Our environment makes extensive use of additional instance creation in jobs (aka multi-node labs). However, we have to share the resources between our build instances and these multi-node labs.

Given that we can easily starve one or the other for resources if we don't balance things carefully, we can regularly have a job start on a lead multi-node instance and then fail because it can't allocate the needed additional instances.

It would be really handy if we could instead of encoding all the instance setup logic in our jobs, instead utilize HEAT orchestration templates that have a lead node identified with a label attached to it. These could then be spun up and wait for Jenkins to attach to the lead node until the entire HEAT stack is online.

I've added the HEAT orchestration developer docs link