Here is how you can add a nodeSelector variable to your k8s work pool base job template. Define a template variable definition for

node_selector

trough the "Advanced" tab when creating/editing your work pool:

"variables": {
    "type": "object",
    "properties": {
      "node_selector": {
        "type": "object",
        "title": "Node Selector",
        "description": "A selector which must be true for the pod to fit on a node. Selector which must match a node's labels for the pod to be scheduled on that node. More info: <https://kubernetes.io/docs/concepts/configuration/assign-pod-node/>"
      },
      ... # other variables

Then create a

nodeSelector

with the jinja template value in

spec

part of the job manifest:

"spec": {
            "containers": [
              {
                "env": "{{ env }}",
                "args": "{{ command }}",
                "name": "prefect-job",
                "image": "{{ image }}",
                "resources": "{{ resources }}",
                "imagePullPolicy": "{{ image_pull_policy }}"
              }
            ],
            "completions": 1,
            "parallelism": 1,
            "nodeSelector": "{{ node_selector }}", # <--- added here
            "restartPolicy": "Never",
            "serviceAccountName": "{{ service_account_name }}"
          }

Once you update template, the

Node Selector

variable will appear on the "Defaults" page when you go to edit your work pool.