Raymond0 (8/31/2011)
We also use SnapManager for backups and restores and clones. And Document Server volume snaps.We now find that when the databases are frozen it may take several minutes for the snapshot to complete. We do have quite a few databases on the LUN. In fact about 90 databases. (Before any guffawing we have no choice but to use SnapManager and we inherited the applications as is.).
I am tending towards what has been written - ie 25 to 35 db per LUN. Now we also have some big indexes and I was tending towards putting the indexes onto their own LUN.
Any thoughts
Raymond0,
The limitations isn't based off a solid number. The limitation is based off a rough estimate of how long SMSQL will need to quiesce the system to complete the snapshot. The length is normally based off the total size of the volume. I'm actually surprised that if the volume is frozen for that long that SQL Server doesn't lose connection to your databases. I had always assumed anything greater than about 30 seconds would cause the databases to go suspect.
My recommendation would be to allocate additional LUNs to separate volumes and move even distribute the datafiles across the volumes. Snapshots are completed at the volume level so if you can arrange to maybe spread your database files across 3 or 4 volumes then you'll limit the total amount of time the system will freeze. I would also think about putting the log files on a separate volume than your datafiles.
In summary, if you have a bunch of small databases then put them on a single volume. If you have 1 or 2 very large databases then I would definitely recommend putting those on their own separate volumes. Keep in mind that since you have multiple databases on a single volume your restores will be streaming and could actually take as long or longer than a native restore.