Question around volumeType: disk for UserVolumeConfig

[kaerbr]

I have a bare metal machine on which I want to install talos (single node learning cluster).
Inside the machine there are 2 SSDs. The first one is for all system partitions. The second disk (disk.transport == nvme) shall be used for local path storage. To provision a UserVolume I have this configuration (myStorageVolume.yaml):

apiVersion: v1alpha1
kind: UserVolumeConfig
name: local-path-provisioner
volumeType: disk
provisioning:
  diskSelector:
    match: disk.transport == 'nvme'

1. Create talos config files:
   • talosctl gen config MY_CLUSTER_NAME https://MY_CONTROLPLANE_IP:6443 --install-disk /dev/sda
2. Apply config & patch/add UserVolume:
   • talosctl apply-config --insecure --nodes MY_CONTROLPLANE_IP --file controlplane.yaml --config-patch myStorageVolume.yaml
3. Checking my created volume:
   • talosctl get volumestatus u-local-path-provisioner -o yaml

   metadata:
   namespace: runtime
   type: VolumeStatuses.block.talos.dev
   id: u-local-path-provisioner
   version: 1
   owner: block.VolumeManagerController
   phase: running
   created: 2025-12-28T19:10:31Z
   updated: 2025-12-28T19:10:31Z
   labels:
     talos.dev/user-volume:
   spec:
     phase: failed
     preFailPhase: prepared
     type: disk
     location: /dev/nvme0n1
     mountLocation: /dev/nvme0n1
     uuid: f37521b6-533c-44ae-a757-524a58a53897
     size: 128035676160
     prettySize: 128 GB
     mountSpec:
       targetPath: local-path-provisioner
       selinuxLabel: system_u:object_r:ephemeral_t:s0
       projectQuotaSupport: false
       parentId: /var/mnt
       fileMode: 493
     errorMessage: "filesystem type mismatch: gpt != xfs"

   • talosctl get volumeconfigs u-local-path-provisioner -o yaml

   metadata:
   namespace: runtime
   type: VolumeConfigs.block.talos.dev
   id: u-local-path-provisioner
   version: 2
   owner: block.VolumeConfigController
   phase: running
   created: 2025-12-28T19:10:31Z
   updated: 2025-12-28T19:10:31Z
   labels:
     talos.dev/user-volume:
   finalizers:
     - block.VolumeManagerController
   spec:
     type: disk
     provisioning:
       diskSelector:
         match: disk.transport == 'nvme'
       partitionSpec:
         minSize: 0
         relativeMaxSize: 0
         grow: false
         typeUUID: 0FC63DAF-8483-4772-8E79-3D69D8477DE4
       filesystemSpec:
         type: xfs
     locator:
       diskMatch: disk.transport == 'nvme'
     mount:
       targetPath: local-path-provisioner
       selinuxLabel: system_u:object_r:ephemeral_t:s0
       projectQuotaSupport: false
       parentId: /var/mnt
       fileMode: 493

I do not understand why the Volume is in a failed state. I mean it seems like there is no partition created on my second drive. But the question is why? I thought the new volumeType: disk is explicitly for this case. So to say syntactic sugar for (myStorageVolume2.yaml):

apiVersion: v1alpha1
kind: UserVolumeConfig
name: local-path-provisioner
provisioning:
  diskSelector:
    match: disk.transport == 'nvme'
  minSize: 10 GiB
  grow: true

Running the installation with this config does just work. (:

So bottom line: The two provided yamls: myStorageVolume and myStorageVolume2 are not the same?
Or in other words: did i find a bug or did I just not understand volumeType: disk? :D

[smira]

You can see the actual error in the VolumeStatus resource (or machine logs), but Talos refuses to use this disk, as it has gpt partition table.

If you have two disks, make sure that the disk selector matches exactly the one you want.

Use talosctl get disks to see the properties of your disks.

[kaerbr]

Thanks for the response!

I managed to solve it by deleting everything from that disk, including the partition table, by using sudo wipefs -a /dev/nvme0n1.