What's best practice backing up DR Groups in EVA BC/CA environment?

[Brad Smithberger]

ENVIRONMENT:
We have (2) EVA8000 running BC/CA. Both have EVA Command View Suite v6.0.2 for XCS v6.0.

Hosts are HPUX servers running 11i v1 running MC/Serviceguard v11.16 with Metrocluster for EVA v1.0. The hosts are configured using MC/Serviceguard/Metro Cluster disk configuration uses DR Groups.

ISSUE:

Though Mirrorclones are obviously best choice, unfortunately I can't use Mirrorclones because they are not allowed with DR Groups. Don't know why either. I can only use containers/snapclones/snapshots.

BTW....Just in case HP is listening....HP NEEDS TO SUPPORT MIRRORCLONES IN DR GROUPS! THIS IS A MUST HAVE!

QUESTION:
What's best practice to backup my DR Groups using CONTAINERS/SNAPCLONES/SNAPSHOTS etc? I intend to automate the process using RSM or SSSU. Any help is most appreciated!

[Ivan Ferreira]

>>> Though Mirrorclones are obviously best choice

I'm not sure about this, if you only will use for backup, demand allocated snapshots seems to fully fit your needs.

>>> What's best practice to backup my DR Groups using CONTAINERS/SNAPCLONES/SNAPSHOTS etc?

I think that there is not too much change if where out of DR Groups.

First, you should make data consistent to take the snapshot, that is, put database in backup mode or shutdown the db, freeze applications, etc.

You can take a snapshot of the source or destination vdisk if you use sync replication. If you use async replication, your snapshots should be done at the source storage.

After snapshot complete, resume I/O to your LUNs.

Probably, your best option is to use RSM because it can handle errors and send notifications. This is hard to do with SSSU.

[Brad Smithberger]

Ivan,

Mirrorclones seem obvious to me becuase my luns are big, and would come in handy when snapping them each day because they only update delta changes made since last snap ( typically 24 hours).

What I've read so far, and sounds like your recommeding it, is to suspend my DR Group links while taking snapshot. That's doesn't sound appealing since I have and HA enviornment, and can't failover with consistent data if my links are suspended. Agree?

I agree with RSM recommendations. Tending lean towards that solution.

Thanks for your input.
Brad