August 30, 2007 at 11:21 am
All,
I just setup Database mirroring utilizing a witness for automatic failover. The weird thing is that I finished the setup at 4am and everything was synch'd up fine. Sometime between 4am and 8am the pricipal role changed on the pricipal server to the mirror server without any outage, hardware failure, or manual failover. Has anybody seen this before?
Also, since this is happening without an outage, is there a script or procedure that will change the pricipal role back to the original pricipal server if in fact the server never had a problem after it's finished rebooting?
Very strange stuff, but I'm hoping somebody can shed some light on this.
-Dave
August 30, 2007 at 11:42 am
Well, one way or another the witness thought there was an outage and initiated failover.
One thing to think about is that mirroring is VERY network dependent. You may want to try playing around with the "ALTER DATABASE ... SET PARTNER_TIMEOUT x" statement if you can't find any other reason for the failover. Network issues can cause failover.
August 30, 2007 at 11:49 am
Aaron,
Thanks for your input. It's greatly appreciated. I'll take a look at what you suggested.
What type of network issues could possibly cause this? Any ideas?
What about a role change script to switch back to the original pricipal server if a reboot occurs for something minor like a windows update?
-Dave
August 30, 2007 at 12:00 pm
Well, I'm of the opinion that you shouldn't fail-back until you've confirmed that everything is fine on the server; so I'm not a huge fan of automatic fail-back. You should be able to just open up the database properties and under the mirroring section click the Failover button.
BOL tells me that the default is 10 seconds of network wait will cause a failover by default. You can change that using the method I described above. At my last job we had some unscheduled emergency network maintenance which disconnected a whole bunch of servers and when it came back online some had failed over.
Are you sure there's nothing in the logs to indicate some other cause of failover? I haven't used mirroring in about a year, but I would think that the witness would log some information about why it initiated the failover.
August 30, 2007 at 12:14 pm
I agree with you statement about not being a huge fan of automatic fail-back until you've confirmed that everythign is fine on the server.
I'll take a look at the ALTER DATABASE syntax you sent me and also take a look at the logs in more detail to see if there was some other cause of failover.
I appreciate all your input on this topic. You are the man!
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply