So, I meant to reply to this a while back but never got to it. :(
I wonder if it would help folks here to share the fedora infrastructure
use case and what we are doing currently.
Basically we use a ton of Fedora vm's for various server use cases
(builders, application servers, etc).
Right now we have ansible call virt-install with parameters we pass in
via ansible host/group/global variables. The virt-install then creates
the vm with that exact memory/disk space/cpus/network config, installs
it using a kickstart file and then once it's up it connects and
configures it.
This has worked really well for us, but it does have a few drawbacks:
It's kind of slow, as it has to boot installer, install, etc.
If something is broken in updates (we usually enable updates) we have to
disable updates and re-run.
A while back we actually considered moving to just using the cloud
image. With that it would be just defining the vm with the
mem/disk/cpu/network we want in libvirt, then copying the cloud image
into the storage, booting it and resizing it. This would be much faster,
but it would mean we would need to apply updates and work out the
resizing.
For our needs we want a install thats just enough to boot, have
NetworkManager/be on the network and have dnf working so we can install
whatever we want. The smaller the better here.
Anyhow, no idea if that revives the conversation, but thought I would
share.
kevin