Here is the thing: i don’t know how it have happened, but in the logs i see an instance of one of my previous deployments running. I don’t see this deployment or this instance anywhere, but the logs are “current”: app reacts on the changes in the environment (i played with setting up and destroying redis instance it depends on).
I’ve restarted app many times, tried to re-deploy it or scale to 0. I also tried to delete the app and create it once again (under the same name). Nothing helped: zombie instance is still there, reporting in logs.
Fly.io team, if you are interested: instance id is a7537e58 running in fra region.
Now, thinking about it - i did, and it was the case where docker entrypoint was configured incorrectly, making the main app to run instead of migration.
Some Zombie VMs are a remnant of some bug with Fly apps using the experimental release_command, but one which Fly allegedly has since fixed. You’d have to nerd snipe Fly engs to get the zombie culled: lingering connections and ghost vms. - #2 by ignoramous