We have a 2 node Microsoft WIndows 2008 Cluster running SQL 2008.
Installed onto both of these Clusters is Snapdrive 6.4 and SMSQL 5.2
Recently we have been asked to restore a single database back from a snapshot, which we did, however it appears to have restored all of the lun, (including all databases !)
even though the snapmanager restore report and log on the filer stipulates that I did only select 1 database (the one I required) and the others were deselected.
Anyone else seen this issue ?
This is expected behaviour I am afraid - SMSQL restores the entire LUN, so if you have multiple databases sharing the LUN, they all will be restored.
For granular backup / recovery perspective each user database should be located in its own dedicated LUN and FlexVol.
Have a look at this TR for more detail:
Thanks for the reply, I'm not 100% sure that is correct, as we have the same setup on a nother 2 node cluster (and the only difference is the Snapdrive version is lower)
and we have performed single database restores before on that cluster ??? And that SQL Server like the one we are having issues with all the database files are on the D: drive
and logs are on the L drive....
I have seen this documentation before, but not seen the bit regarding restoring the whole LUN, but will check
Okay, I stand corrected - single DB per LUN/volume is the recommended best practice, but SMSQL can handle more 'difficult' cases too.
Under 'normal' circumstances SMSQL should use LUN cloning for restore operation (so all databases would be restored). However, it should switch to a copy-based restore method if only a subset of databases on the same LUN is selected.
Probably in your case LUN cloning restore method is used, instead of copy-based.
No worries I thought I was going crazy when you wrote that bit about the restore of the luns - it appears in our instance to be the case that even though one was selected it has restored the Lun, I still think it points at Snapdrive as this is the software link into the restore LUN procedure, where SMSQL will work at the application layer - I will let you know our findings. Thanks for the reply.
Its interesting as have created a Windows 2008 Server with SQL 2008 on it and when select 1 database to restore it restores only 1,
I have now created a 2 node cluster in a test system, and I still get the same issues, when restoring 1 database it restores the whole lun and also brings back files that I deleted, I can replicate this all day - I have a call with NetApp Support and they cant replicate it, have patched to latest version of SD and SMSQL but it still happens, just trying to work out whats different to a standalone cluster and a clustered scenario ? NetApp seem to think it was the location of the SQL databases but I dont wash with that as SQL dont really care...