I have just found a bug in ESXi 4.1 about snapshot and single-file hard disk. It took me quite some time to realize it's related to single-file hard disk and I'm able to reproduce the problem.
This is what had happened:
I was working on an old VM in VMware Workstation 10.0.1.
The VM was created in Workstation 6 or 7 so its hardware level = 6.5-7.x
Its hard disk is in single-file and not preallocated. Maximum size = 16 GB.
I then uploaded this VM to an ESXi 4.1.0-348481 via vSphere Client. So far so good and the VM started without any problem as expected. Then I shut it down and took a snapshot. On the next reboot, I got the following message in text mode on my screen:
A disk read error occurred
Press Ctrl+Alt+Del to restart
As I said earlier, I didn’t know what happened and thought I might have corrupted the hard disk accidentally. I tried to remove the snapshot but that didn’t help. Then after several remakes and re-uploads, I found that if I did a “Delete All” on snapshot, the problem went away!! Then I re-opened storage and found that there are two vmdk files even though there is no snapshot and on fresh upload.
After that, I cloned the hard disk with Ghost and made sure that this time hard disk file is in multiple files. This time, uploaded VM was running without problem even if I created some snapshots and that’s how I realized this serious bug in ESXi 4.1 for snapshot and single-file hard disk.
I have done a little search on VMware KB. The nearest KB is http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1008103 but it doesn’t exactly match my problem.
Information on the ESXi host:
Physical server = Dell PowerEdge R710
ESXi 4.1.0-348481, installed inside an SD card
“scratch” partition in local drive
If anyone has more information on this matter, please tell us.