I’ve had the displeasure of being greeted first thing in the morning by VMWare’s purple screen of death (PSOD) both in my home lab and in production. I seemed to have noticed a trend that once we had more than 2-3 2012 R2 guests on a single ESXi 5.5 host at somewhere round 40 days of uptime the system would purple screen with the message shown below:
After lots of digging and frustration I was able to find a VMware KB article (http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2059053) that identified the issue being associated with the E1000E Virtual NIC adapter. The thing that struck me as odd about this is that the E1000E Virtual NIC is default for the 2012 and 2012 R2 servers in the desktop vSphere client, however the vCenter web client gives you a VMXNet3 adapter by default.
So long story short resolution involves capturing your IP configuration info, shutdown the VM (scheduling downtime like a responsible admin), remove the E1000E adapter and add a VMXNET3 Adapter. Please note that the VMXNET3 does require VMware tools to be installed on Windows Server in order to recognize the adapter. Below are the steps to complete this operation:
Right click your VM and choose edit settings:
Once you’ve selected Edit settings click on your virtual NIC and look at the right hand side of the window to see what type of adapter it is:
To remove the adapter click the remove button at the top of the windows with the Network Adapter Selected. Then click the Add button to add a new virtual NIC:
Make sure to select the VMXNet3 from the dropdown list:
Follow the wizard to victory, then reboot, install VMware tools if not already installed, and reconfigure your IPs. Total downtime for this operation should be 5-10 minutes per 2012 and 2012 R2 VM.
After implementing these steps I have successfully avoided another PSOD in both lab and production. Several other colleagues with similar issues report the fix above resolved the problem for them as well.