[ClusterLabs] design of a two-node cluster
lists at alteeve.ca
Tue Dec 15 21:36:58 EST 2015
On 15/12/15 06:35 PM, Lentes, Bernd wrote:
> Digimer wrote:
>> To prevent a dual fence, where both nodes fence each other when
>> communication between the nodes fail but the nodes are otherwise
>> healthy, you need to set a fence delay against one of the nodes. So when
>> this happens, if the delay is on node 1, this will happen;
>> Node 1 looks up how to fence node 2, sees no delay and fences
>> immediately. Node 2 looks up how to fence node 1, sees a delay and
>> pauses. Node 2 will be dead long before the delay expires, ensuring that
>> node 2 always loses in such a case. If you have VMs on both nodes, then
>> no matter which node the delay is on, some servers will be interrupted.
> Ok. The setp for the ressources is clear now.
> What is about fencing, quorum, STONITH ?
Fencing and stonith are the same thing, just different names.
Quorum is a tool for when nodes are working properly. It's not used in
2-node clusters (or rather, nodes are always quorate).
Fencing/stonith is a tool to force a node that has entered an unknown
state into a known state. It is always required.
> Are there other possibilities than the mentioned fence delay ?
To avoid a dual fence? Depends on the device. Some devices prevent two
log ins and so it's not as much of a concern. Generally though, a delay
with a 2-node cluster is strongly recommended.
> When the interconnect between the nodes goes down, i have a split brain.
No, you have a split-brain when the two nodes both try to proceed
without coordinating with each other, which fencing prevents.
> In every node i have two eth's which will be combined to a bond.
> I have two possibilities to fence, HP ILO and an APC Power switch which is configurable over network.
Use both; IPMI (iLO) first, switched PDUs as the backup in case iLO fails.
> What i read already is that in a two-node cluster i have to diable quorum.
> What is about a quorum disk ? I also read about "tiebrakers" or that STONITH is magically able to chosse the right
> node to fence (i can't believe that).
There is no magic. The node with the delay will win.
Quorum disk allows for quorum, but it is more trouble than it is worth
and it still doesn't remove the need for fencing.
> Helmholtz Zentrum Muenchen
> Deutsches Forschungszentrum fuer Gesundheit und Umwelt (GmbH)
> Ingolstaedter Landstr. 1
> 85764 Neuherberg
> Aufsichtsratsvorsitzende: MinDir'in Baerbel Brumme-Bothe
> Geschaeftsfuehrer: Prof. Dr. Guenther Wess, Dr. Nikolaus Blum, Dr. Alfons Enhsen
> Registergericht: Amtsgericht Muenchen HRB 6466
> USt-IdNr: DE 129521671
> Users mailing list: Users at clusterlabs.org
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org
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