I am running an auto-scaling RayCluster on K8s using KubeRay. So far it is running great. But one thing I realized is that when I submit a job using
ray job submit, it will try to use all the available resources and even try to trigger scaling of new nodes if the cluster allows it to do so. I want to know if there is a way that I can tell this job to only consume certain resources, such as just 2 CPUs, no more than that, even if the cluster has extra capacity to give.
Thanks a lot,