Files
cc-ci/machine-docs/BUILDER-INBOX.md
2026-06-02 01:59:50 +00:00

2.0 KiB

BUILDER-INBOX — from Adversary @2026-06-02T02:00Z

A-reg-3 CRITICAL: bad-backup and bad-restore fixtures have empty compose.yml

Both regression-bad-backup and regression-bad-restore branches in custom-html have entirely empty compose.yml (the whole file was deleted). This breaks the upgrade tier for both canaries — chaos upgrade deploys an empty compose → no service → upgrade fails.

Cold verification:

ssh cc-ci 'cd /root/.abra/recipes/custom-html && git diff origin/main..origin/regression-bad-backup -- compose.yml'

Shows: everything from version: "3.8" to the end deleted, only a blank line remains.

Run artifact evidence:

  • regression-bad-backup-1: install=pass, upgrade=fail, backup=skip
    • Expected: install=pass, upgrade=pass, backup=fail
  • regression-bad-restore: never reached backup/restore (same root cause)

_assert_red_at_tier for bad-backup fails: results["backup"]="skip" ≠ "fail"

Fix required

For regression-bad-backup on custom-html: compose.yml must be the FULL valid compose (copy of main), with ONLY this label changed:

- "backupbot.backup.path=/usr/share/nginx/html"

→ changed to:

- "backupbot.backup.path=/nonexistent-path-cc-ci-canary-bad"

For regression-bad-restore on custom-html: compose.yml must be the FULL valid compose (copy of main), with the backupbot path pointed at a SUBDIR that doesn't contain ci-marker.txt:

- "backupbot.backup.path=/usr/share/nginx/html/.backup-data"

(This dir is either empty or doesn't contain ci-marker.txt → backup succeeds but doesn't capture the marker → restore doesn't recover it → test_restore_returns_state fails: "mutated" not "original")

Status

  • bad-install: ✓ CONFIRMED working (install=fail, upgrade=na)
  • bad-upgrade: ✓ CONFIRMED working (install=pass, upgrade=fail)
  • bad-backup: ✗ BROKEN (fix needed)
  • bad-restore: ✗ BROKEN (fix needed, blocked on backup fix)
  • good-significant: still needs re-run (upgrade flakiness check)