Unknown error returned by VMware Converter Agent

[matt-brewster]

I am experiencing errors trying to recover a large VM using converter. The VM has three disks: one 15GB, one 450GB and one 20GB. The files were created using VCB, and the corresponding VMDK files are split into 2GB chunks. The process fails at 93% every time, with the generic Unknown error returned by VMware Converter Agent[/i] message. If I exclude the large disk the process completes OK, so the error lies with converting the 450GB disk.

The LUN I am restoring to has 550GB of free space, and has a maximum file size of 512GB. I am trying to restore to a 3.0.2 ESX server, connected to an EMC CX300 SAN.

I can post the log files if it helps, but they are rather large and seem to contain many benign errors (i.e. errors that also appear in log files from successful converter operations).

Where do I go from here?

[admin]

It could be you need additional temporary "scratch" space on your VMFS LUN for the restore. Do you have a larger volume/LUN you can try to restore to?

[theanykey]

if you have a blocksize of 2mb (largest file size 512gb), then you have enough space for 450gb physical disk. If your 12gb, 450gb and 20gb are partitions from 1 physical disk, then your vmdk file will be 485gb (1545020), assuming you dont resize them - which is still enough space.

however that being said, assuming you have other VMs currently running, there is virtual memory being swapped (etc) that will utilize current free space.

you will need to find a larger VMFS volume or shutdown currently running VMs (as to determine if this is truly the cause).

[Todaysits]

Assuming space is not your issue...

Attach these disks to another VM and run a chkdsk on them. I have had this error about 10 times now and a chkdsk has fixed it everytime.

Good luck.

Scott

[matt-brewster]

I am in the process of running this problem through with a VMware engineer. So far I have got the restore working if I convert the VM to a GSX VM, then convert it to ESX and shrink the disks slightly (to force a file-level copy), although this is a very convoluted 12-hour restore process at least it's progress.

All three of my VMDK files were separate and each under 512Gb so the block size is not an issue. I tried upping it to 4Mb (just in case) and it still failed.

If I restore just the two smaller disks \*or* the one large disk these processes work - it's just when I try an do it all together that it fails.

I'll keep this thread posted on my progress.

[theanykey]

all 3 of your vmdk files were EACH under 512gb and with a blocksize of 4mb (i believe cap of filesizes no larger then 1024gb) you are OK.

However, if these are simply partitions from the SAME physical disk, then it will produce ONE complete vmdk file (1 vmdk per physical disk). this means 512512512 (3x512 partitions) = more then 1024GB.

So the solution to this problem would be to run converter 3 times (once per partition) to end up with 3 separate vmdk files.

[matt-brewster]

These are definitely not partitions on the same disk, they are three separate disks.

The engineer has got me to try selecting the "modify disk size" option to knock each disk down by a few MB, to force file copy instead of block copy. If I do this it restores OK but takes an absolute age (~13 hours). It's good news because I know I can restore the VM, but that timeframe is too high. I'm going to persist with the engineer to get it working in block copy mode...