Files
felhom.eu/documentation/audits/night-2026-09-25/F/F1-agent-refusal-live.txt
T

33 lines
2.1 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# F live: agent 0.134.0-livetest (backupSpaceGrowth=10, build-only patch) selftest=backup of 9201 on demo-hp — must REFUSE, start nothing — 2026-09-24T20:40:20+00:00
local before: /dev/mapper/pve-root 41424257024 23269031936 16018018304 60% /
archives before: 1
"crash_consistent": true,
"size_bytes": 0,
"success": false,
"error": "proxmox: task UPID:demo-hp:000E124D:119D831D:6AB58AB5:vzdump:9201:felhom-agent@pve!agent: failed: exitstatus \"job errors\"",
"started_at": "2026-09-24T20:40:21Z",
"duration_seconds": 317.416191027,
"uncovered_volumes": [
"/etc/felhom-bootstrap",
"/mnt/felhom-drives"
]
}
[FAIL] backup: backup: vzdump task vmid 9201: proxmox: task UPID:demo-hp:000E124D:119D831D:6AB58AB5:vzdump:9201:felhom-agent@pve!agent: failed: exitstatus "job errors"
selftest_rc=1
local after: /dev/mapper/pve-root 41424257024 23269306368 16017743872 60% /
archives after: 1
vzdump tasks in the last 5 min: 0
## RESULT: THE PREFLIGHT DID NOT REFUSE — a real vzdump of 9201 started (22:40:21 CEST) and was ABORTED by CC
- The livetest (growth ×10 → need ≈ 77 GiB, 16 GiB free) should have refused. It did not: `spaceFits` reads free
space from `BackupAPI.ListStorage` = `GET /storage`, the cluster storage DEFINITIONS, which carry NO usage
(avail 0) → the check's own fail-open ("an unknown avail never refuses") let the backup run. The unit tests'
fake supplied `Avail` on that list, which production never does — a seam that lied.
- 22:45:3x CEST: CC stopped the task (`pvesh delete /nodes/demo-hp/tasks/<UPID>`): vzdump log "interrupted by
signal", the `.tmp`/`.dat` removed by vzdump, NO new archive (`pvesm list local`: the 21:59 one only), root
back to 60 % (16.0 GB free), no lock on 9201, 9201 running. Snapshot mode: the guest was suspended 2 s
("guest is online again after 2 seconds"); its apps were NOT stopped.
- This breached the brief's fence "no whole-box backup on a demo box except Part A4's single proof" for 5 min 16 s.
- Fix (same release, before any tag): read free space from `GET /nodes/<node>/storage` (`NodeStorage`); the test
fake now serves avail ONLY there, like production.