App platform health checks for worker components | Voters

App platform health checks for worker components

Goncalo Oliveira

Health checks can only be set for web components.
Sometimes applications do not expose a public web interface, but they can still provide health checks through HTTP or TCP.
The expectation is that failing the health check would cause the container to be restarted.

May 4, 2023

Tomáš Livora

DigitalOcean restarts workers when the main process exits. But not in case of other problems. For example, our workers get stuck from time to time. They suddenly stop executing their tasks, producing logs, or sending heartbeats. And their CPU keeps running at 100%. It would be nice if DO allowed us to use a custom (HTTP) health check so we don't need to restart these workers manually in such cases.

And BTW, manually restarting a worker is problematic as well. You can restart it by going to Console tab and killing the main process. But that only works when you have a single instance. If you have multiple instances of a given worker running and you're not lucky, you are not able to get to the right console. And since there's no way to restart a component or even the whole app, you actually need to redeploy the app.

Nathan Palmer

This would be useful. If a worker dies I need a reliable way for Digital Ocean to restart it.