XCP-ng
    • Categories
    • Recent
    • Tags
    • Popular
    • Users
    • Groups
    • Register
    • Login

    🛰️ XO 6: dedicated thread for all your feedback!

    Scheduled Pinned Locked Moved Xen Orchestra
    258 Posts 53 Posters 115.9k Views 43 Watching
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • poddingueP Online
      poddingue Vates 🪐
      last edited by poddingue

      Depends which one you mean, because there are two different problems tangled together in this stretch of the thread and they have opposite answers.

      If you mean @escape222's report at #177, where a VM cloned from XO 6 sits at the TianoCore screen for a couple of minutes, then yes, that was an XO bug and it is fixed. PR #9867, merged 26 May, shipped in XO 6.5.0 on 28 May. The boot order was being rewritten whenever no new disk needed provisioning, so an HVM VM created from a template that already had a disk got network pushed to the front whether or not anyone asked for a network install, and the VM burned the PXE timeout before falling through to the disk. It now follows the install method only. Since you are on sources, anything past 6.5.0 has it.

      If you mean the slow UEFI boot @MajorP93 described at #182, with the installing Xen timer and spinlock lines, that one is not an XO bug and no XO patch will touch it. It is a regression in the Linux guest kernel introduced in 6.12.5. The cost lands per secondary vCPU, so the wider the VM, the worse it looks. There is a separate thread with the per-vCPU numbers: Ubuntu cloud images on XCP-ng 8.3 UEFI. Worth noting its title blames console=ttyS0, which we now think amplifies the same bug rather than being a second one.

      I measured that one here this week on a single host, changing only the guest kernel between runs and leaving everything else alone. Ubuntu 7.0.0-30 came in at 50 and 59 seconds across two runs. 7.0.0-31 came in at 0.4. Wall clock reboot to sshd went from 78 seconds to 31.

      The awkward part is the timing. The upstream fix is f24df84cbe05, in stable 6.12.97 and later, 6.18.y, 7.1.4 and later, and 7.2, but no default channel carries it yet. I re-checked the archives this evening: Debian trixie still ships 6.12.94-1, with 6.12.100-1 sitting in proposed-updates for the next point release, and Ubuntu 26.04 still ships 7.0.0-30 in updates, published today, while 7.0.0-31 has been in proposed since 10 August. Rocky and el10 I could not confirm either way. So keep whatever workaround you are on until a named version lands for your distro.

      There is arguably a third one at #183, where @Greg_E had Debian 13 and Windows Server 2022 refusing to boot at all when created through XO-lite with UEFI. As far as I know nobody has retested that since.

      If it is the kernel one you are hitting, this says which version you are waiting for:

      uname -r
      dmesg -T | grep -iE "installing Xen timer|spinlock event"
      
      C 1 Reply Last reply Reply Quote 0
      • C Offline
        coolsport00 @poddingue
        last edited by coolsport00

        @poddingue said:
        Depends which one you mean

        Thanks for the update! I meant the one where the VM was hanging because of incorrect boot order; but, good to know about the other issues. I'm moderately ok on Linux, but not so much so to where I fully understand the long boot time you go into (& divinni in the other thread) 😛

        I had an issue in XO-Lite myself last yr. That sounds a bit like I had; or, I believe the issue was BIOS vs UEFI. If I created the VM with BIOS, I could boot/set it up...but not able to with UEFI.

        Thanks!

        poddingueP 1 Reply Last reply Reply Quote 1
        • poddingueP Online
          poddingue Vates 🪐 @coolsport00
          last edited by

          On the XO Lite one, there's an issue open that sounds like what you're describing, https://github.com/vatesfr/xen-orchestra/issues/9801, filed off @Greg_E's report in this thread.
          It says UEFI VMs created through XO Lite don't boot since the LTS release, with Debian 13 and Windows Server 2022 both failing, while VMs built under 8.2 with UEFI still work.

          Your timing may not line up, since you're saying last year and that issue is scoped to since the LTS release, so I'd rather ask than assume.
          If it does look like yours, a comment on it would carry more than anything I can add, because it hasn't had one since May and I don't think anyone has picked it up. 🤷

          G 1 Reply Last reply Reply Quote 0
          • G Offline
            Greg_E @poddingue
            last edited by

            @poddingue I haven't had time to get back and check this, my lab hasn't even had power on it for about 4 months.

            1 Reply Last reply Reply Quote 1
            • C Offline
              coolsport00
              last edited by

              @poddingue maybe it was this yr? Heck...the days & months just run together on me anymore 😛 I'll check that out. Appreciate it!

              poddingueP 1 Reply Last reply Reply Quote 1
              • poddingueP Online
                poddingue Vates 🪐 @coolsport00
                last edited by

                Thanks!

                1 Reply Last reply Reply Quote 0
                • C Offline
                  coolsport00
                  last edited by

                  Maybe I missed this somewhere...but in XO6 > Quick Actions, is it planned to add a Migration option there? I think that is a needed high action item. I don't see Migrate anywhere in XO6 (yet). BTW, I'm going through all the XO 6.x updates posts. I've not yet finished reading through them, so apologies if it was mentioned in one that I haven't read yet 😛

                  julienXOvatesJ 1 Reply Last reply Reply Quote 1
                  • julienXOvatesJ Offline
                    julienXOvates Vates 🪐 XO Team @coolsport00
                    last edited by

                    @coolsport00 you mean migrate VM ? it's not yet there in XO6, but we'll bring it soon (hopefully in the next 2 months).

                    C 1 Reply Last reply Reply Quote 0
                    • C Offline
                      coolsport00 @julienXOvates
                      last edited by

                      @julienXOvates Ok, great. Sounds good...thank you Julien!

                      Also, do you all have GPU passthrough yet? I'm really needing that function for a VDI setup i have. Thanks!

                      M 1 Reply Last reply Reply Quote 0
                      • M Offline
                        MihaiAmbrosie @coolsport00
                        last edited by

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

                        1 Reply Last reply Reply Quote 0

                        Hello! It looks like you're interested in this conversation, but you don't have an account yet.

                        Getting fed up of having to scroll through the same posts each visit? When you register for an account, you'll always come back to exactly where you were before, and choose to be notified of new replies (either via email, or push notification). You'll also be able to save bookmarks and upvote posts to show your appreciation to other community members.

                        With your input, this post could be even better 💗

                        Register Login
                        • First post
                          Last post