Pod path chased to root cause: MAAS virsh pods incompatible with node-vm volume-ref disks; per-machine power PROVEN working
Dedicated MAAS->libvirt key minted+wired (operator-ruled). Two further
measured findings: (1) MAAS dials the pod from the REGION, not the rack, so
the credential belongs where MAAS dials from; (2) each 503 created a broken
pod server-side outside tofu state (both deleted). Final blocker: domblkinfo
fails on pool+volume disk refs -- reproduced LOCALLY on the rack with an
active pool, so it is libvirtd, not snap/SSH. Making pods work would mean
converting node-vm to file-path disks + re-applying 9 domains. Not needed:
the pod's job was discovery, which PXE already did, and per-machine
power_type=virsh query-power-state returns state=off on the canary. Per
machine power is also the Roosevelt shape (IPMI per node). Awaiting ruling.
CURRENT-STATE untouched pending the ruling.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KiUu1oqt76tWvV4vEC3NAr
1 parent 1616975 commit 0abad6e213a95120cb3dd178df123522e998af56
@JANeumatrix JANeumatrix authored 13 hours ago
Showing 2 changed files
View
docs/audit/stepD-vmhost-20260720.txt
View
docs/changelog-20260719-dc0-deploy-stepB.md