I found an older question in this forum covering the same subject, but it was never answered, figured I'd bring it up again...
Our company's setup has 8 SQL instances running in an active-active Windows cluster with 3 nodes, the normal configuration has 1 instance on the first node, 3 instances on the second node, and 4 instances on the third node (based on the size and usage of the databases in each instance). Once I discovered that only one SnapManager operation could be active per machine, I setup a staggered backup schedule for the agent jobs on all the instances. However, there are occasions where one backup job may take longer than normal to complete and this causes another backup job on a different instance to fail because there's already an active SnapManager operation. Not to mention that if we ever have an instance not running on its preferred node, there will be even more overlaps in the backup job schedules. Trying to schedule all the jobs on all instances so that there's no overlap is not an option, the frequency of backups per instance would not be high enough, not to mention the occassional long runnning job could still cause an overlap. Each instance also has one backup job a night that also verifies the backup in a non-clustered standalone instance, I believe those jobs need to avoid overlaps as well. And, I'm not sure if this matters, but all the nodes are running Windows Server 2008 R2 Ent and all the SQL instances are SQL 2008 R2 Ent.
The way I have the backup jobs setup right now is working, but is very fragile. I'm sure there are other folks out there that have a similar setup, I'm curious as to how others have approached these issues, hopefully there's a better solution than what I've been trying.
