now /dev/watchdog show up with manual start

Edited /etc/init.d/watchdog:

Code:

run_watchdog=1

And still wont start at boot

Code:

● watchdog.service - watchdog daemon
Loaded: loaded (/lib/systemd/system/watchdog.service; static; vendor preset: enabled)
Active: inactive (dead)