Comment 2 for bug 1228362

I am thinking about how updates should happen (not necessarily how they do happen now).

I suggest they should wait for waitconditions too. I guess CFN doesn't do that, but we will definitely need a resource that does all of what the cfn waitcondition does plus wait on updates for canary based rolling updates to work properly.

Anyway, a simpler use case that applies now is the autoscaling group problem raised by Simon Pasquier on openstack-dev

http://lists.openstack.org/pipermail/openstack-dev/2013-September/015205.html

Simon is changing a parameter that controls both the size of the instance group and the count expected in the waitcondition. This is causing the waitcondition to be replaced, but without cfn-hup the same number of signals do not arrive, and thus the stack is "stuck". Perhaps this just needs to be documented prominently and instance groups need to always be configured to use cfn-hup/os-collect-config to watch the count for changes.