We have a lot of NFS datastores, and trying to reduce the number of them. Particularly, for some datastore containing Oracle/SQL databases, we wanted to split them by putting DB in one datastore, and combine multiple OS portions from different VM's into the other datastores, thus, when we take snapshots for the purpose of backing them up and restoring, we can only take snapshots on datastores containing OS's, and without the need to take the snapshot on the datastore containing DB, because DB will be backed up by different DB backup tools. It also means that one VM could be spanned to be more than one datastores, as the result.
Are there any issues or concerns can you thin of with this approach?
I think its a good idea to have database files running on a different datastore may be with better performance and better management of backup.
But, since the entire state of the VM will be saved while taking a snapshot (which includes all the vmdks that it uses irrespective of the datastore), you might not have any advantage wrt VM snapshots.
Hope this helps.
Thanks,
Sandeep
Sorry, what advantage of VM snapshot am I losing? Please advise.
The only thing I know of is that I need to exclude these DB mounts when I restore a VM from Datastore Snapshot, because these DB related vmdk's might not be in the same datastore.
If I take the snapshot on individual VM, then the snapshot should be still useful without DB portion.
Just to reiterate my questions here, or there will be no issues for this type of merging?
The only concern with this approach, and it's something I've seen several times, is now you've purposefully created multiple failure domains where, previously, only one existed per VM. So in your use case that primarily addresses DB VMs, let's say you have four different VMDKs with each one performing a certain function and those, in turn, distributed to different NFS mounts. If any one of those four mounts is interrupted, you will not only lose the capabilities of whatever disk it was holding but it's very possible the entire VM could stop responding (this is not at all uncommon). You're adding complexity here for the sake of data protection, so ultimately you have to contend with and reconcile if that complexity and risk are worth the gains you would make. If you don't have a backup application that uses VADP, you may want to investigate that which gives you the flexibility to exclude certain drives from processing (for example, your DB drive(s)).
Good points, Thanks!
One more followup. Can you foresee any issues when we perform restoration, if we split OS portion with the rest of VMDK's for DB/Apps?
That'll probably depend on your backup application and if it's capable of a split restore like that. Many are not.