September 1, 2005 at 11:38 am
We have an active/active cluster set up on one of our production servers. a strange thing happened which i am not sure if anyone has experienced before. Node 1 is active by default and it failed over to node 2 without anyone manually doing a failover. Only thing that happened was one of the filegroups on one database got filled up and it logged too many error msgs in sql log but is that what triggered the cluster failover by itself. Any thoughts or suggestions why this might have happened/ My cluster.log got overwritten so i couldnt trace back what the problem was .Any help will be greatly apprecaited.
TIA
September 1, 2005 at 12:36 pm
The Event Log on either (or both) Windows servers should have one or more entries that will describe what was going on to cause the failover. Check the Event Viewer.
September 1, 2005 at 2:54 pm
Had a similar event happen where there was something wrong with the heartbeat nic(s) and node2 would think node1 was dead so it would fail over.
September 2, 2005 at 11:59 am
" Only thing that happened was one of the filegroups on one database got filled up and it logged too many error msgs in sql log but is that what triggered the cluster failover by itself. "
Let me dissect this statement.
"Only thing that happened was one of the filegroups on one database got filled up" --> OK, a problem, not a showstopper that I am aware of (unless it's the master database !!!).
"it logged too many error msgs in sql log" --> the only way this could be an issue is if you had run out of disk space on that drive. I've seen huge, 3-4 Gb errorlogs and SQL still keeps tickin' ...
There ought to be messages in the system and application event logs indocating the exact nature of the failover as stated above. Also, you probably should parse out your SQL errorlog (removing all the filegroup space error messages and see what else you can find.
RegardsRudy KomacsarSenior Database Administrator"Ave Caesar! - Morituri te salutamus."
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply