This weekend I was busy with a very obscure VMFS-recovery.
An iSCSI-volume of a Qnap-storage had about 2/3 of its vmdk-files damaged.
Those flat.vmdks that appeared to be damaged actually were moved 224 sectors towards the end of the physical VMFS-volume.
So we could fix some of them by adding another vmdk of the same size to the VM.
Then we booted the VM into Linux and cloned the disks like this:
dd if=/dev/sda of=/dev/sdb bs=512 skip=224.
This obscure offset is something I have never seen before.
The VMFS-volume itself was properly alligned with a 1MB offset for the start of the VMFS-partition.
Has anybody seen something like this before or has a theory which would explain such a behaviour ?
Additional info: no Raid5 problems , no unexpected power failures , behaviour appeared with out any apparent problems or warnings before.
Suggestions or theories are welcome.
Recovery results so far: most important SQL-servers were recovered, we still have some problems with Windows 2008 and Windows 2012 boot disks that are readable but no longer bootable.
The only explanation I can think of is - QNAP?
All jokes aside, this is something that probably only VMware engineering can untangle.