Specify the head node resources separately from worker resources

[mikeprince4]

Hi,

I am trying to run a managed job (sky jobs) with k8s. Is there a way to specify the size (CPUs, memory) of the head node, separately from the workers? We have jobs that appear to die due to out of memory on the head node, but we don't want to have to increase the size for all nodes merely due to the head node dying.

Here's an example of part of the yaml:

resources:
  infra: k8s
  cpus: 1
  memory: 6

num_nodes: 10

Would be nice if we could do something like this:

resources:
  infra: k8s
  cpus: 1
  memory: 6
  head_node_cpus: 2
  head_node_memory: 10

num_nodes: 10

Thanks in advance!