🐈 Task: ersilia delete should not silently exit when deleting a model fetched from DockerHub if it cannot delete its image

[DhanshreeA]

Summary

Refer to the issue

When running ersilia delete eosxxxxx, where eosxxxxx was fetched using its Docker image, if ersilia finds that the Docker engine is not running on the user's machine, it silently exits. In the ideal case, we expect the image to be deleted, however whether or not the engine is active on the user's system, is out of our control. Nonetheless, we need to flag this at the very least and raise a message to user asking them to start docker engine on their system and then run this command.

There is another solution to this that's a bit more robust and upfront: We could utilize the model source information in the model bundle, determining how the model was fetched in the very beginning of the delete command invocation. If it was fetched using DockerHub, we should check whether the engine is running and flag that to the user upfront, instead of deleting things halfway, checking whether the engine is running or not, and then raising this error. In this case, we can prompt the user to start their docker engine and re-invoke the delete command.

In both cases, if the engine is not running, the user would have to start it and re-run the command, however in the second approach, the model wouldn't end up dangling in a half fetched, half not fetched state where model directories are deleted, but the image persists. I prefer this solution, however I am open to suggestions from a usability point of view.

Thoughts @miquelduranfrigola @GemmaTuron?

Objective(s)

No response

Documentation

No response

[miquelduranfrigola]

I like your robust solution, @DhanshreeA. I think this is the way to go.

[GemmaTuron]

I agree the robust solution is best.
I will move this to Ersilia Extension, as it is not crucial at the moment since we are few users of the Hub locally and we can manage, but certainly this will need to be tackled in the future

[DhanshreeA]

Moving this to Ersilia Extension and prioritizing other Docker issues right now.

[DhanshreeA]

Actually, on second thoughts, @GemmaTuron I don't think this is an extension issue, and we cannot simply manage this internally - this is very much an Ersilia maintenance issue because without addressing this issue we would run into what you experienced with a model in issue #1373. So I will assign this P1 and address it accordingly.