I have SQL Server 2008 R2 Enterprise 64bit on Windows 2008 R2 Enterprise 64bit. Each SQL Server has multiple instances, 6 cores, and 32GB RAM on a 10Gb network. Everything is VM btw.
So what I have are multiple databases in each SQL instance, with multiple instances, mirrored in high-safety witnessed autofailover config. There are two primaries and one mirror and each primary serves as the other's witness in this config.
What happens is that randomly the witness thinks the database at the primary is not responding and, therefore, not available and causes a failover to the mirror. But it only does it for some databases in the instance and only some instances on the affected primary. So I end up with a hodgepodge of some DBs failed over and others that are not. It has become maddening and I can't figure it out except that it may be there is too much traffic and I would need to separate the witness function out to it's own server (instead of having each primary witness the other).
I need some thoughts on this as to why it may be happening. Especially since I can force the databases to fail back and all returns to normal without even a blink. Any help or suggestions would be VERY much appreciated!