[ClusterLabs] Two-Node OCFS2 cluster keep rebooting each other

Digimer lists at alteeve.ca
Tue Jun 9 23:39:51 EDT 2015

On 09/06/15 11:04 PM, Jonathan Vargas wrote:
> After reading a bit and chatting at the #suse IRC channel, I'm starting
> to think that for my use case, stonith would not be required.

Always wrong.

Fencing/Stonith is always required. Doubly so with shared storage.

> Since there are only two nodes accesing the same disk, and no additional
> redundant cluster channels, what will work meanwhile is disabling stonith.

You will get a split-brain, it is just a matter of time.

> I consider this also because the application that the Web server hosts,
> is stateless, so if no incoming requests are received by the Load
> Balancer, then, no write operations could be performed.
> At the end, this was the way it worked before HA. What do you think?

If the program(s) can run independently without coordination, you do not
need HA. If not, you need fencing.

Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without
access to education?

More information about the Users mailing list