pg_cron launcher exits with code 1 every minute, no jobs run since Sep 4 (ticket SU-539363)

[ElazarZilber]

Project ref: wxglxjixukbjduemrsnv (Free plan, Tokyo). Support ticket: SU-539363.

No pg_cron job has run since 2026-09-04 12:46 UTC (max(start_time) in cron.job_run_details).
Jobs are active and configured correctly (database=postgres, username=postgres, nodename=localhost).
Settings: cron.database_name=postgres, cron.launch_active_jobs=on, cron.log_run=on.

Postgres logs show this loop every minute:
background worker "pg_cron launcher" (PID ...) exited with exit code 1
pg_cron scheduler started
There are no FATAL / ERROR / password / authentication lines around it.

I restarted the project from the dashboard (db started 2026-11-05 00:41 UTC), the loop continues, and a test job
(insert into a table every minute) wrote 0 rows in 5+ minutes. pg_net worker is running.

Has anyone seen this, or know how to repair pg_cron on a project?

[edu-lvc]

I found two failure modes that may be worth checking before reinstalling or recreating the extension.

1. Check whether cron.runid_seq is out of sync.

Supabase published a troubleshooting article on November 4 describing a pg_cron launcher crash loop caused by duplicate runid values in cron.job_run_details.

You can check the sequence state with:

SELECT
(SELECT last_value FROM cron.runid_seq) AS sequence_value,
(SELECT is_called FROM cron.runid_seq) AS is_called,
(SELECT MAX(runid) FROM cron.job_run_details) AS max_runid;

If the next sequence value is not greater than max_runid, this could explain the launcher failure. The documented recovery is to reset cron.job_run_details, but make sure to back up its history first.

2. Check whether any scheduled job references a deleted role.

There is also a documented pg_cron issue where a missing role causes the launcher to exit with code 1, even when the associated job is inactive.

SELECT j.jobid, j.username, j.active
FROM cron.job j
LEFT JOIN pg_roles r ON r.rolname = j.username
WHERE r.rolname IS NULL;

3. Check the actual scheduler backend and extension version.

SELECT pid, application_name, backend_type, backend_start, state
FROM pg_stat_activity
WHERE application_name ILIKE '%pg_cron%';
SELECT extversion
FROM pg_extension
WHERE extname = 'pg_cron';

Given that restarting the project did not help and no jobs have executed for months, I would investigate these initialization failures before changing job definitions or reinstalling the extension.

Could you share the results of these checks? They should help narrow down whether this is a sequence inconsistency, a missing-role issue, or another launcher initialization failure.