Disk on different datastore getting excluded due to naming conflict?

dsercu · 2023-08-05 00:46:49

I have a VM that I'm trying to do a replica backup for as part of the following saved job:
/vmfs/volumes/629ebcf6-d370a7b0-0955-d4ae529b7524/XSI/XSIBackup-DC/xsibackup \
--replica=cbt \
"VMs(RUNNING)" \
"root@192.168.1.100:remotehost:22:/vmfs/volumes/Data01/Backups/Replicas/$(( $(date +%j) % 2 ))" \
>> /vmfs/volumes/629ebcf6-d370a7b0-0955-d4ae529b7524/XSI/XSIBackup-DC/var/log/xsibackup.log 2>&1

Mostly everything is working fine, but my issue is that one of the VMs has two hard disks configured, and only the first one is getting backed up - the other is getting excluded/skipped. From the xsibackup.log file:
2/30 VM1-flat.vmdk (CBT 1 full) 50.00 GB ESC[90m|ESC[0m Done 0.00%
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^M 3/30 VM1.vmdk 755.00 B ESC[90m|ESC[0m Done 30.83%
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
<snip>
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^M 26/30 VM1-flat.vmdk (CBT 1 full) 80.00 GB ESC[90m|ESC[0m Done 30.84%
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^M 27/30 VM1-ctk.vmdk [skipped excluded]
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^M 28/30 VM1-000002-delta.vmdk [skipped excluded]
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^M 29/30 VM1-000002.vmdk [skipped excluded]
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^M 30/30 VM1-000002-ctk.vmdk [skipped excluded]
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m
^MTotal size: 130.00 GB ESC[90m|ESC[0m Done 100.00%
ESC[90m-------------------------------------------------------------------------------------------------------------ESC[0m

I have confirmed that the scsi0:2.ctkEnabled TRUE setting is in the config file and it shows up in the vSphere config page as well.

The vmdk files are stored in different datastores, but their file names are the same - I'm guessing that is what is confusing the XSIBackup software.

Is this a bug that can be fixed, or do I need to manually rename one of the vmdk files? I have several VMs like this so I would really prefer to avoid the latter if at all possible.

Last edited by dsercu (2023-08-05 00:47:39)

admin · 2023-08-05 10:50:05

(c)XSIBackup-App detects duplicate disks and prepends the datastore name to the disk file when using --backup action. This feature will be available to (c)XSIBackup-DC in next version.

You should name disks uniquely within one VM though. (c)XSIBackup consolidates all disks in one folder in order to produce a working VM when restoring. It does its best at the time to rearrange disks in a new working .vmx file, still that is just a recomposition of the original set up. We can't assume that (c)XSIBackup will find the exact same infrastructure when restoring, as this would be counterproductive in most DR situations, thus we always aim at producing something that will work from a single directory.

When you restore such a VM you will end up having to rearrange things according to the original layout or to a totally new reality. Naming disks uniquely will guarantee that the restored VM will work out of the box without having to eventually edit the .vmx file.

You should rename disks manually to be 100% sure on what is the final result as the VSphere GUI is just an abstraction layer and doesn't always do what you would expect.
Renaming a disk is rather simple:

1/ Stop the VM.
2/ Change the -flat.vmdk file name.
3/ Edit the .vmdk file descriptor to point to the new -flat.vmdk file and remove any CBT related lines. The # Change Tracking File section.

# Change Tracking File
changeTrackPath="VM01-ctk.vmdk"

4/ Edit the .vmx file to point to the new .vmdk file descriptor and delete any CBT related lines, usually at the end of the file if enabled through --enable-cbt argument.

ctkEnabled = "TRUE"
scsi0:0.ctkEnabled = "TRUE"
scsi0:1.ctkEnabled = "TRUE"
scsi0:2.ctkEnabled = "TRUE"

5/ Reboot the VM and eventually run --enable-cbt again.

UPDATE:

There is already a version available (1.7.0.7) which addresses duplicate named disks by prepending the datastore name to them. Please note that when you restore the VM you will have to edit the .vmx file after deciding where to put those disks.

dsercu · 2023-08-05 21:00:22

Thanks for the quick response and the helpful tutorial to rename the disk. I've done that for one VM and will see how the replica job runs now. Btw, I definitely see your point about not naming disks uniquely for one VM even if they are on different datastores, unfortunately that was the configuration when I inherited this setup. :-/

Forum ©XSIBackup: ©VMWare ©ESXi Backup Software

#1 2023-08-05 00:46:49

Disk on different datastore getting excluded due to naming conflict?

#2 2023-08-05 10:50:05

Re: Disk on different datastore getting excluded due to naming conflict?

#3 2023-08-05 21:00:22

Re: Disk on different datastore getting excluded due to naming conflict?

Board footer