Nodes are not given the roles the user assigned to them

[Udi Kalifon]

I assigned specific nodes to the compute and controller nodes, but when I checked after the deployment - the actual roles that these nodes received were random. One node that was provisioned was actually a node that wasn't assigned a role at all, and was supposed to remain available.

[Jiri Tomasek (jtomasek)]

The nodes assignment now does Nodes Tagging. Nodes Tagging is a process which specifies that certain node is designed to be used for certain Role. The actual assignment of node is done via parameter value.

In the same manner user needs to set the flavor name parameter for each role.

Those operations are currently hardcoded in CLI which is wrong approach and needs to be changed.

The solution is to create a mistral workflow which does:
1. Tag nodes to certain role
2. Update flavor to match the minimum common HW configuration of Nodes assigned to it's Role
3. Set the flavor name parameter (e.g. OvercloudComputeFlavor) to match the name of the flavor assigned to role

4.? set the node count parameter (e.g. ComputeCount) to the number of nodes tagged to a role.
 - this one is questionable

Problem is that I am not sure that we want to hardcode the parameter names as that puts requirements on templates used and will get problematic when custom roles are possible. Maybe the solution is to define a Naming conventions for such parameters?