damascus-heartbeat bec3e2f6a9 ci(test): reach the postgres service via service name, not 127.0.0.1 ...

Now that BB1's port 5432 is freed, CI gets past the port-bind step
and actually attempts to run the tests. The next failure is:

  psycopg.OperationalError: connection to server at "127.0.0.1",
  port 5432 failed: Connection refused

The job tried 127.0.0.1:5432 but the postgres service container had
just been brought up and wasn't accepting connections yet. The
'5432:5432' port mapping for self-hosted act_runner is racy — the
host-side bind sometimes lags the service ready signal, and the
job's 'Apply schema' step runs before pg_isready returns 0.

Fix:
- Use the service name 'postgres' as DAMASCUS_PG_HOST. In act_runner's
  service network, the service name is the DNS hostname, no port
  mapping required.
- Drop the now-unused 'ports: 5432:5432' binding.

This is the same fix the cron branch fix/compose-db-volume-self-heal
should land in (PR #7). It's small enough to be a follow-up to this
PR; if the test workflow on BB1's runner doesn't resolve 'postgres',
we can revisit.

Co-Authored-By: Claude <noreply@anthropic.com>

2026-06-24 01:04:54 +00:00

..

test.yml

ci(test): reach the postgres service via service name, not 127.0.0.1

2026-06-24 01:04:54 +00:00