Allow the identification of "volcano nodes"

[jjuliogallegos01]

What is the problem you're trying to solve

My understanding is that volcano assumes all nodes in a k8s cluster are all volcano nodes for compute intensive batch jobs.

However, there are situations in which we could have general-use k8s clusters and add a few HPC nodes to an existing cluster.

K8s provides some mechanisms to force the execution of pods in a specific group of nodes, but those mechanisms would be implemented at Pod level, meaning that the volcano scheduler may want to run a job on specific nodes, but then if we force the use of specific group of nodes at Pod level, that wouldn't be efficient for the volcano scheduler.

Describe the solution you'd like

I think we can add a parameter on the Volcano helm chart values file to identify what nodes are intended for volcano. It could be a selector based on labels or other parameters similar to pods selector.

e.g:

volcanoNodeSelector:
matchLabels:
nodeType: volcano

That allows an efficient identification of the correct nods.

Additional context

I am assuming forcing the execution of pods on specific nodes outside of volcano is not efficient.

[jjuliogallegos01]

node groups should do it