During testing of our system for slow drive performance on a bunch of VMs in one of our VMware clusters (http://communities.vmware.com/thread/159179?tstart=0), we shut off the VMs. After that we adjusted some NFS settings on our Netapp filers and restarted the NFS service. Note, our VMs were completely shut down, but the ESX hosts remained on. They lost connection to the datastores, but reconnected successfully.
We brought up each VM. Three of the four VMs started normally, and their drives used the correct datastores. One of them however did not.
We have an OS, App, and Swap datastore for all VMs to use. C drive is on OS, App is on E, and Swap is on H or something. Our EMC EmailXtender server VM was set up to use two additional datastores for the application. Those were named xtendermq_index (contains message queuing and index drives (D and E)- this machine does not use App01 datastore), and xtender_containers (contains our container volumes, this is drive F.
When we brought up the server again, the swap drive (H) was brought up on the OS datastore, and Drive D and E were again using OS datastore. New .VMDK files were created on the OS datastore, but the old ones are still present on the proper datastores, but the machine isn't utilizing those stores anymore. Now i've got vmdks on the OS datastore named servername_000009.vmdk. It seems VMware decided to create new vmdk files on the OS datastore. I'm not sure what it did with the data to ensure that everything continues to work. The vmdk files on the datastores they were previously using are still present. Are the new vmdk files just pointers to the true vmdk files on the proper datastores?
It's very strange and I'm not sure why this happened. I mean, the VMs were shutdown completely...one would think when they came back up, they'd use the datastores we had the configured the VM harddisks to use. Any thoughts? I hope I was able to explain it properly...
Thanks in advance!
