Crash alerts: know when your app crashes

Until today, finding out something had gone wrong on velixir meant looking. A worker that ran out of memory at 3am restarted quietly, and the first sign was a customer asking why their email never arrived.

Now velixir tells you.

What does velixir show when my app crashes?

When an app or worker dies, the app's Overview says so: which process, how many times in the last hour, and why, in plain words ("it ran out of memory", "it stopped answering its health check"). Underneath are the last lines it printed before it died, which is usually where the answer is.

Will I get an email?

Only when it's serious. A process that keeps crashing on start is put into back-off by the platform, which means it's probably down. That's the moment worth an email, so the team's Owners and Admins get one, once. Nothing more until it has recovered.

Which webhooks can I get?

Add one under Team settings and pick what you care about:

• app.crashed, app.crash_looping and app.recovered
• deploy.succeeded, deploy.failed, build.failed
• cron.failed and release.failed, from yesterday's processes
• rewind.ready and rewind.applied
• approval.requested, when an agent's key asks before deleting something
• billing.low_balance and budget alerts

Crash events are grouped, at most one every 10 minutes per process with a count, so a bad night doesn't become 200 Slack messages. Their payloads never carry your log lines, because a stack trace can hold a secret; those stay on the dashboard.

Paste a Slack incoming-webhook URL and pick the Slack format, and events show up in the channel as one readable line. Or point it at your own code: every request is signed with the Standard Webhooks headers (webhook-id, webhook-timestamp, webhook-signature), so any of their libraries verifies it in a few lines. Use it to tag a release in your tracker, kick off end-to-end tests once a deploy is live, or page someone when a worker crash loops.

A delivery that fails is retried with back-off for about a day. Every attempt is on the endpoint's page with its status code and response, and any delivery can be sent again.

Crash alerts and webhooks are free on every plan. The webhooks docs list every event and show how to verify a signature.