soulsync/webui
Broque Thomas fc573a5f19 AudioDB worker: stop infinite loop on direct-ID lookup failure (#553)
Track enrichment was stuck in a constant retry loop. Logs showed
nothing but `Read timed out. (read timeout=10)` from
`lookup_track_by_id` repeating against the same track ID. AudioDB
itself was being hammered nonstop with no progress.

Cause: when an entity already has `audiodb_id` populated (from a
manual match or earlier scan) but `audiodb_match_status` is still
NULL — an inconsistent state some import paths can leave behind —
the worker tries a direct ID lookup. If that lookup fails (returns
None on timeout, which AudioDB's `track.php` endpoint hits
frequently because it's slow), the prior code logged "preserving
manual match" and returned WITHOUT marking status. Row stayed NULL
→ queue's NULL-status filter picked it up next tick → tried direct
lookup → timed out → returned → infinite loop.

The "preserve manual match" intent was correct: don't fall through
to the name-search path because that could overwrite a manually-set
`audiodb_id` with a wrong guess. Bug was the missing `_mark_status`
call before the early return.

Fix:

* `_process_item` direct-lookup-failure branch now calls
  `_mark_status(item_type, item_id, 'error')` before returning. The
  existing `audiodb_id` is preserved (column not touched). Queue's
  NULL-status filter no longer re-picks the row.

* `_get_next_item` retry-cutoff queue priorities (4/5/6) extended
  from `audiodb_match_status = 'not_found'` to
  `audiodb_match_status IN ('not_found', 'error')`. Same `retry_days`
  window. Transient AudioDB outages still recover automatically;
  permanently-broken IDs eventually get re-attempted once a month
  rather than staying errored forever.

5 new tests in `tests/test_audiodb_worker_stuck_track.py` use a real
SQLite DB (not mocks) so the SQL queries are actually exercised:

  - lookup-returns-None marks status='error' (no infinite loop)
  - lookup-raises-exception marks status='error' (defensive)
  - lookup-success preserves the existing match-success path
  - error-status row past retry-cutoff gets picked up again
  - error-status row within cutoff stays skipped (loop prevention
    works)

Only triggers for entities in the inconsistent `audiodb_id` set +
`match_status` NULL state. Happy path and already-matched /
already-not-found rows unchanged. Full pytest 2698 passed.

Closes #553.
2026-05-11 14:18:55 -07:00
..
static AudioDB worker: stop infinite loop on direct-ID lookup failure (#553) 2026-05-11 14:18:55 -07:00
index.html Tidal: surface Favorite Tracks as virtual playlist (issue #502) 2026-05-10 21:36:22 -07:00