2 node cluster wih 2 EVA3000s

[mc105]

Hello all

We have 2 nodes (active / passive) and 2 * EVA3000s.

I have nothing to do with these EVA’s and I don’t really know what I should be asking our SAN admins for.

We would like to put our data onto these SAN disks – but for DR purposes we should be able to loose one EVA cabinet, and the cluster can still run pointing to the other EVA.

Do I get them to present 10GB from each EVA to our active node, and then use MirrorDisk/UX for replication.

Or is the replication done from the EVAs themselves using DR groups

Thanks

[Jeff_Traigle · Accepted solution]

Depends how your environment is set up.

You could mirror across EVAs at the host with MirrorDisk/UX, but I don't think that's really as safe as it sounds. (We're doing that where I work now between VAs, but I've seen systems hang when an LV goes stale and don't consider this a good HA design. It might protect your data to a point, but it's not necessarily going to keep running without a failover... and you need to be sure you use the "-q n" option on vgchange in the control script.)

You could also take care of it at the EVA level with Continuous Access (CA). It basically mirrors between EVAs in the background. A bit more complicated to deal with in a failover situation, but I don't know any details since I haven't worked with it or had any training on it.

[mc105]

thanks Jeff

The 2nd option sounds much more what we need.

I cant find any documention on EVAs / Continuous Access with ServiceGuard.

Does anyone know where I can find out more on this.

Thanks again

[melvyn burnard]

well the straightforward method would be to mirror between EVA's using LVM mirroring.
quick to do, easy to do, no real cost.

If you want to use DR groups, then you are talking about CA/EVA, and you would need to build this as a Metrocluster, which involves purchasing the Metrocluster toolkit, the CA licenses etc etc. and then configuring the Metrocluster package(s).

Take a look at the following
http://docs.hp.com/en/B7660-90018/B7660-90018.pdf

[mc105]

thanks Melvyn

I'm slightly worried about using MirrorDisk/UX as jeff pointed out we could have problems with it.

I'll read up on MetroCluster

Cheers

[Klaas D. Eenkhoorn]

Hello,

Why not use Mirror UX ?
We have the allmost the same situation here not 2 MC/SG nodes but 4 with HPUX 11.11 and 2 EVA 3000.
For 4 years now we have had no problems with mirroring whats however.

We have 2 major SAP instances running this way without MC/SG tools.
Using Mirror UX both EVA3000's (on dfferent locations) are kept in sync. Failover, no problem !

Really keep it simpel, Use Mirror UX.
No hussel with failover and CA/EVA and turning replication directions of CA the other way.

Having problems with or losing one EVA, No failover will accure but Mirror UX stalls the writes for that eva until that eva is back on line.
With CA this will result in an failover to the other node with corrupt data on the mirrored eva as a result.
eg. Filesystems left open, files left open etc. etc.
CA mirrors the data as is.

Kl@@s

[mc105]

Cheers Klass -

I'll go ahead and setup with MirrorDisk for testing