fix(ci1): revert ISO to Filesystem PVC; CDI v1.65.0 block-upload pod blocked by capability drop

The Block-mode DataVolume migration (commit 0bf47df) hit a CDI v1.65.0 limitation:
the upload-target pod runs as uid 107 with `capabilities.drop: [ALL]`, so it
cannot open the underlying block device:

  blockdev: cannot open /dev/cdi-block-volume: Permission denied
  Saving stream failed: Unable to transfer source data to target file:
  error determining if block device exists: exit status 1

Reverting to a Filesystem-mode PVC + virtctl image-upload pvc, which DID work
(uploaded the 7.7 GiB ISO with valid ISO9660 magic intact). Boot timeout is
unresolved (header docstring captures the open issue + 3 paths to revisit).

The bootOrder swap (1c4145a) and runStrategy migration (87a7d7c) stay landed —
those are correct improvements regardless of the volume-mode question.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Codex
2026-05-08 14:32:52 -05:00
parent 0bf47dfa33
commit 9f6dc1a9d5

View File

@@ -49,58 +49,50 @@ metadata:
pod-security.kubernetes.io/enforce: privileged pod-security.kubernetes.io/enforce: privileged
--- ---
# ISO DataVolume — CDI manages an underlying PVC of the same name and exposes # ISO PVC — populated via CDI virtctl image-upload (CDI is now installed).
# the upload-target pod once it's ready.
# #
# **Why DataVolume + Block volumeMode** (vs the original `kind: PersistentVolumeClaim` # **Volume mode (2026-05-08 status):** Filesystem-mode PVC. A migration to
# + virtctl image-upload pvc): a `volumeMode: Filesystem` PVC stores the upload # `volumeMode: Block` via DataVolume was attempted to address an OVMF SATA
# as `/disk.img` on a mounted ext4. KubeVirt then exposes that file as a SATA # CDROM read timeout, but CDI v1.65.0's upload-target pod runs as uid 107
# CDROM via QEMU. On 2026-05-08 this caused the OVMF UEFI firmware to fail # with `capabilities.drop: [ALL]` and cannot open the underlying block
# Boot0001 with "Time out" reading the SATA CDROM, even with the install ISO # device (`blockdev: cannot open /dev/cdi-block-volume: Permission denied`).
# at bootOrder:1 — see docs/infrastructure/feedback notes below. The ISO # Reverted to Filesystem PVC pending one of:
# content WAS valid (`file` reported "ISO 9660 CD-ROM filesystem data ... # - CDI deployment override granting CAP_SYS_RAWIO to upload pod
# (bootable)"), but the QEMU SATA emulation over a Filesystem-PVC backing was # - Pre-populated PVC via privileged init pod that dd's the ISO directly
# too slow / mis-attached for OVMF's CDROM read window. # - Migration to a different storage class that exposes block devices
# differently (e.g. iSCSI, where Longhorn's CSI mount path may behave
# differently)
# #
# `volumeMode: Block` gives us a raw block device directly — KubeVirt attaches # Population workflow (this PVC, Filesystem mode):
# it to the VM as `/dev/sdX` style storage, OVMF reads ISO9660 sectors directly # 1. virtctl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml image-upload pvc \
# from the underlying block volume, no QEMU virtual file emulation needed. # windows-server-2025-iso -n kubevirt-vms \
# This is the recommended pattern for ISO install media on KubeVirt + Longhorn. # --image-path "$env:USERPROFILE\Downloads\en-us_windows_server_2025_updated_march_2026_x64_dvd_8e06425a.iso" \
# --size 10Gi --storage-class longhorn --access-mode ReadWriteOnce \
# --uploadproxy-url https://localhost:8443 --insecure
# (--uploadproxy-url uses port-forward in practice: `kubectl port-forward
# -n cdi service/cdi-uploadproxy 8443:443 &` first.)
# #
# Population workflow: # **Open boot issue:** even with the ISO at bootOrder:1, OVMF console showed:
# 1. After this DataVolume is applied, CDI creates the PVC and an # BdsDxe: starting Boot0001 "UEFI QEMU DVD-ROM QM00001 " from ... Sata(...)
# upload-target pod. Wait for `phase: UploadReady`. # BdsDxe: failed to start Boot0001 ... Time out
# 2. From BLUEJAY-WS: # Diagnosis confirmed PVC content IS a valid bootable ISO9660 image — the
# kubectl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml port-forward \ # timeout is in OVMF reading from the SATA-CDROM-backed-by-filesystem-PVC.
# -n cdi service/cdi-uploadproxy 8443:443 & # Block mode would likely fix it; see CDI permission issue above.
# virtctl --kubeconfig $env:USERPROFILE\.kube\rke2.yaml image-upload dv \ apiVersion: v1
# windows-server-2025-iso -n kubevirt-vms \ kind: PersistentVolumeClaim
# --image-path "$env:USERPROFILE\Downloads\en-us_windows_server_2025_updated_march_2026_x64_dvd_8e06425a.iso" \
# --uploadproxy-url https://localhost:8443 --insecure --no-create
# (`--no-create` — the DV/PVC already exist, virtctl just streams bytes.)
apiVersion: cdi.kubevirt.io/v1beta1
kind: DataVolume
metadata: metadata:
name: windows-server-2025-iso name: windows-server-2025-iso
namespace: kubevirt-vms namespace: kubevirt-vms
labels: labels:
app: ci-runner app: ci-runner
flowercore.io/managed-by: bluejay-infra flowercore.io/managed-by: bluejay-infra
annotations:
# Tell CDI not to "convert" — keep raw bytes so the underlying block device
# IS the ISO9660 sectors verbatim, not a QCOW2 wrap.
cdi.kubevirt.io/storage.contentType: kubevirt
spec: spec:
source: accessModes:
upload: {} - ReadWriteOnce # Bump to ReadOnlyMany after population for multi-VM use
pvc: resources:
accessModes: requests:
- ReadWriteOnce # Bump to ReadOnlyMany after population for multi-VM use storage: 10Gi # Server 2025 ISO is 7.7GB; 10Gi for headroom
resources: storageClassName: longhorn
requests:
storage: 10Gi # Server 2025 ISO is 7.7GB; 10Gi for headroom
volumeMode: Block # CRITICAL — see header comment above
storageClassName: longhorn
--- ---
# Root disk PVC — empty 200Gi volume that Windows installs into. # Root disk PVC — empty 200Gi volume that Windows installs into.
@@ -410,11 +402,8 @@ spec:
persistentVolumeClaim: persistentVolumeClaim:
claimName: ci1-rootdisk claimName: ci1-rootdisk
- name: windows-iso - name: windows-iso
# Reference the DataVolume (defined above) — CDI creates the PVC of persistentVolumeClaim:
# the same name with volumeMode: Block. The VMI controller blocks claimName: windows-server-2025-iso
# VM start until DV phase is Succeeded (i.e. upload completed).
dataVolume:
name: windows-server-2025-iso
- name: virtio-drivers - name: virtio-drivers
containerDisk: containerDisk:
# Pinned to v1.8.2 (latest stable as of 2026-05-08). # Pinned to v1.8.2 (latest stable as of 2026-05-08).