Vates

People working at Vates

Posts

  • RE: Missing vdi which is a base for a delta

    @blueh2o I think the two things you saw fit together, at least going by what I see in the mirror code. The 1 differencing tag only says that this one backup's disk is a delta, which is normal for an incremental; the key sits further back in the chain, which I think is why the health check could still rebuild the whole ~280 GiB disk.
    And as I wrote a few posts up, the mirror reads the parent of each delta on Ark and then asks each target whether it already has it, so last night's error would mean the Synology is missing d08f1acc for that one VM (fb7fca7d), not that Ark is broken.
    That brings us back to Florent's suggestion from the older thread, deleting that VM's most recent backups on the Synology so they get sent again, though I can't tell you why the Synology lost that base in the first place. At least it's down to one VM out of 56 last night, against 16 in October.

  • RE: Backups in OX v6

    @flakpyro from your screenshot and a look at the code, I think the VM name is already stored on each of those "backup VM" tasks, but the XO 6 task list just shows the generic label instead, so it might be a small change rather than a big missing piece (I haven't checked whether the side panel shows it).
    It could be worth its own post on feedback.vates.tech so it gets votes.
    The closest one I found is about the job's detail panel rather than the task list: https://feedback.vates.tech/posts/60/xo6-backup-summary-is-a-stacked-jumbled-and-cryptic

  • RE: iscsi shows "No Storage Repositories Detected"

    From your discovery output, the target still seems to be there, which is a good start. The docs have a section for the SR part: https://docs.xcp-ng.org/storage/manage-srs#reattach-a-forgotten-sr (with xe sr-probe just below it to find the IQN and LUN, though I haven't tried it on iSCSI myself).
    Before you go further, though, there's something Olivier explained in a similar thread: the VM definitions (name, CPU, RAM, which disks) live in the host's database, not on the SR, so after a reinstall the disks can come back, but the VMs as they were only return with a metadata backup; otherwise it means recreating them and attaching the disks (https://xcp-ng.org/forum/post/93504).
    Did you have an XCP-ng metadata backup job in XO, or one from the host console? Also, from what I read in the LVM storage driver, creating a new SR on that LUN would wipe it, so I'd be careful not to click through anything that offers to create one. Your post got cut off at "Other things d", could you tell us what else you tried?
    @Team-Storage, could someone check I haven't steered okynnor wrong here?

  • RE: [XO6] Bug: Working in multiple XO6 in tabs, breaks the "eye"-sidebar

    I tried your steps on a lab XOA running 6.10, and I could reproduce it, but only once I had unpinned the side panel. With the panel pinned (the pin button at the top of the panel), switching tabs left it alone in my test, so that might do as a workaround for now.
    Did you unpin it at some point, or is your browser window on the narrow side? From a quick look at the code, I think the open or closed state of that panel is kept in the browser's local storage, which all tabs of the same site share, so one tab closing it closes it for the others, though I may be misreading it. You have the steps written down already, so I'd pass them to @Team-XO-Frontend; whether it's meant to work that way is their call, not mine.

  • RE: Large QCOW2 VDI on LVM SR fails to activate with make_chain_rw / Input/output error

    Hello, the current update candidate includes fixes related to this issue. Note that it is still under testing.

  • RE: XCP-ng 8.3 updates announcements and testing

    @Andrew said:

    RPU mostly worked... Every host was updated. All VMs moved and returned. Waiting before re-enabling the load balancer task is stuck waiting...

    @Andrew, xapi 26.1.23-2.1.xcpng8.3 includes fixes and improvements aimed at the RPU issues.

  • RE: packer plugin signing key

    @enkaskal I need to correct my reply above, I got two things wrong. The plugin is ours after all: ddelnano/packer-plugin-xenserver is the repo's old address, it moved to vatesfr/packer-plugin-xenserver (where you filed #212), and GitHub just redirects the old link, which I read on our IaC docs page and jumped to the wrong conclusion.
    The key isn't missing either, I just didn't look in the right place: it's the signing key the Terraform Registry lists for vatesfr/xenorchestra, so you can fetch it from there. 🤦

    curl -s https://registry.terraform.io/v1/providers/vatesfr/xenorchestra/0.41.2/download/linux/amd64 \
      | jq -r '.signing_keys.gpg_public_keys[0].ascii_armor' > xenorchestra-signing.asc
    gpg --import xenorchestra-signing.asc
    gpg --verify packer-plugin-xenserver_v0.12.0_SHA256SUMS.sig packer-plugin-xenserver_v0.12.0_SHA256SUMS
    

    On my machine the last command reports a good signature from 0DD2E92A7A0F682BC29F87073084D82948625D89, still a key in Dom Delnano's name from before the move. I'm honestly not sure how much trust that gives you beyond "the registry vouches for the same key", so I'd leave that part to the folks on #212 who know the release side better than me. Sorry for the detour. 🤷

  • RE: XCP-ng 8.3 updates announcements and testing

    New maintenance update candidates for XCP-ng 8.3 LTS

    This batch of updates contains security fixes, new Windows guest tools, and other improvements. It also changes some defaults of the SSH server: read the OpenSSH section before you update.

    What changed

    Virtualization

    • guest-templates-json:
      • Sync with XenServer guest-templates-json-2.0.17-1.
      • New guest template: Ubuntu Resolute Raccoon 26.04.
      • The Rocky Linux 10 template does not change (15 GB default disk).
    • swtpm: This release contains a fix for CVE-2026-75900, an out-of-bounds read in the swtpm daemon that could cause a crash of the daemon.
    • xcp-ng-pv-tools:
      • Update to XCP-ng Windows Guest Tools 9.2.445.
      • Remove the XenTools-fix-9.2.351.msp hotfix tool for users of 9.2.350.
    • xen: This release is a synchronization with XenServer release 4.17.7-6. It contains the patches for XSA-524 and XSA-525, and a fix for lost timer interrupts. It also fixes a bug in the original XSA-524 fix where the IOMMU device table size on AMD systems was always set to 2 MiB.

    System

    • gnutls: Update to gnutls 3.8.12. This update fixes multiple vulnerabilities, including CVE-2026-3833.
    • libssh2: Update libssh2 to fix five security vulnerabilities. They affect the SFTP/SCP connections that curl and Midnight Commander (mc) make:
      • CVE-2026-58051: potential arbitrary free when listing public keys
      • CVE-2026-66032: use-after-free in SFTP
      • CVE-2026-66033: out-of-bounds read/write with the AES-GCM cipher
      • CVE-2026-66034: out-of-bounds read when listing public keys
      • CVE-2026-66035: heap buffer overflow during ETM (Encrypt-then-MAC) decryption
    • nettle: Update to nettle 3.10. Sync with XenServer nettle-3.10-3, which includes a fix that XCP-ng released previously.
    • openssh: Update to OpenSSH 9.9p1. This update fixes 21 CVEs: CVE-2026-3497, CVE-2026-35386, CVE-2026-35387, CVE-2026-55653, CVE-2026-55654, CVE-2026-55655, CVE-2026-59995, CVE-2026-59996, CVE-2026-59997, CVE-2026-59999, CVE-2026-60000, CVE-2026-60001, CVE-2026-60002, CVE-2026-73281, CVE-2026-73282, CVE-2026-73283, CVE-2026-106552, CVE-2026-106553, CVE-2026-106582, CVE-2026-106583, CVE-2026-106585.
      • Post-Quantum Key Exchange is now available. It is the preferred key exchange by default.
      • FIDO2 keys are now available, but they are disabled by default. See Enabling FIDO2 keys.
      • The SSH client now reads drop-in files in /etc/ssh/ssh_config.d/.
      • Fix a temporary lockout with ssh-copy-id.
      • We hardened the SSH server configuration with our security team. Some changes modify how existing configurations work:
        • sshd refuses SSH user certificates by default. If you log in with certificates, follow SSH user certificates before the update, or you will lose SSH access.
        • Forwarding is limited to the host itself (VNC console and XO Lite/XAPI tunnels continue to work). See Forwarding.
        • MaxAuthTries is now 3, LoginGraceTime is now 60 seconds, and PerSourcePenalties (brute-force protection) is enabled. See What you can safely customize.
        • sshd no longer accepts hmac-sha1 and ssh-rsa (SHA-1). Very old SSH clients can fail to connect.
        • The update restarts sshd and replaces /etc/ssh/sshd_config. Put your changes in drop-in files: see Do not edit our configuration files.
    • openssl: Update to openssl 3.5.7. This is a security update that fixes multiple CVEs, including one rated High (CVE-2026-45447, heap use-after-free in PKCS7_verify()). It also contains a fix for CVE-2026-72897. See the openssl vulnerabilities page for more information.
    • samba: Sync with XenServer: update samba to 4.23.3, to support LDAPS and the Windows security hardening.
    • xcp-ng-deps: Remove ldns and unbound-libs as they are not used (this avoids the frequent security updates of these packages).
    • New and updated dependencies of gnutls 3.8, samba 4.23 and OpenSSH FIDO2 support: libcbor, libfido2, libidn2, libtalloc, libtdb, libtevent. Rebuilds with the updated gnutls: wget, and the optional neon, nut and nbdkit.

    Control Plane

    • xapi:
      • Fix an RPU that could stop in some cases, because xapi lost track of an event.
      • Improve performance for storage migrations with snapshots, with a fix of the migration order.
      • Support adding SSL certificates signed with SHA384.
      • Fix SM plugins that xapi registered twice. This broke SR functions.
      • Fix migration of VDIs with tags to SRs that are not plugged on the source host.
      • Fix incomplete VDI cleanup when HA is disabled while a host is fenced.
      • Fix a host that could stop when it joins the pool, because of a dangling reference.
      • sdn-controller: Add reliability fixes, make sure that xapi applies the rules.
      • This update contains many storage-related changes. We tested each change alone, but not all together. Please test your storage operations.
    • xcp-python-libs: Sync with XenServer xcp-python-libs-3.0.11-1.
    • xha: Fix a potential memory access crash during the activation step.

    Storage

    • blktap: This release contains 2 performance improvements and 2 important fixes:
      • The NBD server of tapdisk does fewer memory allocations. This improves VDI migrations.
      • The polling algorithm of tapdisk sends fewer notifications. This improves the performance in polling mode.
      • Fix a possible VDI inconsistency with Windows VMs, when the Windows driver reconnects a stalled ring.
      • Fix a possible local data corruption during a storage migration with CBT enabled.
    • nbd: Update to nbd 3.25 (sync with XenServer).
    • sm:
      • Fix LINSTOR VDI locking failing during a transient state, which could prevent a user from executing a VDI command.
      • LINSTOR SR scans are now ~35% faster (e.g. ~1m08s down to 44s on a host with 950 VDIs in our tests).
      • Fix diskless LINSTOR resources being cleaned up even though they are still in use.
      • Fix an issue on LVM SRs that could block VDI activation of QCOW2 VDIs on a non-coordinator host.
      • Fix incremental backup failures (SR_BACKEND_FAILURE_460) with CBT on file-based SRs.
      • Clean up gc_running_file on GC exception so that operations on the SR can continue after the GC error.

    Network

    Optional packages

    • libreswan: Remove DNSSEC support. This allows us to remove ldns and avoid the frequent security updates of this package.

    UI

    • xo-lite:
      • [New network] Choose between no NBD, secure NBD and insecure NBD (not encrypted) when creating a network (PR #10540)
      • [VM/System] Fix video RAM displayed in bytes instead of MiB (PR #10486)
      • [Treeview] Add a search bar and virtualize the tree (PR #10414)
      • [New/VM] Hide guest tools ISO SR in new VM ISO selector (PR #10430)
      • [Host] Add possibility to enable/disable a host (PR #10337)
      • [Host] Add possibility to restart a host toolstack (PR #10480)
      • [Host] Add possibility to detach a host (PR #10520)
      • [Side panels] Harmonize card titles across side panels (PR #10289)
      • [VM/New] Fix UEFI VMs booting in BIOS mode when created with a network install or without disk (PR #10521)
      • [Treeview] Press Ctrl+K (Cmd+K on macOS) to focus the treeview search (PR #10492)
      • [Pool] Fix connection, stats and consoles when overriding the pool master with ?master=: use HTTPS by default instead of the protocol and port of the page (PR #9955)
      • [Pool/Host/VM] Add copy button for multiple information on Host/VM Dashboard and Pool/Host/VM System tabs (PR #9976)
      • [New/VM] Fix MAP_DUPLICATE_KEY error that could randomly prevent the VM creation (PR #10552)
      • [Host/Network] Add the possibility to forget a PIF (PR #10239)
      • [Lite] Display dates and times as YYYY-MM-DD HH:mm:ss (PR #10547)
      • [Pool/Hosts] Add a side panel displaying the selected host's details, and a search bar to filter the hosts table (PR #10580)
      • [VM/Host] Add console clipboard management (PR #10625)
      • [VM/Network] Add the possibility to edit a VIF (PR #10505)

    Drivers and Firmware

    • amd-microcode: Update to the 2026-10-01 AMD microcode drop as redistributed by XenServer. Updated CPUs:
      • GN-B1 00a00f11: 2025-04-18, rev 0a0011de -> 2026-06-19, rev 0a0011e1
      • GN-B2 00a00f12: 2025-03-27, rev 0a001247 -> 2026-06-19, rev 0a00124e
      • RS-B1 00a10f11: 2025-06-09, rev 0a101158 -> 2026-07-06, rev 0a10115b
      • RS-B2 00a10f12: 2025-06-12, rev 0a101253 -> 2026-07-01, rev 0a10125b
      • RSDN-A2 00aa0f02: 2025-06-12, rev 0aa0021c -> 2026-07-06, rev 0aa0021f
    • cisco-enic: Update to 4.7.0.6 (sync with XenServer).
      • Add SR-IOV v2 support with PF/VF mailbox communication and lifecycle management.
      • Add administrative queue support and improve queue/IRQ handling.
      • Improve PF/VF link-state, MTU, shutdown, and SR-IOV teardown handling.
      • This is a large update of the driver. If you have Cisco hardware that uses enic, please test it.
    • intel-microcode: This update installs the Intel IPU 2026.4 microcode from the XenServer 20260929-1 release.
    • mellanox-mlnxen: Fix ConnectX-4 Ethernet and RDMA devices that the driver did not register correctly.

    Optional packages

    • r8125-module-alt: New alternative driver (Realtek 9.018.00). Various bug fixes for 2.5G Realtek NICs. It does not support the 5G devices: use the main driver for them.

    Versions

    • amd-microcode: 20260923-1.1.xcpng8.3 -> 20261001-2.1.xcpng8.3
    • blktap: 3.55.5-12.3.xcpng8.3 -> 3.55.5-12.4.xcpng8.3
    • cisco-enic: 4.5.0.7-1.xcpng8.3 -> 4.7.0.6-1.1.xcpng8.3
    • gnutls: 3.3.29-10.2.xcpng8.3 -> 3.8.12-1.1.xcpng8.3
    • gpumon: 24.1.0-99.1.xcpng8.3 -> 24.1.0-100.1.xcpng8.3
    • guest-templates-json: 2.0.16-1.1.xcpng8.3 -> 2.0.17-1.1.xcpng8.3
    • intel-microcode: 20260925-1.xcpng8.3 -> 20260929-1.xcpng8.3
    • libcbor: 0.11.0-3.1.xcpng8.3
    • libfido2: 1.14.0-7.1.xcpng8.3
    • libidn2: 2.3.4-4.xcpng8.3
    • libssh2: 1.11.0-1.xcpng8.3 -> 1.11.0-1.2.xcpng8.3
    • libtalloc: 2.1.16-1.el7 -> 2.4.3-1.1.xcpng8.3
    • libtdb: 1.3.18-1.el7 -> 1.4.14-1.1.xcpng8.3
    • libtevent: 0.9.39-1.el7 -> 0.17.1-1.xcpng8.3
    • mellanox-mlnxen: 5.9_0.5.5.0-3.1.xcpng8.3 -> 5.9_0.5.5.0-3.2.xcpng8.3
    • nbd: 3.24-1.xcpng8.3 -> 3.25-2.1.xcpng8.3
    • nettle: 2.7.1-9.el7_9 -> 3.10-3.1.1.xcpng8.3
    • openssh: 9.8p1-1.2.6.xcpng8.3 -> 9.9p1-33.2.xcpng8.3
    • openssl: 3.5.5-1.3.xcpng8.3 -> 3.5.7-1.1.xcpng8.3
    • rpm: 4.11.3-32.el7 -> 4.11.3-32.1.xcpng8.3
    • samba: 4.10.16-25.3.xcpng8.3 -> 4.23.3-1.1.xcpng8.3
    • sm: 3.2.12-25.2.xcpng8.3 -> 3.2.12-25.3.xcpng8.3
    • swtpm: 0.7.3-12.1.xcpng8.3 -> 0.7.5-1.xcpng8.3
    • wget: 1.21.4-1.1.xcpng8.3 -> 1.21.4-1.2.xcpng8.3
    • xapi: 26.1.21-1.4.xcpng8.3 -> 26.1.23-2.1.xcpng8.3
    • xcp-featured: 1.2.1-6.xcpng8.3 -> 1.2.1-7.xcpng8.3
    • xcp-ng-deps: 8.3-14 -> 8.3-16
    • xcp-ng-pv-tools: 8.3-19.xcpng8.3 -> 8.3-21.xcpng8.3
    • xcp-python-libs: 3.0.10-1.1.xcpng8.3 -> 3.0.11-1.1.xcpng8.3
    • xen: 4.17.7-4.2.xcpng8.3 -> 4.17.7-6.2.xcpng8.3
    • xha: 25.2.0-1.1.xcpng8.3 -> 25.2.0-1.2.xcpng8.3
    • xo-lite: 0.26.0-1.xcpng8.3 -> 0.27.0-1.xcpng8.3

    Optional packages

    • fakechroot: 2.20.1-2.el7
    • http-nbd-transfer: 1.7.0-1.xcpng8.3 -> 1.8.0-1.xcpng8.3
    • libreswan: 4.12-2.3.3.xcpng8.3 -> 4.12-2.3.4.xcpng8.3
    • nbdkit: 1.8.0-4.el7 -> 1.8.0-4.1.xcpng8.3
    • neon: 0.30.0-3.el7 -> 0.30.2-6.1.xcpng8.3
    • nut: 2.8.0-2.1.xcpng8.3 -> 2.8.0-2.2.xcpng8.3
    • python3-py: 1.4.32-2.el7
    • python3-pytest: 2.9.2-5.el7
    • r8125-module-alt: 9.018.00-1.xcpng8.3

    Test on XCP-ng 8.3

    Warning: if you log in to your hosts with SSH user certificates, or if you edited /etc/ssh/sshd_config, read the OpenSSH section before the update.

    yum clean metadata --enablerepo=xcp-ng-testing,xcp-ng-candidates
    yum update --enablerepo=xcp-ng-testing,xcp-ng-candidates
    reboot # required: new Xen hypervisor, microcode, and drivers
    

    The usual update rules apply: pool coordinator first, etc.

    What to test

    As usual, normal use and anything else you want to test. These checks help us most:

    1. SSH: deploy a public key with ssh-copy-id, then connect with the key. It must work. If you use SSH tunnels (VNC console, XO Lite) or FIDO2 keys, test them too.
    2. Storage: enable CBT on the disks of a running VM, migrate the VM with its storage, then run an incremental backup. Also do your usual backups, restores and snapshots. They must complete, and the data must be correct.
    3. If you have Cisco hardware that uses the enic driver, make sure that the NICs have link and pass traffic. If you use SR-IOV on them, test it too.
    4. If you use XOSTOR, do SR scans and VDI operations. They must complete without errors.

    If a check fails, tell us the hardware, the VM type, and provide the logs (/var/log/xensource.log, /var/log/SMlog, /var/log/secure).

    Test window before official release of the updates

    ~6 days

    We would like to thank users who shared feedback since our last call for testing:
    @acebmxer, @Andrew, @XCP-ng-JustGreat, @JeffBerntsen, @hoh, @flakpyro, @robertblissitt, @Greg_E

Member List

olivierlambertO olivierlambert Group Owner
22,523 Posts 3,413 Reputation
Marc.pezinM Marc.pezin Group Owner
63 Posts 84 Reputation
DanpD Danp Group Owner
2,969 Posts 452 Reputation
julien-fJ julien-f Group Owner
744 Posts 280 Reputation
stormiS stormi Group Owner
3,239 Posts 1,247 Reputation
gduperreyG gduperrey Group Owner
200 Posts 213 Reputation
poddingueP poddingue Group Owner
471 Posts 95 Reputation
O OlivierFL Group Owner
1 Posts 2 Reputation
MarieGardeM MarieGarde Group Owner
1 Posts 2 Reputation
lmolinaL lmolina Group Owner
11 Posts 16 Reputation
C CesaireMH Group Owner
2 Posts 0 Reputation
JeremyLARDENOISJ JeremyLARDENOIS Group Owner
3 Posts 1 Reputation
S sylverea Group Owner
0 Posts 0 Reputation
ArnaudGFA ArnaudGF Group Owner
0 Posts 0 Reputation
asollierA asollier Group Owner
0 Posts 0 Reputation
All-KiA All-Ki Group Owner
2 Posts 1 Reputation
LucienLassalleL LucienLassalle Group Owner
11 Posts 12 Reputation
eveaneE eveane Group Owner
3 Posts 0 Reputation
M MrQ Group Owner
5 Posts 4 Reputation
W warren.postdam Group Owner
0 Posts 0 Reputation