[ClusterLabs Developers] How does the two-node Active/Passive clusters fence the faulty node, and keep the healthy node running

[Digimer]

On 2018-12-07 8:28 a.m., zhongbin wrote:
> hi.
> I have a two-node Active/Passive clusters.When heartbeat was
> lost,Active-node  and Passive-node  are both fenced(reboot) by each
> other at the same time.But I don't want the Active-node to
> reboot,because I don't want the resources on the Active-node being
> interrupted.So,I set a delay-time ,before the Passive-node reboot the
> Active-node,and the Passive-node is always fenced.
> 
> Is there any problem with this? Do you have any suggestions?

You can prevent this by adding 'delay="15"' to the stonith configuration
for the primary/active node. This way, when the passive node looks up
how to fence the active node, it sees the delay and pauses. The active
node looks up how to fence the passive node, sees no delay, and fences
immediately. The passive node dies before the delay exits.

If the active node does actually fail, then after the 15 second delay,
it proceeds with the fence action as normal.

PS - Your question came to the list three times...

-- 
Digimer
Papers and Projects: https://alteeve.com/w/
"I am, somehow, less interested in the weight and convolutions of
Einstein’s brain than in the near certainty that people of equal talent
have lived and died in cotton fields and sweatshops." - Stephen Jay Gould