none
Hyper-V cluster rolling upgrade with CSVs

    Question

  • We have a 4 node 2012 R2 Hyper-V Cluster using CSVs for storage that I want to upgrade to Server 2016.  I know when upgrading from 2012 to 2012 R2 I had to build a new cluster and could not introduce new cluster nodes and live migrate VMs to the new nodes because I was using CSVs.  I looked through the documentation here https://technet.microsoft.com/windows-server-docs/failover-clustering/Cluster-Operating-System-Rolling-Upgrade about rolling upgrades but did not see anything about any issues with CSVs.  Can I do the rolling upgrade using CSVs without any issues then?
    Thursday, March 02, 2017 9:57 PM

Answers

  • "Can I do the rolling upgrade using CSVs without any issues then?"

    Though the engineers have done a great job on this new capability, there is no way anyone can guarantee there will be no issues.  Many have performed this upgrade successfully.  Some have had issues.  To be safe, backup your environment before starting.  That way, whether you perform an in-place upgrade or a new cluster, you have your environment backed up to be used to recover.


    . : | : . : | : . tim

    • Proposed as answer by Leo HanModerator Thursday, March 16, 2017 7:12 AM
    • Marked as answer by WSUAL2 Friday, March 17, 2017 4:21 PM
    Friday, March 03, 2017 2:34 PM
  • Hi,

    drain/evict a node and install it fresh with Windows Server 2016, if you don´t lift up the Cluster Functional Level and  issues happens, you could roll back. Drive the Cluster in mixed mode for a certain period of time, migrate VMs and proof the state of Health.

    Regards,
    Marcel

    https://www.windowspro.de/marcel-kueppers

    I write here only in private interest

    Disclaimer: This posting is provided AS IS with no warranties or guarantees, and confers no rights.

    • Marked as answer by WSUAL2 Friday, March 17, 2017 4:19 PM
    Friday, March 03, 2017 2:54 PM

All replies

  • Hi WSUAL2,

    >>Can I do the rolling upgrade using CSVs without any issues then?

    I have seen some posts about the issues caused by rolling upgrade. And the issues are difficult to troubleshoot.

    Before more related official documents become available, I suggest you build a new cluster with 2016 Server and migrate the roles and data.

    Best Regards,

    Leo


    Please remember to mark the replies as answers if they help.
    If you have feedback for TechNet Subscriber Support, contact tnmff@microsoft.com.

    • Proposed as answer by Leo HanModerator Thursday, March 16, 2017 7:12 AM
    • Unproposed as answer by WSUAL2 Friday, March 17, 2017 4:21 PM
    Friday, March 03, 2017 5:04 AM
    Moderator
  • "Can I do the rolling upgrade using CSVs without any issues then?"

    Though the engineers have done a great job on this new capability, there is no way anyone can guarantee there will be no issues.  Many have performed this upgrade successfully.  Some have had issues.  To be safe, backup your environment before starting.  That way, whether you perform an in-place upgrade or a new cluster, you have your environment backed up to be used to recover.


    . : | : . : | : . tim

    • Proposed as answer by Leo HanModerator Thursday, March 16, 2017 7:12 AM
    • Marked as answer by WSUAL2 Friday, March 17, 2017 4:21 PM
    Friday, March 03, 2017 2:34 PM
  • Hi,

    drain/evict a node and install it fresh with Windows Server 2016, if you don´t lift up the Cluster Functional Level and  issues happens, you could roll back. Drive the Cluster in mixed mode for a certain period of time, migrate VMs and proof the state of Health.

    Regards,
    Marcel

    https://www.windowspro.de/marcel-kueppers

    I write here only in private interest

    Disclaimer: This posting is provided AS IS with no warranties or guarantees, and confers no rights.

    • Marked as answer by WSUAL2 Friday, March 17, 2017 4:19 PM
    Friday, March 03, 2017 2:54 PM
  • My plan was to drain/evict a node and install it fresh with 2016.  

    I just learned that I need to change the network that the cluster resource itself and the network of the OS for the cluster nodes are on so I am trying to factor that into this equation as well.  Is there a process for changing the network and IPs that the cluster and its nodes are on without any downtime or VMs jumping from node to node as the host OS IPs are changed?  The cluster management and live migration networks and IPs will be staying the same.  

    Friday, March 03, 2017 3:50 PM
  • Host IP addresses are completely separate from the guest IP addresses.  No different than two separate physical machines.  Changing host IP addresses should have no effect whatsoever on the VMs.  I suggest making the IP changes on the existing cluster to make sure you have things working properly before performing the rolling upgrade.  Better to keep control of the changes and ensure one set of changes is working properly before moving to the next.

    . : | : . : | : . tim

    Saturday, March 04, 2017 1:27 PM
  • That sounds good. I was hoping that would be the case. So in what order should I make the IP changes? Cluster IP address first and then the hyper-v nodes? Or the nodes first and then the cluster IP address? I want to make sure that I do it in the right order so I don't crash the cluster and VMs. Thanks.
    Saturday, March 04, 2017 4:14 PM
  • You don't need to change the Cluster CNO and IP Address. The process is rolling, node per node. Evict and add the node to the existing Cluster. Configure Node/Host IP Addresses like the old Node before including the Hyper-V Switch. Use the same name as before and think about case sensitive. Before that, live migrate the VM's. You don't need to change the VM configuration.

    https://www.windowspro.de/marcel-kueppers

    I write here only in private interest

    Disclaimer: This posting is provided AS IS with no warranties or guarantees, and confers no rights.

    Saturday, March 04, 2017 4:54 PM
  • I know I don't need to change the cluster IP and node IPs for the the rolling upgrade. I need to change the network that they are currently on to a more secure one first and then do the rolling upgrade. So in what order should I make the IP changes? Cluster IP address first and then the hyper-v nodes? Or the nodes first and then the cluster IP address? I want to make sure that I do it in the right order so I don't crash the cluster and VMs. I need to do it during working hours so I can have the network guys flip ports on the nodes to the new vlan, hence why I want to make sure I have the exact process down before I do it. Thanks.
    Saturday, March 04, 2017 6:31 PM

  • Hi,
    Are there any updates?
    You could mark the reply as answer if it is helpful.
    Best Regards,
    Leo

    Please remember to mark the replies as answers if they help.
    If you have feedback for TechNet Subscriber Support, contact tnmff@microsoft.com.

    Thursday, March 16, 2017 7:12 AM
    Moderator
  • I was able to change the IPs of the Cluster nodes without any issues.  Now I just need to wait for my storage vendor to certify that it supports Windows Server 2016 and then I will start by draining the roles on one Hyper-V node, evict it, re-build it with a fresh copy of 2016 and then so on and bring up the cluster functionality level to 2016 once I verify everything is working.  All VMs will be backed up before attempting this.
    Friday, March 17, 2017 4:21 PM