[ClusterLabs] Suicide stonith
kwenning at redhat.com
Fri Apr 15 08:42:39 EDT 2016
On 04/15/2016 02:21 PM, Digimer wrote:
> On 15/04/16 03:12 AM, Andrey Rogovsky wrote:
>> How to suicide works?
>> I want to reboot node when it lost networking and get out from cluster.
> Stonith can't really on the failed/lost node behaving in any predictable
> way. So self-fencing, ssh fencing and what not are not going to work.
> Now, if your machine has a hardware watchdog timer, that can work. I've
> not used it myself, but others here may have insight on that.
Basically you can use sbd (https://github.com/ClusterLabs/sbd) to take
care of the watchdog and to stop triggering it in cases where you want
the node to reliably reboot.
Integration with pacemaker might look a little different depending
on which distribution or pacemaker-version you are on.
Primary intent why sbd came into life was to distribute poison pills
via shared storage but you can use it with a hardware-watchdog
More information about the Users