For this to work, a kernel driver for the aformentioned device needs to be loaded. The module may be found in: /lib/modules/`uname -r`/kernel/drivers/char/watchdog It can be configured to try and fix things (such as networking) before triggering a reboot. As 'repair.sh' may be able to do. See the examples included in /usr/doc/watchdog-5.6/examples This package also contains the 'wd_keepalive' program by Marcel Jansen which "can be run during critical periods when the normal watcdog shouldn't be run. It will read from the same configuration file; it will do no checks but will keep writing to the device. To enable the watchdog: make sure a kernel module gets loaded for it (at boot) such as the 'softdog' or an actual hardware driver... Then add the following to rc.local: # Start the watchdog timer deamon if [ -x /etc/rc.d/rc.watchdog ]; then /etc/rc.d/rc.watchdog start fi