For the partitioning part, the docs say NVIDIA vGPU isn't supported on XCP-ng, since it depends on a proprietary XenServer component that can't be redistributed: https://docs.xcp-ng.org/compute#vgpu. So as far as I can tell, the "20, 40 or 80 GB per VM" split won't work through vGPU. Passing a whole GPU to one VM is the documented path, on the same page. I haven't seen anyone report on H100 or H200 cards here, though, so I can't tell you if that class of card brings its own surprises. There's also a blog post doing LLM inference that way on a much smaller Tesla T4: https://xcp-ng.org/blog/2025/07/28/your-own-gpu-powered-llms-with-xcp-ng/. There's an open design discussion upstream in XAPI about GPU partitioning, MIG included (https://github.com/xapi-project/xen-api/pull/7260), but that's the XenServer side, and I don't know if or when it would reach XCP-ng. As GPU partitioning matters for your project, it might be worth opening a request on https://feedback.vates.tech/ with what you described here (H200s, LLM inference, several tenants per card, fixed VRAM per VM). From what I've seen, a concrete use case like yours carries more weight there than a plain "please support vGPU", and other people needing the same thing can vote on it.

@poddingue Yes, reads pretty much like the same error / root cause.

Thanks, that helps. Some VMs fail to the Synology but pass to S3, and some the other way round. So I think the backups on Ark are probably fine (they pass health checks, as you said at the start) and it has more to do with what each target already holds, but I could be wrong. I'd also flag the several delta jobs with different retentions all feeding Ark, I just don't know the mirror code well enough to say whether that matters. @Team-XO-Backend, posts 6 and 10 have the log and the pattern if someone gets to this one.

@mdm thanks for checking Reconnect. If I'm reading the code right, both actions have the same mistake, so one fix should cover both. You already wrote the steps out in your first post, so if you're up for it, an issue on https://github.com/vatesfr/xen-orchestra/issues would put it where the frontend team tracks their work. Their bug form asks for the XO version (or commit if you're on sources) plus your Node and XCP-ng versions, which I don't know for your setup.

Hello @ludalm13, Thanks for your contribution and sorry for not giving an answer earlier. We initially planned to review your PR while doing some additional work on the audit plugin, but our planning evolved and that work was postponed. We'll review your PR in the following days.

Just as an update on this. Did get the migrations working. Still haven't gotten the backups working. Pool is still stable otherwise. Have managed to get backups by using Veeam so it's not to much of a problem.

Thanks for the information, @ph7 . That's a typo, for sure.

@nick.lloyd that's something out of this plugin, that's ceph native. most clusters should run on msgr2 native. which ceph versions are you running?

@zorro So with purge off, Veeam stayed clean for a few nights, which fits what I saw in the lab test. If I read the docs right, the cost is that XO's reference snapshot keeps its data on the SR again, which you'd notice more on thick-provisioned storage: https://docs.xen-orchestra.com/backups-and-dr/backup-types/incremental_backups#purge-snapshot-data-cbt. Storage hasn't said yet whether purge breaking another tool's chain is expected, so I'd still call this a workaround rather than the answer. If a CBT warning comes back when you add the other VMs, the Veeam task id and the time would probably help them. And keeping both makes sense to me, since Veeam doesn't do the cross-site replication you need XO for.

@yannsionneau I don't use Ubuntu, but I have a few linux vm's ranging from Debian 13-14 to RHEL 10 and in all of them the passthrough works fine.

@fred974 Hello, Could you please provide the logs produced by XenClean after its execution? If your VM couldn't boot after XenClean, XenBootFix executed from a recovery environment should get your VM working again. Instructions are available here: https://docs.xcp-ng.org/troubleshooting/windows-pv-tools/#windows-fails-to-boot-hangs-inaccessible_boot_device

FYI, our approach with vectura is only temporary, we have other plans in few months to get rid of this potential cat & mouse game Broadcom could do by slightly change their protocol (without even talking about the potential legal risk with it).

I'm also investigating on a solution to make XCP-ng opportunistically resize device BAR; in case the firmware didn't do it (or has no support for it).

Thanks @jr-m4, reporting the small ones is useful too. From what I can see in the code, the VM view of the restore page has that refresh button and the Metadata view only has a link back, so I think your screenshots show a real gap and not a glitch on your side. I honestly don't know whether XO 5 still takes this kind of small UI fix now that most of the work goes into XO 6, so I'll leave that to @Team-XO-Frontend to decide.

Thanks @cocoon, and I owe a correction: when I said in https://xcp-ng.org/forum/post/104865 that 6.4 would make these alerts disappear for good, I overstated it. If I'm reading the code right this time, the XO server sends a ping every 30 seconds stamped with its own clock (https://github.com/vatesfr/xen-orchestra/blob/master/@xen-orchestra/rest-api/src/events/event.class.mts), and the browser compares that stamp with its own clock and shows the banner once the gap passes 40 seconds (https://github.com/vatesfr/xen-orchestra/blob/master/@xen-orchestra/web-core/lib/packages/remote-resource/sse.store.ts). So https://github.com/vatesfr/xen-orchestra/pull/9681 widened the margin from 32 to 40 seconds, but a browser whose clock runs ahead of the XO server by more than about 10 seconds would still see it. That might be why @GregBinSD saw it on the Windows PC and not on the tablet, though that's only a guess on my part. Comparing two different clocks looks fragile to me, but @Team-XO-Frontend will know far better than I do whether there's a reason it's done that way.

@irtaza9 What do you see if you run top or xtop, or "xe task-list" when you start the process?

For some reason, we have the same issue (missing disks / snapshot-of) on our environment. (XO from source commit a5077) This is a major issue for us. We use Terraform to manage our VMs, and the missing VDIs are causing Terraform to fail. As a result, we are currently unable to manage our VMs using Terraform. ➜ dacsvm-db02 git:(main) tofu plan -var-file="dacsvm-db02.tfvars" module.vm.xenorchestra_vm.vm: Refreshing state... [id=e36fcd8a-b88c-42f7-6d61-2aa639f63b4a] Planning failed. OpenTofu encountered an error while generating this plan. ╷ │ Error: Could not find client.VDI with query: {VDIId:cad7d739-35c2-4585-9991-4c2f3fc6531c SrId: NameLabel: NameDescription: Size:0 VBDs:[] Parent: PoolId: Tags:[]} │ │ with module.vm.xenorchestra_vm.vm, │ on ../modules/vm/main.tf line 16, in resource "xenorchestra_vm" "vm": │ 16: resource "xenorchestra_vm" "vm" { │ ╵

@Velocity1192 If you can wait a little, we are atively working on the new XO6 backup job uI, that will come with some changes : one schedule per job, but with conditionnal actions You will run an incremental job, and will be able to define that the schedule of sunday is a full,and even run a healthcheck on it The ETA is on 6.11 ( next month release )

@acebmxer I haven't run your script, so I can't vouch for the log lines, but the systemctl status check looks like the safer of the two to me. It shows the binary the service is really running. Whichever node comes first in your shell can be a different one if more than one Node is installed. For XOA I'd go with what @Danp said above.

For anyone else who steps through this, I've also installed the latest xen-guest-agent with the below process. Repeat for each host in the environment, 1 at a time from your Helper VM. ### ssh into your helper ### oc debug node/fedora-cmp-1/2/3 chroot /host cd /var/tmp curl -LO https://gitlab.com/xen-project/xen-guest-agent/-/jobs/6041686376/artifacts/raw/RPMS/x86_64/xen-guest-agent-0.4.0-0.fc37.x86_64.rpm sudo rpm-ostree install /var/tmp/xen-guest-agent-0.4.0-0.fc37.x86_64.rpm mkdir /var/tmp/xga && cd /var/tmp/xga rpm2cpio ../xen-guest-agent-0.4.0-0.fc37.x86_64.rpm | cpio -idmv ### From your helper... DRAIN YOUR WORKERS ### oc adm drain fedora-cmp-1/2/3 --ignore-daemonsets --delete-emptydir-data ### log back into fedora-cmp-1/2/3... ### sudo systemctl reboot ### from helper ### oc adm uncordon fedora-cmp-1/2/3 ### login to fedora-cmp-1/2/3 ### chroot /host sudo systemctl enable --now xen-guest-agent And then from XO this is what your VMs will look like. [image: image.jpeg]