#190 VMFS5 Disk Clone Fails With Clonezilla Live 2.1.2-43-amd64 and Versions and Newer

open
None
3
2014-04-17
2013-12-03
ArtieMan
No

I've come across a significant and reproducible bug that I'd like to report here. I've been trying to create Clones of various machines with ESXi 5.0 and ESX 5.1 hypervisor installed on them and a single VMFS-5 local Disk/LUN datastore within ESX using Partclone. In my test cases, the ESX host that I'm cloning has a single virtual machine created in it with a thin-provisioned single virtual disk (running an OS anywhere from Windows 2003, Windows 2008 to various Linux versions... they all fail in the same fashion regardless).

This error functionality that I'm about to describe only occurs in newer versions of the Clonezilla live disk when the datastore is of type VMFS-5 and does not happen with VMFS-3 datastores. This functionality also worked just fine in Clonezilla Live 1.2.12-60 AMD64, but has become broken at some point between that version and version 2.1.2-43.

Here is what I'm seeing (2 different variants of the VMFS-5 functionality being broken on different versions of Clonezilla and ESX):

1) When tested on ESX 5.1, using Clonezilla Live 2.1.2-43-AMD64, I can create the clone image of my ESX 5.1 host machine and get no reported errors. When I install the created image to a different machine and boot it up, ESX starts fine and I am able to redetect my datastore and the VM that it contains and can also see my original Virtual Machine in the list of available VM's. Everything looks OK until I actually try to start up the Virtual Machine. When I start the virtual machine up, the virtual disk appears blank and unformatted. It would appear that the VM's virtual disk data is not being copied over within the datastore to the target machine although the rest of the VM's attributes have been copied over.

2) When I attempt to create a Clone of the ESX 5.0 host machine using Clonezilla Live 2.2.0-31 AMD64 or 2.1.2-43-AMD64, I get a fatal error during the clone creation phase:

*** glibc detected *** partclone.vmfs5: free()  invalid next size (Fast) 0x00000000025a71a0   ***
====== Backtrace   =======

(I'm leaving memory dump info out for brevity here)

====== Memory map: =======

(I'm leaving memory dump info out for brevity here)

/filesystem.squashfs/usr/sbin/partclone.vmfs5

(I'm leaving memory dump info out for brevity here)

[vsyscall]
Checking the disk space...
Failed to use partclone prgram to save or restore an image!
Press "Enter" to Continue"

The image creation has failed at this point and the only way to break out of it it is with CTRL-ALT-DELETE, at which point it says "System going down for reboot now, sending kill signal", and then I get a bunch of PartClone screens that pop up anyways and attempts to finish cloning the rest of the disk partitions before finally shutting down the system.

I can provide more detail if necessary.

Again, I'd like to emphasize that this functionality all works in Clonezilla Live 1.2.12-60 AMD64 but became broken at some point between then and the current versions of Clonezilla Live.

Discussion