Fly not rolling back when a bad image is deployed

I’m deploying my Node.js app in a docker image with LiteFS and SQLite. Due to today’s issues Deploying to Fly via GitHub Action failing I deployed from local. For some reason my local build of the Docker image resulted in a bad build that wouldn’t start up. Instead of rolling back to a previous image, Fly just completely shut down my primary regions and failed to bring them back up.

I sent a bunch of logs to @benbjohnson2 (discourse won’t let me upload them here) which you could probably ask him for if you’d like. I’m happy to send them to whoever. There are a lot of logs and I’m not sure what part of them is very useful.

Oh, and I was streaming when this happened so feel free to watch that if you like (I didn’t show logs on stream though): EpicWeb.dev Live stream 🌌 Writing the Deploying Distributed Node Tutorial - YouTube

I chatted with @kentcdodds on Discord but I’ll post here for others to see. I think the rollback issue is because LiteFS is not passing through the error code from the application process. It’s simply returning exit code 0 when the process is done.

I added an issue to track this (#264) and I’ll get it fixed up in the morning.

1 Like