How to diagnose recurring Creating release fails with “An unknown error occurred”

I am deploying a Deno based app running Dinatra. The initial deploy always seems to work. Subsequent deploys seem to always fail though. When I destroy the app, recreate, and re-deploy it, that works. Unless I destroy and re-deploy to the same name which fails.

The error isn’t helpful of course as it’s unknown. Is there any way to get a more detailed error log?

You could try LOG_LEVEL=debug fly deploy to get a bit more detail about the deployment process.

And maybe also in a separate tab/window, run fly logs too.

A combination of those should show what’s up. Like if a healthcheck is failing, an in-app exception thrown etc.

1 Like

Thanks @greg! I’ve been running logs to see what happens but as it crashes just after pushing the image and creating the release, the running instances are never touched and no new ones spin up. And now my other app, which until now has have no deployment issues, is also doing that. :frowning_face:

[edit] The last one was user error. I made an invalid tweak to the toml, and LOG_LEVEL=debug was :rocket: spot-on to diagnose. Thanks for that!

Now back to the original crashing item…

1 Like

Hmm for the Deno server the graphql deploy image command errors out with a 500 so seeing this info is very helpful. The image gets pushed, just fails on deploying the image…

A deploy to the same-named Deno app succeeded. And two more. I don’t know what happened but something has improved.

I added a SIGINT handler to hopefully help handle the shutdowns better… but looks like it wasn’t needed. Still, I can shut down the server and clean things up so that’s a :heavy_plus_sign: .

If any folks can speak to any changes, that’d be good to know.

Which region are you deploying in? There were a few infrastructure hiccups yesterday that may be related.

Thanks @joshua. My initial deploy region was lax, then switched to ord. I had issues with both as I recall. Again it was only the Deno Dinatra based app. The NATS cluster was fine throughout though it’s a 3 node cluster on lax, iad, and ord.