Data Backup and Recovery

snapshots and NFS datastores on Vmware


I have question on nfs datastores in vmware.   We have setup separate volumes and then export them as NFS datastores on our vmware environment,   But my question is that is it worth enabling snapshots on these volumes (nfs datastores).  Do most people do this, is it worth doing.  I am debating if we need snapshots.



Not really, because the data won't be consistent. Use VSC for snapshots instead.


In all honesty, it's a bit hairy subject

VSC consistency relies on VMware snapshots. These in turn are quite often temperamental, so in many cases VMware snapshots for VSC backups are turned off anyway. So the end result is "a" snapshot, as good, as any other taken on the filer side. Also, a crash-consistent backup of OS partition will work just fine in most cases - it is like pulling a plug suddenly from a server, running check disk may be required, but in principle it should work. It is a different story for databases and/or mail stores, where ideal approach is to use relevant SnapManager product. For Microsoft apps though VSC can potentially do data consistency as well  via VMware tools & VSS integration, but it can't do more advanced stuff like log trimming.

That said, using VSC offers some other benefits for OS backups, like scheduling, backup catalogue, etc.




We ran shapshots on our NFS datastores, and it saved our butts several times.  As others said, the data is only crash consistent, but 99% of the time we have been able to recover from that when needed.  We've been able to crack open a vmdk from a snapshot and extract a single file on Redhat.  I've seen notes of a similar process for windows, but I've never seen it.  Radek is right about databases - the one time the vmdk snapshots didn't work for us was trying to get into a mysql database on a restored guest.

I haven't played with any of the tools to manage the snaps and quiescing the guests, so I can't speak to those.

One thing - the size of the snapshots gets BIG.  Of course this depends on the data change rate, but in general they are BIG.  Something that we never got around to doing was removing the swap partition from the volume being snapped - if you can do that, your change rate should go way down.

Hope that helps...