Understanding ipmitool(1) watchdog

ipmi

I saw a following crontab entry on a server:

/usr/bin/ipmitool mc watchdog reset

According to ipmitool manual, it will reset the IPMI Management Controller watchdog timer. As I understand, this command resets the watchdog timer back to 300s. Once the timer reaches 0, the system is rebooted. However, is there a some sort of watchdog automatically set to place during the installation of ipmitool? What parameters are checked(by default) by this watchdog?

PS. Unfortunately I don't have a server with IPMI chip so I can't test this myself.

Best Answer

Ok, so here is the deal.

1) "As I understand, this command resets the watchdog timer back to 300s." Yes this command /usr/bin/ipmitool mc watchdog reset will do just that reset the ipmitool back to its "Initial Countdown" value which is 300s by default.

2) Once the timer reaches 0, the system is rebooted. Well that depends how often your reset cron job runs. By the default when you enable watchdog daemon these things happen:

You set it to run after you enable ipmi watchdog in BIOS and start the ipmi service with "service ipmi start", which activates both ipmi drivers and ipmi_watchdog driver.

Then after you activate the actual watchdog service by issuing the command:

ipmitool mc watchdog on

the watchdog timer would reboot the hardware every 300 seconds, but at the same time a software timer resets the hardware watchdog timer every 5 seconds. So it is constantly being reset back to 300 seconds. It is done for the purpose where if your machine gets hung the hardware timer will countdown to 0 and cause a reboot within 300 seconds. So i am not really sure why you have that line in the cron, since if the systems get hung cron will not be called to reset watchdog back to 300s. Maybe initial software clock service was not configured correctly or some other related issues.

Hope this helps.

Related Topic