2

We have a strange issue when attempting to create a snapshot in one of our regional environments. We have 4 VMs, 2 of them have mulitple VMDKs spread onto different datastores. When a snapshot is created, all the VMDKs (except for the first), seem to vanish. If you look at the properties of the VM, the path to the VMDKS points to the datastore where the primary VMDK is, and the file name is some random garbage. If the snapshot is deleted everything returns to normal. Has anyone ever seen this?

I'm using ESX3.5

Thank you in advance.

Andy

==============UPDATE==============

Here is some more information. I just created a test machine with 7 VMDKs. 1 for the OS. and 6 others for data. All of the VMDKs are in seperate datastores. I take a snapshot of the machine, all of the 6 VMDKs loose their reference to the actual VMDK files. The all point to 64KB VMDK files in the datastore where the OS VMDK is located. These 64KB vmdks didn't exist until the snapshot was taken. When the snapshot is deleted, everything goes back to normal.

1 Answers1

1

Yes, this is normal operation for ESX 3.x. Basically it (incorrectly I think) puts all the differential VMDKs on the same datastore as the .vmx file. For example, if you have a VM with a VMDK_1 file on DS1 and another VMDK_2 on DS2 and you create a snapshot, you will see two new files on DS1:

  • VMDK_1-000001.vmdk
  • VMDK_2-000001.vmdk

This can be real disaster if DS1 has very little extra storage and your VMDK_2 file is heavily used. We had a situation happen to us where the C: drive of a file server was on DS1 and the file sharing drive (F: in our case) was on DS2. A snapshot was taken during a VCB backup and ran for about a week before it was discovered. By the time it was discovered, DS1 was pretty much out of space and we almost lost a server while trying to consolidate the snapshot... it was a bad deal.
So you want to make sure the Datastore with your .vmx file on it has enough room to save the snapshot .vmdk files.

Scott Lundberg
  • 2,364
  • 2
  • 14
  • 22