Re: Virtual Server and Cluster

Have you read this document?
http://download.microsoft.com/download/d/c/b/dcbdbbc4-a654-4c35-b4ad-67dbba35782e/VSHostClustering.doc

I setup the same using this doc as a guide. Manual shutdown of a host is a special case for failover, and there is a section in that document which describes a simple batch file which runs on shutdown that essentially stops the cluster service. This will cause your guests to first save their memory state to disk and then the guests should failover to the other node. This works in conjunction with the vbs scripts you add as resources.

One thing you may be noticing is that the resources are actually failing over, but the vm's don't start. This could be due to overprovisioning memory. This works albeit flaky at times but we run about 8 developer workstations from a pair of clustered SAN boot x64 R2 hosts. Works great *most* of the time.

To answer your question though, doing a start->shutdown is a valid test, but does not replace the case where one server simply dies (i.e. power loss, bugcheck, lockup, etc). The safest method, IMHO, is to manually disconnect the storage connections and all ethernet connections, for all intents and purposes this will appear to the remaining node as a power failure. This will test the ability for the remaining node to detect the partner is dead, and clear the reservations on the disks (via a hierarchical break_reservation), and then reserve them and start any resources dependent on those disks (your vb scripts).

~kenny

"E2tech" <E2tech@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message news:7B44C07A-B51B-46EE-9BE0-36A67641C59A@xxxxxxxxxxxxxxxx
I have 2 identical Windows 2003 servers configured in a cluster. I setup
Virtual Server per MS documentation and it is working fine. I can manually
move cluster groups from one to the other with no problem.

A problem happened when I went to do a failover test. I performed a Start >
Shutdown on one server thinking the other would pick up the cluster.
However, it didn't...the one server shutdown and the cluster never came up on
the other. I have scripts in the cluster resource to ensure VMs start on
cluster move and when I ended the Virtual Server in task manager the cluster
came up.

Is performing a shutdown on one server not a good way to test Virtual Server
cluster failover? Does anyone know if there is Microsoft documentation
explaining a Virtual Server operation in a cluster as it pertains to failover
testing?

.

Relevant Pages

  • Slow failover on Microsoft Virtual Server Cluster
    ... node Virtual Server cluster. ... I have two groups within the cluster: ... When this was first set up and I was testing failover with one or two guests, ...
    (microsoft.public.windows.server.clustering)
  • Re: Demonstration of Failover Capabilities
    ... What Rodney was suggesting was that you build a test SQL cluster and use it ... to demonstrate failover. ... Here is a link to a webcast on exactly how to build your a Virtual Server ...
    (microsoft.public.sqlserver.clustering)
  • Re: Failover fails when power failure is tested
    ... I have a two node cluster configured ... shutdown by shutting down power a failover is not executed. ... stops the cluster service and the entire cluster is unavailable until one ...
    (microsoft.public.windows.server.clustering)
  • Re: Virtual Server and Cluster
    ... Virtual Server per MS documentation and it is working fine. ... move cluster groups from one to the other with no problem. ... Shutdown on one server thinking the other would pick up the cluster. ...
    (microsoft.public.windows.server.clustering)
  • Re: Why Cluster in a test Environment?
    ... If your apps are not failover aware, ... SQL Server 2000 then takes anywhere from 30 seconds to minutes before ... a cluster but not appear in a standalone server. ... A non-clustered environment would not allow accurate stress testing do ...
    (microsoft.public.sqlserver.clustering)