What´s the best situation?

[Filippo_1]

Hello Guru´s

I´d like to know which is the best situation to configure a 6 node cluster considering the environment below:

Site 1:

- 1 Superdome Server with 4 Npartitions
- Eva 8000

Site 2(backup site):

- 1 Superdome Server with 2 Npartitions
- Eva 3000

Every server is on the same subnet and all the disk storages can see each other.

So the thing is, shoud I use Quorum Server? Or Not? What happens if my Site 1 crashes? Will the quorum server keep the cluster UP and RUNNING on Site 2 (backup site) only with 2 nodes out of 6?

Thanks,

Filippo

[Steven E. Protter]

Shalom,

Lock disk is only supported for two nodes cluster. Lock disk only works if both nodes have access to the same disk. That appears to be the case with your situation.

Beyond this you need a quorum server.

Properly configured SG depening on which type you use will keep high availability even at great distance.

SEP

[Rita C Workman]

If your running two sites, then I must presume your not building a 6 node local cluster but rather a 4 node cluster and a 2 node cluster.
So, then quorum server would work for the 4 node cluster and you could do lockdisk for the 2 node cluster.

Suppose my next question would be are you campus/metro or higher level of clustering ?

Just an observation,
Rita

[Filippo_1]

Hi,

I´ve been reading the serviceguard documentation and it says that quorum server shoud be used.. and doesn´t says that must be used. I thought of not use the quorum server at all.

The question that comes to my mind is, what happens if 4 of my 6 node cluster server fails.. and only 2 nodes remain running along with the quorum server?? Will the cluster keep working only with 2 nodes or the cluster will halt?

See, on the situation above more than 50% of my nodes are down! So in my opinion the cluster will halt and the quorum server will be useless..

Regards,

Filippo

[Duncan Edmonstone · Accepted solution]

Filippo,

You are correct - if you have a 6-node cluster and 4 nodes fail simultaneously (because they are all on the same power or at the same site) then the remaining 2 nodes will not be able to form a cluster, because they cannot achieve quorum (>50% of nodes present). If you are staisfied that the 4 nodes at site 1 are really down you can of course manually restart the cluster with just the 2 nodes at site 2 as in:

# cmruncl -v -n node5 -n node6

The quorum server is only used when a cluster is unable to achieve quorum because exactly 50% of nodes are present - at all other times standard cluster membership rules apply.

Now if the nodes failed 1 at a time, you'd be OK because in each case you'd still be able to achieve a quorum of >50% based on the previous cluster membership, but given that the systems are all in the same complex, and at the same site thats not likely to be the case.

A better solution here might be 3 nodes (nPars) at each site - then a site failure will cause the quorum server to be used and your cluster will stay up.

Oh, and Steven - a cluster lock disk will work in clusters of up to 4 nodes:

http://docs.hp.com/en/B3936-90100/ch03s02.html#d0e2502

HTH

Duncan

[Filippo_1]

Nice answer Duncan. Points for you!

That solves my problem.

Thanks.

Filippo