Monitoring Supabase Cron – including apps built with Lovable

Setup

Set up in three steps

1. Create a heartbeat

   In your account, create a new one under “Heartbeats”, with the same rhythm as your cron job and some leeway. You receive a ping address of its own, such as https://runcheck.de/a8k3m2x9q1.

2. SQL job: the check-in after the work

   If your job does its work in SQL, add net.http_get with your ping address after it, for example in the SQL Editor. Both run in one transaction, and pg_net only sends requests once it has committed: if the work fails, no check-in goes out. The pg_net extension has to be enabled.

   select cron.schedule('nightly-cleanup', '0 3 * * *', $$
     delete from sessions where expires_at < now();
     select net.http_get('https://runcheck.de/a8k3m2x9q1');
   $$);

3. Edge Function: the check-in as the last line

   If your job calls an Edge Function – for emails or syncing with other services, say – the check-in belongs inside the function, before the response. Only then does it mean: the work is done. The time limit makes sure the function never waits for us.

   Deno.serve(async (req) => {
     // ... your work ...
     await fetch('https://runcheck.de/a8k3m2x9q1', { signal: AbortSignal.timeout(10000) }).catch(() => {});
     return new Response('ok');
   });

The ready-made lines for Linux, PHP, Python, Node.js, n8n, Make, Supabase and Vercel are on the page Cron job monitoring. Your own ping address appears in your account once you create the heartbeat, with the same lines ready to copy.

When something fails

What happens when a run fails

If the job does not run – project paused, job deactivated, error in the query – or the Edge Function fails before its last line, the check-in is missing. After the leeway we alert you; the next check-in brings the all-clear.

FAQ

Common questions about Supabase Cron

I built the app with Lovable and never wrote SQL. Can I still do this?

Yes. Ask Lovable in the chat to call the address https://runcheck.de/a8k3m2x9q1 at the end of your scheduled job – with a ten-second time limit and without letting an error there abort the job. The lines above show what the result should look like.

Does the check-in slow my job down?

No. pg_net works asynchronously: the request goes out in the background after the transaction ends. In the Edge Function, AbortSignal.timeout limits the wait to ten seconds, and errors while checking in are ignored.

How do I see which jobs exist?

With select jobname, schedule, command from cron.job; in the SQL Editor. Recent runs are in cron.job_run_details.