We're running a Mirror setup (SQL Server 2005 SP3) with high availability mode, so a mirror, principal and a witness. We run maintenance at night which reorganizes the indexes. This causes a lot of changes in our log files which need to be synchronized with our mirror.
Now we have alerts setup to mail us when the mirror states change, we're filtering synchronizing/synchronized so we're left with the important connection lost / fail over messages. Every now and then during the maintenance however we get mailed that the principal has lost connection with the mirror. The mirror sometimes also gives a message that it loses the principal (not always though!)
Now this is leaving me rather clueless as to what is happening, I have increased the PARTNER TIMEOUT to 120 seconds (from the 10 which is default) Which I'd expect to be more than enough? Now I'm looking at the mirroring monitor at the moment and I'm noticing that around the time I'm getting the alerts the transfer of actual data also seems to grind to a halt. After an X amount of time (usually around a minute, sometimes more though) the transfer rate goes up again, sometimes it does this multiple times per database.
Now this leads me to believe there is something wrong with our network, something is unstable so to speak, how ever I'm not an DBA or experience Network engineer (I'm a software engineer who makes due with what little experience I have :x) so any suggestions/pointers in the right direction would be highly appreciated.
If this is normal behavior and if I should just disable the alerts during the maintenance that's something we could do as well (I actually intended to do just that until I noticed data transfer also stops) but I'd like to be sure that this is the case and that I'm not ignoring an actual problem which could escalate in the future.
Edit:
Found the following error: Error: 1479, Severity: 16, State: 2.
That lead me to the following KB http://support.microsoft.com/default.aspx?scid=kb;en-us;959006&sd=rss&spid=2855 but we're already running SP3 (which is where this problem is apparently corrected >.<)
