Could have some interaction with #825

There is a fairly thin one-way line of communication with the main process (which kills the task). Have a look at cleanupTask in koji.daemon (and the functions it calls, like _doKill).

We always send a SIGINT first "to let mock mock try to clean up." This could be caught by the child, and it could take better cleanup actions. Perhaps we need to selectively extend the timeout here.

Also, we might not be finding all the children in the main kojid process, perhaps that is part of the task-side cleanup too.

Possibly we could allow task handlers to declare a cleanup method that is run in the main process before the killing occurs. We'd still have to figure out how trigger things (perhaps a different signal, or perhaps same with a longer timeout and/or deeper monitoring).

@mikem Any update on this?

@breilly do you recall the details on this?

Brendan is out for a couple of weeks. @mohanboddu are you running into this problem in Fedora?

@dgregor I just want to leave a note of what we talked:

Brendan is out for a couple of weeks. @mohanboddu are you running into this problem in Fedora?

Yes, we are hitting it in Fedora. So, either Kevin or I manually go and clean up the vm's when something like this happens.

I guess its to much to ask for to get this fixed with in a month (by F30 GA) with Brendan not around. But let us know :smile:

Will be fixed on imagefactory side. On koji side it means only raising required imagefactory version in spec after it will be merged/released.

https://github.com/redhat-imaging/imagefactory/pull/431

Let's close it here and leave it on image-factory.

This issue has been migrated to Fedora Forge:
https://forge.fedoraproject.org/koji/koji/issues/832

Please continue any further discussion there.