I have heard numerous war stories about the difficulty of diagnosing vMotion problems.

This week while researching vMotion configuration issues, I encountered a situation where a user had DRS and HA enabled on a cluster that contained six hosts.  Five of the hosts were consumed to about 40%, but one of the machines wasn’t running any VMs at all.  Additionally, the user had multiple VMs that were cloned and he had selected to be created on different hosts in the cluster.  The user expected the VMs to balance themselves across all hosts in the cluster, but clearly this wasn’t the case.

It all sounded pretty normal to me, except for the part about one host not running any VMs.  Digging in a little deeper it was discovered that the DRS Automation Level was set to manual and the cloned VMs all had attached CDROMs and serial devices.  Any of these was enough to prevent automated load balancing from working. 

I have heard of other users with vMo inhibitors.  One had a 64-bit VM that wouldn’t vMO because he had a bit set wrong in one of the host’s BIOS.  During a training class, one student confessed that he hadn’t been able to get vMo to work for more than a year.  It was a big part of his reason for taking the class.

Clearly vMotion has lots of configuration dependencies and it is can be very difficult to get them all configured correctly.  But once configured correctly, VMs can quickly move from host to host as need to balance the load.  After you get it set up the challenge is knowing where the VMs are at any point it time.

Virtual Product Management, vMotion

This entry was posted on Friday, February 6th, 2009 at 4:23 am and is filed under Virtual Product Management. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.