Comment 6 for bug 1829896

> Any chance the looping to talk to the metadata server is blocking being able to respond to the poweroff in a timely and graceful fashion?

Maybe, I'm not sure. I thought cloud-init would eventually timeout if it couldn't get a response from the metadata API, like after 10 seconds or something, but I don't know if it would retry but clearly it is retrying. So yeah if it retries 20 times with 10 seconds between, that would be longer than we even have a build_timeout config in tempest (196 seconds) so it would run up against the power off.

I wonder if there is a way to configure that retry loop in cloud-init for our gate images...