A old container (allocation) in my app is stuck and can't be terminated.

1

A old container version cannot be terminate. It’s marked to be stop but it’s still running for many hours.

Screenshot from 2023-01-28 18-26-08
Screenshot from 2023-01-28 18-26-081383×264 21.1 KB

I’ve tried fly scale count 0 and v13 is terminated successfully but v8 is still stuck.

I didn’t try delete fly app but I think this is such a destructive action.
There’s no way to force terminate a specific allocation id.

Is there any way to fix this?

Thanks

===

Further update:
My v8 app status is shown below.

v8      true    rollback        cancelled       Reverting to version 6                                  36m38s ago

===

Update 2:
I managed to ssh into the container and kill the app’s process. It just restarted and continue to be in running state.

===

Update 3:
The container was terminated! I don’t know what happened, maybe fly team has done something.

Thanks everyone.

2

I have a similar issue with an outdated instance hanging around in “desired=stop, status=running” state. I tried manually stopping and restarting the vm with the cli, but to no effect.

Rollouts were really slow in the affected region (nrt) earlier today as well, while others (ams) are working just fine, so perhaps it’s region dependent.

1 Like
4

Phew! I’m not the only one having trouble with the region nrt. The container in nrt is super slow for every status, and every command. And eventually it seems failed to deploy. As hyperborea said, I also think it is region dependent problem because the other regions like hkg, sjc have no problem with same deploy.

5

Out of curiosity does your app configure a kill_timeout?

We usually don’t forcefully kill those until the timeout is reached

6

Yes, it’s configured for 5 seconds.

7

The container was terminated! I don’t know what happened, maybe fly team has done something.