[ClusterLabs] Problem using IPMI for fencing

[Andrei Borzenkov]

В Tue, 3 Mar 2015 19:14:14 +0100
Jose Manuel Martínez <jose.martinez at fcsc.es> пишет:

> Hello everybody.
> 
> I'm trying to build an active/passive cluster for the Lustre filesystem.
> Pacemaker is working fine in most situations except one: If a node goes 
> out of power in a 2-node cluster, and I am using fence_ipmilan as 
> fencing resource (for HP iLO2), the alive node is not able to takeover 
> the resources of the failed node. It tries to check the fencing device 
> trying to reboot it, but as the node is dead (no power), the IPMI 
> interface does not answer.
> 
... 
> ...but these operations never happen. If it can't fence the dead node, 
> the resources are not takeovered.
> 
> This is an infinite loop and resources are never takeovered.
> 

This is by design. If cluster failed to fence node, it cannot takeover
- this means potential data corruption.

> Is there a way to say the cluster what to do in this case?.
> 

Your options are

- declare node down manually. It can be done using "crm node clearstate
<node>". You are responsible for ensuring that node is really not
running any resource.

- use different or additional fencing resource like managed power
  outlet or SBD to add redundancy to fencing methods.