SQL Server Performance

Failover behavior on a two node active-passive for a network failed situation

Discussion in 'SQL Server 2008 Questions [Archive Only - Closed]' started by WingSzeto, Oct 28, 2008.

  1. WingSzeto Member

    It is my first time setting up clustering and I do this with SQL 2008 standard 64 bit on a Windows 2008 64 bit enterprise. Node 1 has one NIC go to a public network switch 1 and another NIC go to a private network switch 2 (for heartbeat). Node 2 has one NIC go to the same public network switch 1 and another NIC go to switch 2. We don't have any network teaming or network load balance setup. Again it is a active-passive cluster setup.
    When I unplugged the network cable from Node 1 to swtich 1, the cluster failed over to node 2 successfully. When I plugged the Node 1's network cable back to switch 1 and unplugged the network cable from Node 2 to switch 1, the cluster doesn't fail over from Node 2 to Node 1. Under active-passive clustering, should I expect the cluster fail over from node 2 to node 1? If it should, what's wrong? If it shouldn't, does it require some steps to make it fail over back like reboot or something?
    wingman
  2. Saurabh Srivastava New Member

    As you are using WIndows 2008/SQl2008 combination I would suggest you to use cluster validation tool to find out any configuration issues. Microsoft has made great improvements in this tool to make cluster troubleshooting easier for us. Please post what you findings and solution to help others.
  3. WingSzeto Member

    Yes, I did the configuration validation and it passed all tests.
    Do you know what the expected failover behavior for the failover test I mentioned in my original email? SQL 2008 may have some improvements on clustering but I think this type of networking test should be similar to the SQL 2005 version. Do you have any white paper and blog I can take a look?
  4. Saurabh Srivastava New Member

  5. WingSzeto Member

    In the Windows 2008 link, are you able to open the 'WS2008 Failover Clustering Datasheet.doc' document? It said it is a Word document but it is all garbage characters in there.
    In any case, these documents didn't answer my questions. I think my scenerio should also apply to SQL 2005 clustering. I like to hear from that SQL 2005 group and see what they will expect from SQL 2005 clustering if the network failure test conducts in SQL 2005 clustering setup.
    wingman
  6. Saurabh Srivastava New Member

    Yes, 2008 link is throwing garbage. Earlier it opened fine now it seems to be some problem with link.
  7. atwr2 New Member

    We are also experiencing this issue. Have you come up with a fix?
    We have a 2 node Windows Server 2008 x64 Enterprise Edition Cluster in Active/Passive. We can fail over fine from Node1 to Node2 when the Public network fails. However when we bring the other node back up (by replacing the network cables) the node won't fail back to it if then Node2 fails with a network failure.
    It just doesn't seem to even try, it just marks the resource as failed and leaves it at that and to recover we basically have to wait until the servers Network recovers and then manually bring the cluster back up.
    If anyone is able to shed some light on this it would be greatly appreciated!

Share This Page