2009-03-02 08:09 PM
Hi. I have a question about snapvault and exchange archiving. The following documentation explains how to use snapvault to archive snapmanager for exchange snapshots:
One part stood out in particular:
IMPORTANT: Keep in mind that SnapVault restores the entire qtree, so any other data that might be on the primary storage qtree will get overwritten with the data from the SnapVault secondary storage when performing a SnapVault restore.
It goes on to list an example of a typical Exchange architecture:
Database: Mailbox Store
Database: Public Folder Store
My question would be if I was snapvaulting each of these 3 qtrees and one day the Finance database in StorageGroup_3 - LUN4 became corrupted and I wanted to do a snapvault restore, would that mean that all of the databases contained within that qtree would be restored?? That would mean that LUNS 3, 5,6,7 would contain out-of-date data just because LUN4 needed to be restored. Not very acceptable.
Am I incorrect on that assumption? To be able to restore a database from archive it would seem that I'd have to architect my Exchange to contain a single database LUN for every StorageGroup. That way I'd have a qtree for each StorageGroup and could therefore perform a snapvault restore without worrying about it overwriting other LUNS.
How difficult would this be to accomplish for an already existing layout that looks like the third qtree (above) that contains several LUNS in the Storage Group? Is it even possible to create a qtree for a volume that contains LUN(S) already without destroying that data? For example
LUN8 - HR
LUN9 - Payroll
Would I be able to create a qtree within that volume that contains LUNS 8,9 without destroying any data or would the qtree have to be created first before any LUNS are created under it? I'm thinking the qtree would have had to have been created first as snapdrive needs the full path to the LUN. I guess that would mean a disconnect and reconnect of the snapdrive LUNS to get the new paths - which means downtime for Exchange.
It just seems that to be able to use snapvault to archive and restore Exchange backups means that the Exchange layout really needs to be thought out BEFORE actually implementing the layout. Doesn't seem like an easy thing to accomplish for an already existing Exchange layout since most people don't use qtrees and/or layout their Storage Groups to contain single database LUNS.
While a single databse per Storage Group is fine for Exchange 2007 (50 Storage Groups, 50 databases per SG), it's not neceassirly ok for Exchange 2003 (4 SG's with 5 databases per SG).
Maybe someone who is using snapvault to archive their Exchange backups can share their experiences?
2009-03-03 10:10 AM
You could simply use SnapVault to restore to a second location?
The reason for having the backups based on Storage Groups is that is the only way Exchange provides consistency points. Logs are generated per Storage Group and not per mailstore, so snapshotting each mailstore separately would be difficult to get solid consistency points.
But yes, you could easily architect it so that each mailstore was in it's own qtree within a single volume. But not easily from an existing infrastructure, it would be a migration process.
Hope that helps a little?