System Center Virtual Machine Manager 2012 Using WSUS To Update Hyper-V Cluster Hosts & Other Fabric Servers

One very neat feature in System Center Virtual Machine Manager 2012 (SCVMM2012), which is currently in Béta, is the integration with WSUS to automate the patching of Hyper-V cluster hosts (+ the Library servers, SCVMM servers and the update servers, i.e the fabric). The fact that SCVMM 2012 will give you the complete toolset to take care of this is yet a great addition to the functionality available in Virtual Machine Manager 2012. More and more I’m looking forward to using it in production as it has so many improvements and new features. Combine that with what’s being delivered in System Center Operations Manager (SCOM2012) and the other member of the System Center family and I’m quite happy with what is coming.

But let’s get back to the main subject of this blog. Using WSUS and SCVMM2012 to auto-update the Hyper-V cluster hosts without interruption to the virtual machines that are running on it. Up until now, we needed to script such a process out with PowerShell even tough having SCVMM2008R2 makes it easier since we have Maintenance Mode in that product which will evacuate all VMs from that particular host, one by one. The workflow of this script looks like this:

  • Place the Host Node in Maintenance Mode in SCOM 2007 R2 (So we don’t get pesky alerts)
  • Place the Host Node in Maintenance Mode in SCVMM2008R2 (this evacuates the VMs from the host via Live Migration to the other nodes in the cluster)
  • Patch the Host and restart it
  • Stop Maintenance Mode on the host node in SCVMM2008R2 (So it can be used to run VMs again)
  • Stop Maintenance Mode on the host node in SCOM 2007 (We want it to be monitored again)
  • Rinse & Repeat until all Host nodes are done. Depending on the size of the cluster you can do this with multiple nodes at the same time. Just remember that there can be only one Live Migration action taking place per node. That means you need at least 4 nodes to do something like Live migrate from Node A to Node B and Live Migrate from Node C to node D. So you need to work out what’s optimal for your cluster depending on load and number of nodes you have to work with.
  • Have the virtual machines redistributed so that the last host also gets its share or virtual machines

Now with SCVMM2012 we can do this out of the box using WSUS and all of this is achieved without ever interrupting any services provided by the guests as all virtual machines are kept running and are live migrated away from the host that will be patched. If you’re a shop that isn’t running System Center Configuration Manager you can still do this thanks to the use of WSUS and that’s great news.  There is an entire sub-section on the subject of Managing Fabric Updates in VMM 2012 already available on TechNet. But it goes beyond the Hyper-V host. It’s also the SCVMM server, the library server, and the Update Server that get patched. But don’t go wild now, that’s the entire scope of this. That means you still need regular WSUS or SCCM for patching the virtual machine guests and other physical servers. The aim of this solution is to patch your virtualization solution’s infrastructure as a separate entity, not your entire environment.

So how do we get this up and running? Well, it isn’t hard. Depending on your needs and environment you can choose to run WSUS and SCVMM on the same server or not. If you choose the latter please make sure you install the SWSUS Administration Console on the SCVMM server. This is achieved by downloading  WSUS 3.0 SP2 and installing it. Otherwise, just use the WSUS role from the roles available on Windows 2008 R2. This handles the prerequisites for you as well. It is also advisable to install the WSUS role on a separate server when your SCVMM 2012 Infrastructure is a highly available clustered one. For more information see http://technet.microsoft.com/en-us/library/gg675099.aspx . Time-saving tip: create a separate domain account for the WSUS server integration, it can not be the SCVMM 2012 service domain account.

Make sure you pay attention to the details in the documentation, don’t forget to install the WSUS 3.0 SP2 Administration Console on the SCVMM 2012 server or servers and to restart the SCVMM service when asked to. That will safe you some trouble. Also, realize that this WSUS Server will only be used for updating the SCVMM 2012 fabric and nothing else. So we do not configure anything except the operating system (W2K8R2) , and the languages needed. All other options & products that are not related to virtualization are unchecked as we don’t need them. Combine this with dynamic optimization to distribute the VM’s for you and you’re golden. A good thing to note here is that you’re completely in control. You as the virtualization infrastructure / SCVMM 2012 Fabric administrator control what happens regarding updates, service packs, …

You do need to get used to the GUI a bit when playing around with SCVMM2012 for the first time to make sure you’re in the right spot, but once you get the hang of it you’ll do fine. I’ll leave you with some screenshots of my lab cluster being scanned to check the compliance status and then being remediated. It works pretty neatly.

Here are the hosts being scanned.

You can right-click and select remediate per baseline or select the host and select remediate form context menu or the ribbon bar.

The crusader host is being remediated. I could see it being restarted in the lab.

New KB Article 2494016 Related to Windows Server 2008 SP1 Hyper-V: Stop error 0x0000007a When Using CVS in Redirected Access

Well not a day after my blog post Extra Info on Clustering & Hyper-V with Dynamic Memory When You Start With Windows Server 2008 R2 SP1on important hotfixes for Hyper-V clustering with Windows Server 2008 R2 SP1 Microsoft releases a new hot fix for issue below. I’ll add it to the post to keep up to date.

Stop error 0x0000007a occurs on a virtual machine that is running on a Windows Server 2008 R2-based failover cluster with a cluster shared volume, and the state of the CSV is switched to redirected access

The KB article with instructions on how to get the hot fix is here: http://support.microsoft.com/kb/2494016/en-us?sd=rss&spid=14134

The scenario is detailed as follows:

Consider the following scenario:

  • You enable the cluster shared volume (CSV) feature on a Windows Server 2008 R2-based failover cluster.
  • You create a virtual machine on the CSV on a cluster node.
  • You start the virtual machine on the cluster node.
  • You move the CSV owner to another cluster node, and you change the state of CSV to redirected access.
  • The connection that is used for redirected access is switched to another connection when one of the following scenarios occurs:
    • The cable for local area network (LAN) is disconnected.
    • The related network adapter is disabled.
    • The connection is switched by using Failover Cluster Manager.

In this scenario, you receive a Stop error message that resembles the following in the virtual machine:

STOP 0x0000007a ( parameter1 , parameter2 , parameter3 , parameter4 )
KERNEL_DATA_INPAGE_ERROR

Note

  • The parameters in this Stop error message vary, depending on the configuration of the computer.
  • Not all "0x0000007a" Stop error messages are caused by this issue.
  • You may also receive other Stop error messages when this issue occurs. For example, you may receive a "0x0000004F" Stop error message.

Extra Info on Clustering & Hyper-V with Dynamic Memory When You Start With Windows Server 2008 R2 SP1:

Here’s a quick “heads up” if your starting to use or thinking about using Windows Server 2008 R2 SP1 for your Hyper-V clusters. The most common issues I’ve seen in the wild are:

  1. https://blog.workinghardinit.work/2011/04/01/kb2230887-hotfix-for-dynamic-memory-with-windows-2008-standard-web-edition-does-not-apply-to-without-hyper-v-editions/ This one is being worked on and the hotfix will be re-released to support the “Without Hyper-V” SKU of Windows Server 2008 SP2.  It’s a simple oversight but one that can be important when your Hyper-V clusters are filled with that SKU.
  2. We also got bitten by this one Déjà vu Bug: The network connection of a running Hyper-V virtual machine may be lost under heavy outgoing network traffic on a computer that is running Windows Server 2008 R2 SP1, but the hotfix was already available luckily.
  3. And than one to head and to read the TechNet forum about Cluster Validation Bug In Windows 2008 R2 SP1 – Disk has a Persistent Reservation on it. They are also working on a fix. I’ve written a blog post on this and I suggest you read it and also take note of the discussion in the TechNet forum.

    UPDATE: The hotfix for issue 3 has become available today, April 26th 2011 as announced on the TechNet forum here:

    A hotfix is now available that addresses the Win2008 R2 service pack 1 issue with Validate on a 3+ node cluster. This is KB 2531907. The KB article and download link will be published shortly, in the mean time you can obtain this hotfix immediately free of charge by calling Microsoft support and referencing KB 2531907.   Update 27/05/2011 Here is the link: http://support.microsoft.com/kb/2531907/en-us?sd=rss&spid=14134

An other one that I haven’t seen in the wild is:

Windows Server 2008 R2 installation may hang if more than 64 logical processors are active. There is is a workaround and a hotfix for this one.

Issue: When you try to install Windows Server 2008 R2 on a computer that has more than 64 logical processors, Windows Setup may stop responding in one of the following operations:

  • Initialization of Windows Setup
  • One of the two restarts that are required to complete Setup

Cause: This issue occurs because of an error in the Network Driver Interface Specification.This issue occurs because of an error in the Network Driver Interface Specification (NDIS) driver.
When a computer has more than 64 logical processors, the NDIS driver does not correctly handle some operations. Therefore, the computer encounters stop responding issues and other system failures.

I don’t have any nodes under my care who have more than 64 logical processors so that’s why I guess Smile But with ever more cores available you it’s bound to happen in the near future.

Update 2: To keep me busy this KB article was released within 24 hours of me posting this blog on a BSOD with CSV and redirected access for witch a hot fix is available

Stop error 0x0000007a occurs on a virtual machine that is running on a Windows Server 2008 R2-based failover cluster with a cluster shared volume, and the state of the CSV is switched to redirected access

The KB article with instructions on how to get the hot fix is here: http://support.microsoft.com/kb/2494016/en-us?sd=rss&spid=14134

The scenario is detailed as follows:

Consider the following scenario:

  • You enable the cluster shared volume (CSV) feature on a Windows Server 2008 R2-based failover cluster.
  • You create a virtual machine on the CSV on a cluster node.
  • You start the virtual machine on the cluster node.
  • You move the CSV owner to another cluster node, and you change the state of CSV to redirected access.
  • The connection that is used for redirected access is switched to another connection when one of the following scenarios occurs:
    • The cable for local area network (LAN) is disconnected.
    • The related network adapter is disabled.
    • The connection is switched by using Failover Cluster Manager.

In this scenario, you receive a Stop error message that resembles the following in the virtual machine:

STOP 0x0000007a ( parameter1 , parameter2 , parameter3 , parameter4 )
KERNEL_DATA_INPAGE_ERROR

Note

  • The parameters in this Stop error message vary, depending on the configuration of the computer.
  • Not all "0x0000007a" Stop error messages are caused by this issue.
  • You may also receive other Stop error messages when this issue occurs. For example, you may receive a "0x0000004F" Stop error message.

Cluster Validation Bug In Windows 2008 R2 SP1 – Disk has a Persistent Reservation on it

Pretty soon after the RTM of Windows 2008 R2 SP1 release we were discussing a bug on the TechNet forum (Hyper-V Cluster issues after applying Win2008 R2 SP1 on a 3 node Cluster!) here. If you have a Windows 2008 R2 SP1 cluster with more than 2 nodes you get the following warning:

List Potential Cluster Disks

Disk with identifier 2sef8cdf has a Persistent Reservation on it. The disk might be part of some other cluster. Removing the disk from validation set

“Normally” you would expect such a warning if the LUN ever belonged to another cluster and it needs the old reservation cleared. To do that you would use following command on the node that throws the warning (where in this example the disk is disk 2 in disk manager/diskpart) and after making sure it is not in use anywhere else in the SAN

"cluster node clusternode1 /clearpr:2"

However this is not the cause here as were most others in this discussion. And I’m pretty no san software or MPIO software is putting a reservation on there either so what is this? A bug? Well yes, it has been confirmed by Microsoft support that is is indeed a bug an that is fix will be made available by April 18th2011 .

This was not a show stopper bug, but it could be one if you needed to add a host to a cluster and confirm all is well and supported. However if you’re certain you’ve done everything right you can choose not to run cluster validation.

I will update this blog with more information when the fix becomes available.

UPDATE:  The hotfix has become available today, April 26th 2011 as announced on the TechNet forum here:

A hotfix is now available that addresses the Win2008 R2 service pack 1 issue with Validate on a 3+ node cluster.  This is KB 2531907.  The KB article and download link will be published shortly, in the mean time you can obtain this hotfix immediately free of charge by calling Microsoft support and referencing KB 2531907. Update 27/05/2011 Here is the link: http://support.microsoft.com/kb/2531907/en-us?sd=rss&spid=14134