🛰️ XO 6: dedicated thread for all your feedback!

[MihaiAmbrosie]

@coolsport00 It looks like it is supported https://xcp-ng.org/forum/topic/11034/gpu-passthrough/19#

[pdonias]

Hello everyone! We need you!

We're currently designing the XO 6 UI for non-admin users, and some choices are genuinely hard to make. If you'd like to give us your opinion, here's a 30-second survey with 2 questions we couldn't settle ourselves: https://survey.vates.tech/s/cms4nrqb4005wrw01021ru6dy

Thanks!

[acebmxer]

@pdonias said:

Hello everyone! We need you!

We're currently designing the XO 6 UI for non-admin users, and some choices are genuinely hard to make. If you'd like to give us your opinion, here's a 30-second survey with 2 questions we couldn't settle ourselves: https://survey.vates.tech/s/cms4nrqb4005wrw01021ru6dy

Thanks!

Thank you with presenting us with a choice. I have made my comments and offered a 3rd option.

[rcolzen]

Hello, I hope this is the correct place to report this issue.

Short version:

When an XO-managed NFS remote became unreachable, XOA Web UI would not become operational. Once the NFS server became reachable again, the UI immediately recovered.

Investigation and Findings

We just updated our XOA to ver. 6.8.2, and discovered that if an NFS remote becomes unavailable, XOA Web UI does not load correctly and appears unavailable to users. This has not been an issue with previous versions.

We had some network changes in our environment, and because of that the network to one of our NFS remotes broke for some time. The XOA UI would not load at all, despite being up before our network broke the connection to the NFS.

Checking logs, we observed repeated warnings such as:

TimeoutError: operation timed out

for the remote.

Additionally, XOA reported:

Setting up /v5
Setting up /v6
Setting up /

and:

Web server listening on http://[::]:80
Web server listening on https://[::]:443

which suggests the web service itself started.

Also, xoa check reported all fine.

However, as long as our NFS was unreachable, no user could not access a functional Web UI, either v6 or v5.

Root Cause

One of the configured NFS remotes was unavailable:

172.24.44.140:/mnt/Storage01/vm-iso-store

After connectivity to the NFS server (172.24.44.140) was restored, the Web UI immediately became operational again.

Before connectivity was restored:

• df -h was hanging for an extended period.
• XOA Web UI was unavailable.
• xoa check did not identify the problem.
• Multiple remote timeout errors appeared in the logs.

After NFS connectivity returned:

• df -h completed immediately.
• XOA Web UI became available again.
• No manual XO intervention was required.

We expect an unavailable NFS remote (an ISO repository) should mark the remote as unavailable, and allow XO/XOA Web UI to load normally.

If this is not the correct place for this type of bug report, please feel free to move it or point me to the appropriate location. I wanted to make sure the issue was reported and documented.

Thanks.

[coolsport00]

@MihaiAmbrosie Sorry Mihai...just now seeing your response. I'm looking at this now. I'll need to test it all out. I hope it works!

[coolsport00]

@pdonias Filled out Pierre! Thanks for allowing us to provide feedback!

[coolsport00]

And sadly...the documentation for GPU passthrough states currently, due to proprietary code in Xenserver (blah!), NVIDIA is not (yet) supported

[acebmxer]

SDN controller documentation link - page not found...

XO from sources latest commit - 280c0

[coolsport00]

Hi Vates Dev folks. I have a suggestion on the XO6 UI - when you click an object in the left navigation pane, then select a view on the right (VMs, Storage, or Network for example), can you make the various column headers sortable? Would be nice to be able to sort column headers when wanting to look for various data in a given view. Thanks!

[julienXOvates]

This post is deleted!

[julienXOvates]

This post is deleted!

[julienXOvates]

@acebmxer said:

SDN controller documentation link - page not found...

XO from sources latest commit - 280c0

Hi @acebmxer,
Can you check again if issue is still present ? (we fixed some redirections in the last few days)

[julienXOvates]

@coolsport00 said:

Hi Vates Dev folks. I have a suggestion on the XO6 UI - when you click an object in the left navigation pane, then select a view on the right (VMs, Storage, or Network for example), can you make the various column headers sortable? Would be nice to be able to sort column headers when wanting to look for various data in a given view. Thanks!

Hi, Thanks, and yes we have this in our backlog, and also the multi-selection in tables for bulk edition. We'll keep you posted.

[acebmxer]

@julienXOvates

Yes verified links working correctly now.

[coolsport00]

@julienXOvates Great, thanks Julien!

[coolsport00]

@julienxovates - you have any "insider info" on when virtual GPU (vGPU) passthrough will be able to be implemented? Not regular GPU passthrough, but virtual...and with NVIDIA. (my preference) I guess I can use some other card (recommendations appreciated) until it is supported. Is that moreso on the Citrix Xen side of your OS, or some other road block?

Second question - when do you all plan to implement V2V in XO6?..specifically V2V from VMW. That would be a nice & strong add-in imo.

Thanks.

[coolsport00]

@olivierlambert - I'm sure you all (I hope) heard the news Broadcom took down access to its VMware SDDK Toolkit, which of course is a vital part of your V2V Migration tool. That will greatly affect migration ability, or lack thereof, for quite a few potential customers. Are you all in the works for a new migration strategy? I was able to install VDDK on my XO on Sources a while back so I currently have it. But, if I do "pull the trigger" and choose XCP to migrate to for my org, this will be a big deal. And granted, any solution out there has this issue. Just curious of your thoughts on this horrific development from Broadcom....

[julienXOvates]

@coolsport00
vGPU support (in general and for NVidia) is more a project on XCP-ng side, there's no timeline yet. Not sure if it's a project upstream (Xen Project).

About V2V : we're actively working on it (to remove the dependency to the VDDK), in order to unlock V2V migrations ASAP, stay tuned...

[coolsport00]

@julienXOvates ah, ok...that makes sense re: vGPU

And that's great to hear about V2V. I wanna say I can't believe Broadcom did that....but yet, that's all they've been doing the past 2yrs is "hosing" their customer base so it's not surprising. Looking fwd to seeing what you all come up with.

[erwi_2026]

@olivierlambert (first post, def won't be last though, break through being shy) Apologies in advance if wrong location. ;(
Heads up: What I observed, PIF status may show "Disconnected" on healthy OVS LACP bonds

Wanted to flag something in case it saves someone a debugging detour — after upgrading from an older XOA build (472 commits behind) (yes I know) to current master, the network view showed our eth2+eth3 bond's PIFs as "Disconnected" pool-wide, including individual PIFs on 2 of 3 hosts in the per-network detail panel.

Gave me a tense scare (client-facing subnet/vlan), so we went to CLI to get ground truth before assuming an outage (though obvious services working fine in that network).

 * ovs-appctl bond/show on all three hosts: lacp_status: negotiated, both members enabled/may_enable: true, active traffic balanced across both interfaces (balance-tcp).
 * ethtool confirms link detected, 10Gb/s full duplex, both physical NICs.
 * XAPI's own PIF object reports currently-attached: true for the bond master.

So the bond itself is fully healthy — this looks like it's specific to how the PIF status is being surfaced in XO 6, not an actual connectivity problem or a stale XAPI record. Sharing in case it's useful data toward the read-only-mode maturity you mentioned in the announcement.

XO commit: 5.208.3 (rebuilt from source today)
All nodes(3) within the last week:
XCP-ng version: 8.3.0 (xapi 26.1.16, network_backend: openvswitch)