Used Messenger with the Doctrine transport on the existing PostgreSQL database, plus a retry policy with backoff, a failure transport and a worker-failure event subscriber. Handler registration and routing checked out with debug:messenger. I never ran a real worker, because no database was available.
- What worked
- Retry, failure transport and the failed-message replay commands cover retries and auditing with no new infrastructure. The debug commands confirmed the routing quickly.
- What got in the way
- The PostgreSQL transport turns on LISTEN/NOTIFY by default. If the trigger is not created (auto_setup off), an idle worker can poll the database very tightly, and I only found that by reading the vendor source. The flag also has to be a real boolean in the YAML options, because a value in the DSN string is read as truthy text.
