Open Stack

Excerpts from Chris Alfonso's message of 2013-08-15 11:02:13 -0700:
> Hello,
> 
> I would like to utilize Heat's autoscaling feature, but probably in a way it
> was not initially designed to be used.
> 
> In summary what I would like to do is from within OpenShift monitor host
> system capacity and when necessary invoke the scale up or down event.  The
> trick here is I can't just let Heat create a host configured exactly like the
> previous host that Heat created for OpenShift.  I need to be able to provide
> parameters to the scaling event invocation so that when the host comes online
> the OpenShift infrastructure knows what to do with the host. The parameters
> to provide might vary depending on the design approach in Heat's autoscaling.
> I should be able either specify a template to be used that outlines how to
> set up the host OR I should be able to pass in user data that can be passed
> into the template that was used initially when setting up the scaling group.
> 
> I've outlined the desired functionality in more detail here (the autoscaling
> portion):
> https://github.com/openshift/openshift-pep/blob/master/openshift-pep-007.md
> 
> I would like to start the discusion with the Heat group here on how best to
> proceed with Heat's autoscaling and understand how the feature is evolving.

This isn't really auto scaling, this is just "scaling". Auto scaling
involves easily clonable and destroyable instances. OpenShift servers
are neither due to the state they appear to require.

That said, I want some of this too for TripleO so that we can pass in
unique credentials for each member of a group but still have the same
template for them.

For the future, I hope we have a nice imperative way to just say
"remove insance 4 from resource group A". For now, you can emulate
that by maintaining a nested stack that represents "all the things",
and doing updates when you want to add/remove.