Failover Clustering and Network Load Balancing Team Blog
Hi Cluster Fans,
Windows Server 2008 Service Pack 2 (SP2) is now available for download on the web! This blog post will walk you through performing a “rolling upgrade” from Windows Server 2008 RTM to Windows Server 2008 SP2. A rolling upgrade allows you to progressively upgrade your cluster nodes while keeping your resources highly-available, however you may have some downtime while moving the resource groups between nodes. What happened to Service Pack 1? Windows Server 2008 RTM was actually SP1 as it was released around the same time as Vista SP1.
Note that when moving between Windows Server 2003, Windows Server 2008 and Windows Server 2008 R2 a migration is needed because mixed-mode clusters are not supported. A migration guide is at: http://technet.microsoft.com/en-us/library/cc754481.aspx.
In this guide I use a 2-node cluster so when I split the cluster into 2 logical groups, each group will have 1 node. If you have larger clusters, you can walk through this guide in as many iterative steps as you need to gradually upgrade all of you nodes to SP2 while ensuring that the remaining nodes can sustain the running workloads of the services and applications in your cluster and you maintain quorum.
1) Download Windows Server 2008 SP2 for your cluster nodes:
· x86: http://www.microsoft.com/downloads/details.aspx?FamilyID=a4dd31d5-f907-4406-9012-a5c3199ea2b3
· x64: http://www.microsoft.com/downloads/details.aspx?FamilyID=656c9d4a-55ec-4972-a0d7-b1a6fedf51a7
· ia64: http://www.microsoft.com/downloads/details.aspx?FamilyID=e890b3cf-972b-483f-a2ff-03f6aefac6f8
· ISO: http://www.microsoft.com/downloads/details.aspx?FamilyID=9f073285-b6ef-4297-85ce-f4463d06d6cb
2) Move all resources to the first node or subset of nodes (in my 2-node cluster I will move them to node 1). A best practice is to also pause the node(s) you are about to upgrade by right-clicking the node name and selecting 'Pause'. This will prevent resources from failing over to them (this has been skipped in this blog post).
3) Upgrade the other node or subset of nodes (those which are not hosting any HA resources)
4) Move resources to the second node or group of nodes (if it was paused, resume it first). It is best to be in this 'mixed version' state for as little time as possible. You may want to control the failover of your Core Cluster Group or Witness Disk which can be done using PowerShell or Cluster.exe, however this is not required as these resources will automatically failover during the upgrade.
5) Upgrade the first node or group of nodes (those which are not hosting any HA resources). You may want to pause the node(s) first.
6) Redistribute your resources across your nodes. Resume the node(s) first, if necessary.
All of your nodes are now running Windows Server 2008 SP2 and you have finished you rolling upgrade. A KB article for adding service packs to any version of clustering is available at: http://support.microsoft.com/default.aspx/kb/174799.
Thanks,Symon PerrimanProgram ManagerClustering & High-AvailabilityMicrosoft
Is it safe to say you're upgrading the node without any active resources on it at that time?
Correct, you should move any active resources off the node before you upgrade so you can gracefully fail them over. However if you do an upgrade with resources still on the node, it should still work as the resources will failover during the upgrade, but it has not been tested, so we would recommend the safer experience with the planned failover.
PingBack from http://www.ditii.com/2009/06/13/rolling-upgrade-to-windows-server-2008-sp2-failover-clustering-from-rtm/
There is an issue when you install Windows 2008 SP2 remotely, if you have RDP to the server. your installation shows installing physicaly standing at the front of servers.After RDPing to server & once SP2 starts installing your connection is lost, Microst should maintain connection &Admin shouldn't use 3rd party software to connect Server using KVM over IP.
I think Microsoft should address this if you are installing remotely.
PingBack from http://wheelbarrowstyle.info/story.php?id=1537
Losing your remote desktop connection is expected when the machine reboots during installation (or any time a machine reboots). This does not impair any functionality. Once the machine is upgraded you will be able to reconnect via remote desktop using the same credentials and machine information. The steps documented above were performed entirely using remote desktop.
PingBack from http://patioumbrellasource.info/story.php?id=2835
PingBack from http://www.windowsaffinity.com/?p=2441
You mention "A best practice is to also pause the node(s) you are about to upgrade by right-clicking the node name and selecting 'Pause'." Have you tried this? In fact you will see that you can't install the service pack onto a node that is paused. You will get the following error during the validation
"The machine is clustered, but the cluster is not online or cannot be
accessed from one of its nodes. To continue determine why the cluster is not
online and rerun setup instead of rerunning the rule since the rule can no
longer detect a cluster environment correctly."
So what is the best practice?
Great question. You will want to run Validation *after* you have updated all of your nodes. Ideally you want to get out of this 'mixed mode' cluster as quick as you can as the cluster may not be as stable during this time (and running Validation will mean you're in mixed mode for even longer).
Once you have upgraded all the nodes and they're online (not paused), Validation will complete successfully.
Clustering & HA
I performed the steps outlined above.
I moved all active resources from Node B to Node A.
I paused Node B.
I applied updates and Service Pack 2 to Node B.
After rebooting, the Failover Clustering feature was no longer installed. Node A shows Node B as "down" at this time.
I reinstalled the Failover Clustering feature on Node B. Now the cluster service will not start on Node B. How can I bring Node B online so that I can update Node A?
Hi David B,
Did you resume (unpause) the node after installing the hotfix? The pause will persist over the reboot.
If you did this and are still having issues, please send me an email by clicking on the 'email' link on the upper right.