none
Geocluster SQL 2008: 2 nodes + 1 Witness (3 datacenters)

    Question

  • Cluster: Windows 2008 R2 + SQL2008 R2.
    3 Datacenters. 1 datacenter (Site3) is used to Witness. Replication exists beetween other datacenters (Site1, Site2).
    Site 1 has 2 SQL Server nodes. Site 1 has 2 SQL Server nodes. All instances are actives on a Site (Site1 or Site2)

    What would happen in these situations? (All SQL Instances are actives on Site1)

    1) Site1 loses communication with Site3
    2) Site1 loses communication with Site3 and Site 2
    3) Site3 loses communication with Site2
    4) Site1 fail and Site 2 loses communication with Site3
    5) In this case of have a cluster with only 2 votes (for example nodes on site1), How could I start the clúster? Situation2)

    Sunday, October 30, 2011 6:00 PM

Answers

  • 1) Depends which node has ownership of the FSW resoruce in the cluster. Assuming Site 1 owns the FSW resource, the FSW resource would fail and move to Site 2. If Site 2 owns the FSW resource, nothing should happen.

    2) Site 1 will attempt to arbitrate for quorum and will fail. Cluster service will terminate at Site 1 and resources will be moved to Site 2

    3) Same as #1 though obviously the sites are reversed. 

    4) Cluster shuts down on all nodes until the cluster is able to form with 2 or more votes or a 'forcequorum' occurs. If you meant both 1 & 2 lose access to 3, the cluster would remain fully functional, except the FSW resource would fail.

    5) start the cluster with the 'forcequorum' flag. See http://support.microsoft.com/kb/947713 for more details.

    Hope this helps.


    Visit my blog about multi-site clustering - http://msmvps.com/blogs/jtoner
    Monday, October 31, 2011 3:02 AM
  • 6) Yes, the "Current Host Server" is the current owner of the cluster core resources. You can also see this in command line using a "cluster group" or "cluster res" command

    7.1) The cluster will shut down on all nodes

    7.2) Does this connectivitiy issue still exist? If so, cluster service will continue to fail until either the Forcequorum occurs or the connectivity is restored.

    8.1) Cluster service will fail on Site2. All resources from Site2 will failover to Site1.

    8.2) Does this connectivitiy issue still exist? If so, it depends on which nodes you start first and/or which node successfully arbitrates for the FSW. Whichever node gets the FSW, will bring cluster and all resources online. The other node will continue to try, but will fail until connectivity is corrected.

    9) Not sure what you mean...DAC is built using the same technology as clustering and is pretty much the same concept. Quorum is a majority of votes and if you don't have a majority of votes, your cluster goes down.

     


    Visit my blog about multi-site clustering - http://msmvps.com/blogs/jtoner
    Tuesday, November 08, 2011 6:55 PM
  • Correct, SQL should not be affected by communications failure with Site3. The only group that would potentially be affected would be the group that contains the FSW resource...which should be your "Cluster Group".
    Visit my blog about multi-site clustering - http://msmvps.com/blogs/jtoner
    Wednesday, November 09, 2011 2:15 PM

All replies

  • You asked this question on the SQL forums as well. Try to only post in one place.
    Allan Hirt Blog: http://www.sqlha.com/blog Author: Pro SQL Server 2008 Failover Clustering (out now!)
    Sunday, October 30, 2011 9:52 PM
  • 1) Depends which node has ownership of the FSW resoruce in the cluster. Assuming Site 1 owns the FSW resource, the FSW resource would fail and move to Site 2. If Site 2 owns the FSW resource, nothing should happen.

    2) Site 1 will attempt to arbitrate for quorum and will fail. Cluster service will terminate at Site 1 and resources will be moved to Site 2

    3) Same as #1 though obviously the sites are reversed. 

    4) Cluster shuts down on all nodes until the cluster is able to form with 2 or more votes or a 'forcequorum' occurs. If you meant both 1 & 2 lose access to 3, the cluster would remain fully functional, except the FSW resource would fail.

    5) start the cluster with the 'forcequorum' flag. See http://support.microsoft.com/kb/947713 for more details.

    Hope this helps.


    Visit my blog about multi-site clustering - http://msmvps.com/blogs/jtoner
    Monday, October 31, 2011 3:02 AM
  • Ok Thanks

    6) Is "Current Host Server" the owner of the FSW Resource?

    7) A node Site1 has ownership of the FSW resoruce in the cluster, and loses connectivity with Site3 and Site2. Also, Site2 loses connectivity with Site3.

        7.1) I suppose that cluster will stop....

        7.2) I shutdown all cluster nodes on Site1 and Site2. After, I start all cluster nodes... What happend??

    8) A node Site1 has ownership of the FSW resoruce in the cluster, and loses connectivity with Site2. Also, Site2 loses connectivity with Site1.

       8.1) I suppose that cluster will not stop....

       8.2)  I shutdown all cluster nodes on Site1 and Site2. After, I start all cluster nodes... What happend??

     

    9) What about Datacenter Activation Coordination (DAC) ?? (Exchange 2010) Does Cluster 2008 have anything like this??

    Tuesday, November 08, 2011 6:07 PM
  • 6) Yes, the "Current Host Server" is the current owner of the cluster core resources. You can also see this in command line using a "cluster group" or "cluster res" command

    7.1) The cluster will shut down on all nodes

    7.2) Does this connectivitiy issue still exist? If so, cluster service will continue to fail until either the Forcequorum occurs or the connectivity is restored.

    8.1) Cluster service will fail on Site2. All resources from Site2 will failover to Site1.

    8.2) Does this connectivitiy issue still exist? If so, it depends on which nodes you start first and/or which node successfully arbitrates for the FSW. Whichever node gets the FSW, will bring cluster and all resources online. The other node will continue to try, but will fail until connectivity is corrected.

    9) Not sure what you mean...DAC is built using the same technology as clustering and is pretty much the same concept. Quorum is a majority of votes and if you don't have a majority of votes, your cluster goes down.

     


    Visit my blog about multi-site clustering - http://msmvps.com/blogs/jtoner
    Tuesday, November 08, 2011 6:55 PM
  • John, In this case of 1) Site1 loses communication with Site3

    1) Depends which node has ownership of the FSW resoruce in the cluster. Assuming Site 1 owns the FSW resource, the FSW resource would fail and move to Site 2. If Site 2 owns the FSW resource, nothing should happen.

    If Site 1 owns the FSW.... I suppose that SQL Instances on Site1 don't move to Site2, only the ownership of the FSW, not??

    Wednesday, November 09, 2011 9:54 AM
  • Correct, SQL should not be affected by communications failure with Site3. The only group that would potentially be affected would be the group that contains the FSW resource...which should be your "Cluster Group".
    Visit my blog about multi-site clustering - http://msmvps.com/blogs/jtoner
    Wednesday, November 09, 2011 2:15 PM