September 30, 2003 at 7:22 am
I am manually causing a failover in a testing environment by pulling a NIC cable on Node 1. At times the failover takes forever and the only thing to force the failover it is to stop and start the cluster service.
any suggestions?
October 3, 2003 at 8:00 am
This was removed by the editor as SPAM
October 3, 2003 at 10:11 am
No suggestions, since I don't know your setup and how you clustered everything.
But something to think about...when you 'pull the cable' you may not be doing something that requires a failover. The cluster is still operational, it just isn't connected to the network. Nothing went wrong with the cluster itself. Usually a failover is caused by a hardware problem (for example: one node failing, OS failure on one node, power loss on one node, etc).
-SQLBill
October 3, 2003 at 11:02 am
As SQLBill pointed out, the cluster is probably not even initiating a failover. If you have two NICs in each node, one for just cluster heartbeat traffic and one for network access, pulling access to the network on one of the nodes won't necessarily cause a failover. It depends on a lot of things, like what resources are owned by what node, resource dependancies, etc.
The best test of failover (although not necessarily the best for the hardware) is to just power off one node completely. I always include this when testing a clustering setup. Actually, it's a good test of the hardware to make sure the second node comes back up. If it doesn't, it's an opportunity to practice your disaster recovery plan 🙂
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply