none
Apply SP1 to SQL 2012 clustered on a Windows 2012 failover cluster

    Question

  • For earlier versions of clustering there was an ability to prevent a resource from failing to a particular node. I used this when service packing clusters to ensure that no SQL instances would attempt to fail back to the node I was patching. That ability doesn't seem to exist anymore. ANd pausing or stopping a node causes rule failure in the SP setup.

    Has anyone successfully performed this patching and are there any best practises or tips to follow?


    Regards, Matt Bowler MCITP, My Blog


    • Edited by matt.bowler Sunday, December 22, 2013 10:40 PM
    • Changed type matt.bowler Friday, December 27, 2013 1:07 AM Incorrectly categorised
    Sunday, December 22, 2013 10:39 PM

Answers

  • Hello Matt,

    This is still available, though I'm not sure where you remember it being. You can choose individual resources in a resource group and can also do it at the resource group level as well. It's in the properties of each level and allows each resource or resource group to be allowed to run on only the nodes selected.

    Here is the KB on it (though they use the command line): http://support.microsoft.com/kb/958734

    I'd patch the instances on the nodes that the instance is not currently active (if more than 1), pause the node and reboot if necessary. Continue rolling until services need to be moved or your patched nodes outweigh your unpatched nodes at which point I'd move the services and patch all remaining nodes.

    -Sean


    Sean Gallardy | Blog | Twitter

    • Marked as answer by matt.bowler Friday, December 27, 2013 1:07 AM
    Monday, December 23, 2013 12:18 AM
  • Hi Sean,

    Sadly that isn't available in the Windows Cluster 2012 GUI (or at least I can't find it). But your post did point me in the right direction and I was able to activate the cluster.exe command line interface and use this tool to achieve the rolling update.

    http://blogs.msdn.com/b/clustering/archive/2012/04/06/10291601.aspx

    http://technet.microsoft.com/en-us/library/cc723245.aspx

    The relevant cluster.exe commands are RemoveOwner and AddOwner.

    I'm still curious to know what the recommended best practise is and why the possible owner list is not able to be managed via the GUI.




    Regards, Matt Bowler MCITP, My Blog


    • Edited by matt.bowler Monday, December 23, 2013 8:54 AM
    • Marked as answer by matt.bowler Friday, December 27, 2013 1:07 AM
    Monday, December 23, 2013 2:06 AM

All replies

  • Hello Matt,

    This is still available, though I'm not sure where you remember it being. You can choose individual resources in a resource group and can also do it at the resource group level as well. It's in the properties of each level and allows each resource or resource group to be allowed to run on only the nodes selected.

    Here is the KB on it (though they use the command line): http://support.microsoft.com/kb/958734

    I'd patch the instances on the nodes that the instance is not currently active (if more than 1), pause the node and reboot if necessary. Continue rolling until services need to be moved or your patched nodes outweigh your unpatched nodes at which point I'd move the services and patch all remaining nodes.

    -Sean


    Sean Gallardy | Blog | Twitter

    • Marked as answer by matt.bowler Friday, December 27, 2013 1:07 AM
    Monday, December 23, 2013 12:18 AM
  • Thanks for the response Sean - but I'm using SQL Server 2012 on top of Windows Server 2012 - and cluster.exe has been removed. It has been replaced by powershell cmdlets. I've scoured lists of these but can't see a way to prevent a resource (or group) from failing onto a particular node during maintenance. It may be that this is no longer necessary on Windows 2012?

    My case is quite simple at this stage - I am building a POC rig for our websites and want to run a few scenarios - patching being one. We have a 2 node windows cluster with 2 named SQL instances. These would normally run in 'active - active configuration' but the plan for patching is to fail both to one node, remove the passive node from possible failover partner list, patch passive node... rinse and repeat....


    Regards, Matt Bowler MCITP, My Blog


    • Edited by matt.bowler Monday, December 23, 2013 1:06 AM
    Monday, December 23, 2013 1:04 AM
  • Matt,

    The KB was old but it gave the right idea which is still applicable.

    If you open up failover cluster manager (for the visual interface instead of using the cmdlets) and find your resource groups, you can set this in the properties of the resources:


    Sean Gallardy | Blog | Twitter

    Monday, December 23, 2013 1:43 AM
  • Hi Sean,

    Sadly that isn't available in the Windows Cluster 2012 GUI (or at least I can't find it). But your post did point me in the right direction and I was able to activate the cluster.exe command line interface and use this tool to achieve the rolling update.

    http://blogs.msdn.com/b/clustering/archive/2012/04/06/10291601.aspx

    http://technet.microsoft.com/en-us/library/cc723245.aspx

    The relevant cluster.exe commands are RemoveOwner and AddOwner.

    I'm still curious to know what the recommended best practise is and why the possible owner list is not able to be managed via the GUI.




    Regards, Matt Bowler MCITP, My Blog


    • Edited by matt.bowler Monday, December 23, 2013 8:54 AM
    • Marked as answer by matt.bowler Friday, December 27, 2013 1:07 AM
    Monday, December 23, 2013 2:06 AM