Copying vmdk files, checksums don't match

[2k2f4i]

I am running esxi 4 currently, and I am getting ready to migrate my vitrual servers to an esxi 5 box. I added a stand alone sata drive to the esxi 4 server. I shut down the virtual server and copied the files from the old datastore (a raid 5 array) to the stand alone drive (datastore). The virtual server has multiple drives so I have 4 vmdk files to copy. I originally copied the vmdk files using the datastore browser. Once the copy completed I used the md5sum command to check the integrity of the copy compared to the original and the checksums do not match. I then attempted to use the cli and issued the vmkfs -i command to copy the files individually (one at a time). Again the checksums do not match. The block size on both datastores is the same (4mb) so I am unsure why my checksums are not matching after copying. Is this normal? The server has no snapshots etc.

I used this link for the correct procedure-

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=900

[gregsn]

What kind of controller is the standalone drive connected to?  Is it an on-board controller of some sort or is it connected to your RAID controller? 

[2k2f4i]

Both the sata array and standalone sata drive are connected to the same sata raid controller (ARC-1110). Also just copied the vmdk files again, and I was able to create a new virtual machine and mount the disks to it without errors although the checksums still don't match? 

[gregsn · Accepted solution]

It could be possible that there are bit errors occurring during the copy.  I have seen this happen even with high end RAID controllers which usually get fixed with firmware updates.  You might want to check if there are any updates for your controller.

To find out if you are getting bit errors, try the following:

Make one copy from the array to the single disk and check the md5 on the single disk.  Note it down.  Make a second (presumably identical) copy from the array to the single disk and check the second md5 on the single disk.  If the two md5's on the single disk match, then there is something in the copy mechanism which is altering the files slightly and resulting in a different md5 from the original on the array.  If they don't match, then it is likely that you are getting bit errors from the controller.

[continuum]

I would not expect that the md5 checksum matches after vmkfstools -i

If you want to use checksums create them inside the VM

[2k2f4i]

Thanks all! Gregsn was correct in the issue being limited to the sata controller. When I first setup the box, I had to manually install the driver for the controller since the controller was not "officially" supported in esxi 4. I copied different vmdk files from the internal scsi array and the checksums matched on both the 2TB drive and the scsi array.

I did have issues however when I tried to disconnect the solo 2TB sata drive, and connect it to my new server. The drive could not be added as a datastore. I ended up wiping the drive, adding it as a datastore to the new server and installing openfiler. I setup a virtual disk on the 2TB drive in openfiler. I created a lun there and hooked up the old server to the new openfiler lun via iscsi through bmware. I have copied a few vmdk's and the checksums match! It is slower than directly connecting the 2TB sata drive to the old server, however it should be easy to configure the new server to use the same iscsi lun and my checksums matching make it worth it.

Thanks all!

John