Xen Orchestra 5.110 V2V not working

[farokh]

So this will be a little convoluted, so bear with me.

I am evaluating replacing my ESXi hosts with XCP-NG. I currently have one XCP-NG host running 8.3.

I have 2 XO machines set up, one is the official one that requires a contract to use all the features and one is a community edition that I installed on Ubuntu 24.04 LTS using the installer script on GitHub (at https://github.com/Jarli01/xenorchestra_installer).

The official XO seems to be stuck at 5.109.1 and claims there are no upgrades available.

On the CE I used the command line update script and it appears to have gotten updated to 5.110, although it's hard to verify everything since the CE disables everything having to do with updates. At the command line, when I checked, I'm running node v22.19.0 and rpm 10.9.3, which are newer than the ones on the official XO.

So, I had tried to use V2V under 5.109.1 and then discovered that it won't work with the version of ESXi I'm running (in this case 7.0u3).

Then I saw the blog about V2V under 5.110 being able to handle all versions of ESXi, so I gave it a try on a relatively small VM that I can easily restore if I need to.

I thought I had followed the instructions in the blog as well as the V2V interface and installed everything required, but now I'm not so sure it all worked.

When I go to the Import From VMware, i see the following:
nbdinfo : ok
nbdkit : ok
nbdkitPluginVddk : ok
vddk :
There are green check marks in front of the first 3 (which have ok), but there's no check mark on the vddk and no ok following it.
I downloaded it from Broadcom and dragged it into the box that originally appeared asking me to install it. However, now I don't get prompted to install it.

So, when I attempted to migrate this one VM, The VM got powered off (not shut down), then nothing. No more tasks on the ESXi host, and nothing happened on the XO screen.

I finally gave up, clicked disconnect and tried to power on the VM on the ESXi host. It failed, complaining that there was a snapshot file missing, which, if I recall correctly, is what happened when I tried to migrate it using the previous version.

I deleted the VM and restored it from a backup, so it's back up and running.

Any ideas on what I should do next?

Thanks!

[olivierlambert]

Hello there!

1. For XOA, are you sure you are on the "latest" release channel? Version reported on XO sources doesn't mean anything, you should see the commit number. There's no "stable/latest" on anything outside XOA
2. For the rest, it's really weird, @florent might ask some specifics when he's around

[farokh]

OK, I switched the release channel to latest (I assumed that was for betas and such), and it updated to 5.110.

Now, when I tried to use Import From VMware, it prompted me to install the ndbkit, which I did, it then prompted me to upload the vddk library from Broadcom, which I did and clicked on the Import and install the Vddk library button. It's been sitting spinning now to something like 30 minutes. How long should I wait before reloading the page or something?

Thanks.

[olivierlambert]

@farokh I think it's a known bug, just refresh the page

[farokh]

OK, I refreshed the page and I got the same thing I got in the CE with the list of items with OK, etc. Here's a screenshot:

So, is this correct and I should be able to do a warm import from ESXi 7.0u3?

Thanks.

[olivierlambert]

It's the way it should work yes, having warm migration. What kind of storage are you using on VMware side?

edit: since you lost once your VM, double check your backup on VMware side in case, better safe than sorry.

[farokh]

The VM I'm looking to import is running on ESXi 7.0u3, and the storage is on local VMFS storage.

I have a backup, which is what I used to restore it last night, so I'm good that way.

Just so I'm clear, do I have to do anything in order to prep before starting the import?

Thanks.

[olivierlambert]

I don't think so (at least from the top of my head)

[farokh]

OK, so I started the import. These are screenshots from the XO and the ESXi.


So, I started the process, and on the ESXi side, it powered off the VM, and then nothing else (as per the task list).
I left it alone until now, and at some point, the import screen on XO went away, and it shows the imported VM.

However, this time, when I powered up the original VM on the ESXi host, it did power up without giving me an error.

When I checked the event log on the ESXi host, I found the following:

So, the XO machine clearly downloaded files, but there was nothing at all displayed on the import screen with a status.

So, other than abruptly powering off the VM, it appears to have worked.

I'll post more once I've played with the imported the VM, but I guess there's still the issue where it's powering off the VM (not even doing a shutdown).

[farokh]

Another thing to note, the original VM had 5 total network interfaces, and the imported machine has 5 network interfaces, but all 5 were connected to the same interface (which was the only one selectable during the import).

[farokh]

Starting to play with the imported VM, and there are 2 snapshots, both from the import process. I'm going run the VM with them for now and then try deleting them once I see it's working.

[florent]

@farokh If my understanding it correct : it migrated the source VM correctly , without breaking anything on esxi side, but it would be better if the V2V gives more feedback to the user ?
Also, you would like to be able to select different networks as target ?

[farokh]

@florent That's a good summary, however, V2V did power off the VM before starting the import, which defeats the point of a warm migration, and the power off is abrupt, not allowing the OS to cleanly shutdown, which means that there's a potential for data loss.

Thank you.

Edit: I was just looking at the blog post for 5.110 and I noticed this:
The interface has been polished as well. There’s now a progress bar visible from the start of the migration, speed metrics are recorded in the VDI, and snapshots are created at each step. This allows the process to pause and resume safely without needing to create multiple VMs. Finally, UEFI VMs are no longer forced into Secure Boot mode, avoiding previous boot issues.

I guess I'm confused as to why I didn't see any of this?

[florent]

@farokh do you have at least one snapshot on the source ? The data migrated while the VM is running are the data before the last snapshot

The power off is effectively a call to powerOff, is there a better alternative (that can work with or without the vmware tools installed) ?

the progress bar are visible in the task view, or on the disk tab on the VM being imported ( not yet in the form , this will probably wait for the XO6 version of this page )

[farokh]

@florent Yes, I made sure that I had a snapshot.

I guess my main question is why the VM is being powered off? It's supposed to be a warm migration, meaning the VM is running while the data is migrated, correct?

But, assuming that you have to power the VM down, and I haven't done any programming or scripting with ESXi, but I can tell you that the web interface can tell if the VM has the tools installed. If the tools are installed, you should be able to send a Shutdown request, which should shut the machine down gracefully, allowing the OS to perform tasks prior to powering off. You can wait a few minutes and if the machine hasn't powered off yet, you could then issue the power off command to stop it.

Next time I test a migration, I'll check the task and/or disk to see the status.

Thanks.

[farokh]

Hi, just wondering if there's any update on why the VMs are being shutdown for the migration.

Thanks.

[olivierlambert]

We made a patch release of XOA, improving some aspects. Have you tried it?

[farokh]

@olivierlambert No, I wasn't aware. I'll take a look.
Thanks.

[farokh]

Is there a blog or forum post about the updates? I can't seem to find anything.

Thanks.