machines transition through 'down' state on the way to being fully deployed

[Jason Hobbs]

Machines often end up in 'down' state while they are being setup. They don't need any attention, it just seems to be some side effect of how juju works.

An example status yaml: http://paste.ubuntu.com/26526625/

This is the same 'down' state that machines end up in when they fail to deploy and are actually in an error state that requires attention.

Since 'down' state is being used both for transiently 'down' machines that are actually just deploying, and for really down machines that require operator attention, there is no way to programatically tell the difference between these two states.

This means we can't tell 'juju wait' or other tools to give up on a deployment when a machine is really down and needs operator attention. We have to instead wait for the deployment to timeout, which adds hours to finding out we've failed in test runs and in automated deploys that use juju wait (our field deployment tools do this).

It seems to me that 'down' should only be used for systems that are really down, and that cases where juju is transiently setting a machine state to 'down' are bugs.

[Tim Penhey (thumper)]

This is a side-effect of how presence is currently implemented.

We do have plans to change the way the agent presence is handled, and that should fix this too.

[Canonical Juju QA Bot (juju-qa-bot)]

This bug has not been updated in 2 years, so we're marking it Low importance. If you believe this is incorrect, please update the importance.