We're currently running a vCenter farm with 4.1 vSphere hosts. I'm setting up a fresh 5.1 farm in our test environment.
To test out some guests in the new farm, I've mounted up the NFS stores that are currently being used in our 4.1 test environment. I've unregistered a few guests from the 4.1 farm (remove them from inventory), then I register them and power them up in the 5.1 farm. In the 5.1 farm, the guest CPU utilization is VERY high. The processors spike to 100% when doing the most menial tasks (like opening up Notepad). If I unregister it from the 5.1 farm and put it back in the 4.1 farm, everything is fine again. I've upgraded VMware Tools, verified that Virtualization, Hyperthreading, etc. are enabled in the BIOS, but I can't seem to figure out why there is such a drastic change in CPU utilization.
I've migrated one of these VM's to local storage and experienced the same issue, so this rules out latency or anything with NFS. Has anyone else experienced similar issues with 5.1? Any other ideas of what I should check for?
very interesting problem and a good test one must say -- can you share esxtop o/p while the guests are on esx 5i host -- also please share messages file from the hosts
what does the CPU utilization look like from the guest ? does it happen with guests of other OS types?
Sorry about the delayed response. The problem turned out to be the unsupported host firmware version that we were running in our Cisco UCS servers. We had been using 2.0(1). We upgraded to 2.1(1) (it wasn't totally clear from Cisco's interoperability matrix whether or not this was supported), but the problem continued to occur. The sweet-spot version ended up being 2.0(4). Very weird issue, but everything is running great now.