This ticket is a duplicate of https://pagure.io/taskotron/libtaskotron/issue/239

Also, everyone please try to come up with use cases where we would actually prefer to run directly on virt-host instead of a disposable client, outside of performance. I'm not sure if there are such use cases, but it would be nice to think about it before we make our implementation depend heavily on disposable clients-only.

I'm not thinking of many cases that aren't covered by something similar to the cloud image testing case that's described in the docs. Image building, maybe (assuming we do that as a task) but other than that I think it's mostly a performance issue of spin-up time vs. execution time.

FWIW, I don't think it'd be all that complicated to keep some vm clients around, stick them in a separate builder and send a subset of tasks to them.

Image building looks like a good candidate for non-disposable (and non-VM) clients. So maybe we can generalize that anything that requires virt support (running VMs) will probably need a bare-metal machine in a non-diposable client mode. Unless we properly investigate nested virt and find out this is no longer an issue.

Is this something that we still want to look into? Everything seems to be running fine and we've solved the image building issue by building them outside taskotron on bare-metal.

I think this can be closed

I'm taking silence as "no objection" and closing the ticket. Re-open with explanation if you think this still needs to be done.