----- Original Message -----
One other idea:
We simply do this by default for all cloud images, without a timeout - if no cloud-init metadata is provided, you can log in to the hypervisor console and see an autogenerated root password.
I like that. So just to make sure I understand what you mean here: - No new boot menu item - Whenever cloud-init fails to find a source, we generate an expired root pw, disable ssh pwauth, and autologin - Atomic Developer Mode is started by a user command (e.g. a /usr/bin/atomic-devmode)
Is that correct?
One issue here is that cloud-init takes some time before realizing that it has no source. It tries to connect to the default metadata server address. It uses a 50s timeout it seems. We would probably have to reduce this to something more sensible (5s? 10s?).