Description of problem: If you have an instance which disappears on the backend cloud provider, dbomatic will move its state to "vanished," and then destroy the instance on subsequent queries. However, we never actually do anything with the deployment. This leaves you in a bizarre state, where you have a Deployment in running state, but with zero instances. There are a few things we should do here (either stop deleting instances in this case, or also delete the deployment), but in any case, our changes to the instance need to propagate up to the deployment.
Oh, there isn't an actual :status attribute on deployments, it's just derived from the instances. So this is just a bug -- a deployment with no instances shouldn't be in "running" state.
I sent out a series of patches to the list: http://lists.fedorahosted.org/pipermail/aeolus-devel/2012-February/009068.html I ended up combining this with 795794 as well in the patchset. They're distinct but related issues.
Pushed 3 patches to master to address these issues. The commit hashes are in https://bugzilla.redhat.com/show_bug.cgi?id=795794
To test this I think either Dave or myself will have to shut down a vsphere or rhevm hypervisor that hosts a running instance.
good 2 go with aeolus-conductor-0.8.0-35.el6.noarch