topic Re: Clustering RedHat Linux in Operating System - Linux

Clustering RedHat Linux

Andrew Moody_1 — Tue, 26 Jan 2010 12:55:58 GMT

Gents

I understand the Serviceguard for Linux is being deprecated. We are looking to implement a RedHat cluster over about 300km. Storage will be replicated asynchronously. Do you know if this sort of distance cluster is doable in RedHat Cluster Suite or do we have to use Symantec or some other vendor to get clusters working over this distance with RedHat?

Any pointers would be gratefully recieved.

Re: Clustering RedHat Linux

Ivan Ferreira — Tue, 26 Jan 2010 20:32:05 GMT

Please check this document:

docs.hp.com/en/5900-0380/5900-0380.pdf

"There is, as yet, no supported configuration or solution on RHCS that can be used for setting up an extended distance cluster."

Re: Clustering RedHat Linux

Steven E. Protter — Wed, 27 Jan 2010 00:52:19 GMT

Shalom,

Such a scenario, unsupported might be possible with enough band width. Otherwise RHCS won't do it for you.

A manual fail over scheme might work with continuous data replication through a fast line.

SEP

Re: Clustering RedHat Linux

Andrew Moody_1 — Wed, 27 Jan 2010 10:02:13 GMT

Thanks Gents.

I had already downloaded that document, the XDC option for Linux mentioned in the document seemed more of a Metro Cluster style option which is why I'd not taken it as gospel. I didn't believe frankly that it wouldn't be possible to cluster RHEL in a similar fashion to a Continental Cluster or Veritas Cluster Server with Global Cluster Option, in it's native form.

Two RHEL clusters, 300km apart asynchronously related so as to be able to take on each others workloads in disaster. My only problem is that the VCS option isn't supported on Itanium, so it's not an ideal fit either.

Re: Clustering RedHat Linux

Serviceguard for Linux — Wed, 27 Jan 2010 17:07:46 GMT

Even SGLX did not have the option that you are referring to - asynchronous over that distance. CLX does support asynchronous replication but I'm not sure if you are planning to use that.

I agree with the earlier response that you should set up 2 clusters and a method (with human intervention?) to switch between the clusters after a failure. When you are using asynchronous replication, with the possibility of data loss, making sure you don't have a "false failover" may be a good tradeoff.