[ClusterLabs] Antw: Re: [EXT] Re: Two node cluster without fencing and no split brain?

[kgaillot at redhat.com]

On Mon, 2021-07-26 at 12:21 -0400, john tillman wrote:
> > On Mon, Jul 26, 2021 at 4:53 PM john tillman <johnt at panix.com>
> > wrote:
> > > > > 
> > > > > Maybe explain how it should work:
> > > > > If the two nodes cannot rech each other, but each can reach
> > > > > the ping
> > > > > node,
> > > > > which node has the quorum then?
> > > > > 
> > > > 
> > > > Guess both - which is what is played down as 'disadvantage' in
> > > > the
> > > > description
> > > > below ;-)
> > > > 
> > > > 
> > > 
> > > It is not perfect, I agree, but it may be better than nothing at
> > > all.
> > > 
> > > As for how it worked in my head:
> > > 
> > > I would have used the switch's IP address as the "ping" tie
> > > breaker; a
> > > common connection point between the two nodes.  My assumption is
> > > that if
> > > there was network loss by Node A then it would lose quorum. In
> > > the mean
> > > time Node B would still reach the switch, achieve quorum, and
> > > start/move
> > > resources.
> > > 
> > 
> > You seriously misunderstand what "split brain" means. Once more -
> > both
> > nodes are up, neither node can contact another node, both nodes can
> > ping switch. What should each node do now?
> > _______________________________________________
> > Manage your subscription:
> > https://lists.clusterlabs.org/mailman/listinfo/users
> > 
> > ClusterLabs home: https://www.clusterlabs.org/
> > 
> > 
> 
> 
> They would continue running their resources and we would have split
> brain.
> 
> So there is no safe way to support a two node cluster 100% of the
> time. 
> But when all you have are two nodes and a switch ... well, when life 

If it's an intelligent (SNMP-managed) switch, you could use the
fence_ifmib agent to fence a node by cutting off its network access
(which is safe if the node doesn't have access to a shared resource
such as a disk by means other than the network being cut off).

> gives
> you lemons ...
> 
> Thank you again for the response and I apologize for beating a dead
> horse
> here.
> 
-- 
Ken Gaillot <kgaillot at redhat.com>