VDI not showing in XO 5 from Source.

[olivierlambert]

Thanks!

[wilsonqanda]

Hello All,

Found some roundabout solutions you may want to try:
https://xcp-ng.org/forum/post/101370

[Pilow]

@wilsonqanda tried your workaround on a halted VM, and it worked !
If i snapshot -> disk still invisible
delete snapshot - > disk still invisible

but

if i snapshot --> disk still invisible
revert snapshot (with take snapshot option) --> disk APPEAR again
delete the two snapshots -> disks still there

edit : even without the take snapshot before revert, it is working, tried on another VM

[wilsonqanda]

@Pilow Lol glad I was able to help. Its only if you have a few VMs having hundreds will be a nightmare... giod luck to those that has that issue. In the meantime I will use the method I mention for now.

[Pilow]

@wilsonqanda @anthoineb @olivierlambert

we are now impacted on a SR that didn't have the problem... now all disks are invisible and DASHBOARD/HEALTH show some strange dates on base copies...

halt/snap/revert correct the issue but needs to be done on each VM

all VDIs seem to be regrouped strangely on unique base copy

there is something seriously wrong happening

no impact on production of VMs though, just management issues in XOA web ui
even backups are OK.

halp.

[Pilow]

i could be wrong but in

/usr/local/lib/node_modules/xo-server/dist/xapi-object-to-xo.mjs

I see a lot of

if (obj.is_a_snapshot || obj.$snapshot_of !== undefined)

seems to be a way of managing both "old version" and "new version" (or "broader version") to define if a VDI is a snapshot
there has been an evolution in the code at some point ?

and someone somewhere in recent updates (10 dec) forgot the || on some important action

my recent SR problem appeared when I was snapshoting/deleting VDI/reverting snapshot
all VMs suddenly didn't have visible VDIs anymore on the whole SR.

I could be wrong, but the fact that ALL VDIs are now seen as snapshot seems smelly with this ||

https://github.com/vatesfr/xen-orchestra/commit/85596da79217070bf4431135bbb5b0d2cf04e45b

0 spacotte-vates committed to vatesfr/xen-orchestra

fix(xo-server): improve handling of xapi snapshots (#9231)

In some cases, when converting an xapi object into an xo object, the xapi.is_a_snapshot param may be set to false even tho the xapi.snapshotOf refers to another xapi object. Handle this case better.

[wilsonqanda]

@Pilow now that you mention that everything is seen as snapshot i remember all my ISO are seen as snapshots too... so most iso cannot be mounted until i drag them in a folder then put them back in same folder so XO could reprocess all the iso correctly. This might be a related issue to all the snapshots for the VMs.

[Pilow]

@wilsonqanda strangely my ISOs were not impacted, except for the default XCP TOOLS iso that seems gone

the "base copy" regroup bug on homepage of the SR, and the orphan base copies in the DASHBOARD/HEALTH seems to be linked to same problem

we have much fast clones VMs depending on those base copies seen as orphan

we are narrowing on the problem, just need the dev to level it, have faith

[wilsonqanda]

@Pilow Check the iso does it have a snapshot icon in the drive?

Next it show up fine until u want to create a vm in the new vm pg it think they are snapshots and you cant select the ISOs. This is when the problem arise.

Yep the dev here are super reaponsive. Have been a long term user and here and love their support . I mostly just been playing with things and learn so much by tinkering

[Pilow]

@wilsonqanda it's an SMB iso SR on my end, I think this is why it is not impacted

can't snapshot a file on Microsoft SMB share

[wilsonqanda]

@Pilow my is SMB too but it saw the issue. I use windows to login and drag all iso into a new created folder in the iso folder and drag it back out delete the newly created folder and all my iso is back to normal. It was a quick thing.

[Pilow]

@wilsonqanda HAHA. problem resolved instantly for me on last SR I had the troubles

playing with the "rescan all ISO SRs" made the guest tools iso reappear

AND ALL MY OTHER VDIs ON THIS SR/HOST that were invisible

donnnnn't ask me HOW...

but this trick didn't work on other local RAID5 SR that have the problem

[Gheppy]

I have the same problem with 2 servers.
The solution for me was to:

• migrate the VM between servers with different SR or to a 3rd one.
• export-import, virtual machines that are not that important and may have downtime.

[Pilow]

@Gheppy if you can have downtime snapshot/revert works

[limezest]

I'm seeing this same problem in my lab, and I've found a symptom that I'm calling "VDI super-parents"

There is a VDI metadata field called "snapshots:" if your VM has snapshots, this field will be populated with a list of the snapshot VDI UUIDs.
If your VM has no snapshots, this field will be empty.
If your VM is a super-parent, it will have a list of VDI UUIDs that are actually the primary disk for other, unrelated VMs. In this case, XO5 will hide those VDIs from the list of disks because they've been determined to be snapshots.

As far as I can tell, XO5 has two methods for figuring out which VDI are snapshots:

1. It looks at the "is-a-snapshot:" field
2. It builds a map of parents and children from the "snapshot-of:" and "snapshots:" metadata fields (as @pilow found)

XO6 seems to only look at the "is-a-snapshot:" metadata field.

Now here's the question: why is this happening? I only started seeing this behavior after installing XO6 preview.
Maybe it was due to an incomplete backup operation?
I have a bunch of VMs which are clones of a template.
I use iSCSI SRs.

Doing a xe vdi-copy to a new SR collapses the snapshot chain and makes new VDI metadata, and then I did a vbd unplug and destroy, then connected the copied VDI to a new VBD. This fixed the problem for the first test VM that I messed with.

[olivierlambert]

@anthoineb have we found a possible explanation?

[wilsonqanda]

Hello All,

This issue seem to completely go away on a older branch/tag if i build it from below for XO:
xo-lite-v0.17.0

So the issue I would think is coming from the newer version of XO. Not sure best way to test this. So I am going back to an older commit.

If anyone has any suggestion I can try helping as its an annoying issue.

[andrewperry]

We have been quietly suffering without the time to try and resolve it for the past couple of months.

I have now spent the day trying to resolve it in our environment, as we have one SR having this problem with some hundreds! of vdi.

I am having the same problem whether running the docker container version of XO CE or the local install on a VM we've been using for ages.

It seems that the api call from the VDIs tab in v6 (Disks in v5) may be triggering a call to the wrong url, without the /rest/v0 prefix:

sudo journalctl -u xo-server -n 300 --no-pager

2026-02-26T06:01:57.169Z xo:rest-api:error-handler INFO [GET] /vms/[[UUID]]/vdis (404)

I had the same experience as some others for a while, a couple of months ago, where it would not show up in the v5 UI but was showing in v6, but very quickly after that it stopped working in either.

I know that these are VDIs with a snapshot in the chain, for example a parent VDI that may have two snapshots from it.

I had thought the issue may have something to do with https://github.com/vatesfr/xen-orchestra/pull/9381, as it was around this time that I saw the problem in v6 - but I see that this topic was started just before Christmas so there must have been something else too, perhaps that is when it emerged in v5 and this later patch then surfaced it in v6.

If I run curl with the /rest/v0 prefix to the url I don't get the 404.

I hope this helps to track it down!

MathieuRA opened this pull request in vatesfr/xen-orchestra

closed fix(rest-api): fix getVmVdis and enhance the type #9381

[florent]

@andrewperry hi this is an identified issue on xapi / storage side
this commit is making it visible .

Pinging @anthoineb here for more info

[andrewperry]

@florent Thanks, if there is something I need to do to coalesce the vdis to avoid disaster, that would be good to understand. I had been thinking it was just an XO issue that would not affect running vms and their vdis.