SQL DBs are suspect after test failover

[paulymo]

I have been testing DR with SRM for one of our clients and the SQL databases come up as suspect during the tests.  Here is a typical SQL server build for this customer:

OS = Windows 2008

SQL = SQL Server 2005

OS/SQL Data = NFS datastore

DB/Logs = Windows iSCSI LUNs (vcenter/SRM is unaware of these LUNs)

vCenter = 4.1

SRM = 4.1

Storage = NetApp FAS2040

I've tried failing over with SRM and manually creating FlexClones of the volumes and attaching the VMs to vcenter.  Both methods produce the same result.  The one odd thing is any DB that is on the same partition as the OS (C drive) will mount without issue. 

Any thoughts?

[dcoz]

Are you replicating the LUNs that contain the Databases and Logs?

[paulymo]

Yes, the LUNs are being replicated with SnapMirror from Primary to Recovery site. 

All of the databases are set to Full Recovery and are being backed up with Avamar from EMC. 

[admin]

The fact that the DBs that are on C: mount is a good hint.  When replicating databases like SQL or Exchange we have to make sure that the volumes hosting the DBs and Logs are replicating as part of a set to ensure we maintain write consistency.  If not then there will be problems getting them to mount with out playing with a bunch of recovery commands.  For the DBs on C: are the logs also on C:?  If so then I'm betting that the DBs and Logs are replicating and different times.

-alex

[paulymo]

I figured out the Read Only LUNs were mounting instead of the FlexClone copy.  I removed the RO LUNs from the iGROUPs and add them back manually during a test.