remove priority/affinity from template -- not compatible with Kueue

Author: dgrove-oss

src/codeflare_sdk/templates/base-template.yaml

@@ -3,11 +3,7 @@ kind: AppWrapper
 metadata:
   name: aw-kuberay
   namespace: default
-  #new addition
-  labels:
-    orderedinstance: "m4.xlarge_g4dn.xlarge"
 spec:
-  priority: 9
   resources:
     Items: []
     GenericItems:
@@ -103,16 +99,6 @@ spec:
             #pod template
             template:
               spec:
-                #new addition
-                affinity:
-                  nodeAffinity:
-                    requiredDuringSchedulingIgnoredDuringExecution:
-                      nodeSelectorTerms:
-                      - matchExpressions:
-                        - key: aw-kuberay
-                          operator: In
-                          values:
-                          - "aw-kuberay"
                 containers:
                 # The Ray head pod
                 - name: ray-head
@@ -197,15 +183,6 @@ spec:
                 # finalizers:
                 # - kubernetes
               spec:
-                affinity:
-                  nodeAffinity:
-                    requiredDuringSchedulingIgnoredDuringExecution:
-                      nodeSelectorTerms:
-                      - matchExpressions:
-                        - key: aw-kuberay
-                          operator: In
-                          values:
-                          - "aw-kuberay"
                 containers:
                 - name: machine-learning # must consist of lower case alphanumeric characters or '-', and must start and end with an alphanumeric character (e.g. 'my-name',  or '123-abc'
                   image: quay.io/project-codeflare/ray:latest-py39-cu118

src/codeflare_sdk/utils/generate_yaml.py

@@ -87,16 +87,6 @@ def update_names(yaml, item, appwrapper_name, cluster_name, namespace):
     lower_meta["namespace"] = namespace
 
 
-def update_labels(yaml, instance_types):
-    metadata = yaml.get("metadata")
-    metadata.pop("labels")
-
-
-def update_priority(yaml, item):
-    spec = yaml.get("spec")
-    spec.pop("priority")
-
-
 def update_custompodresources(
     item,
     min_cpu,
@@ -147,10 +137,6 @@ def update_custompodresources(
         sys.exit("Error: malformed template")
 
 
-def update_affinity(spec, appwrapper_name):
-    spec.pop("affinity")
-
-
 def update_image(spec, image):
     containers = spec.get("containers")
     for container in containers:
@@ -219,7 +205,6 @@ def update_nodes(
 
         for comp in [head, worker]:
             spec = comp.get("template").get("spec")
-            update_affinity(spec, appwrapper_name)
             update_image_pull_secrets(spec, image_pull_secrets)
             update_image(spec, image)
             update_env(spec, env)
@@ -370,8 +355,6 @@ def generate_appwrapper(
         cluster_name,
         namespace,
     )
-    update_labels(user_yaml, instance_types)
-    update_priority(user_yaml, item)
     update_custompodresources(
         item,
         min_cpu,

tests/test-case-bad.yaml

@@ -6,7 +6,6 @@ metadata:
   nam: unit-test-cluster
   namspace: ns
 spec:
-  priority: 9
   resources:
     GenericItems:
     - custompodresources:
@@ -58,15 +57,6 @@ spec:
             serviceType: ClusterIP
             template:
               spec:
-                affinity:
-                  nodeAffinity:
-                    requiredDuringSchedulingIgnoredDuringExecution:
-                      nodeSelectorTerms:
-                      - matchExpressions:
-                        - key: unit-test-cluster
-                          operator: In
-                          values:
-                          - unit-test-cluster
                 containers:
                 - env:
                   - name: MY_POD_IP
@@ -115,15 +105,6 @@ spec:
                 labels:
                   key: value
               spec:
-                affinity:
-                  nodeAffinity:
-                    requiredDuringSchedulingIgnoredDuringExecution:
-                      nodeSelectorTerms:
-                      - matchExpressions:
-                        - key: unit-test-cluster
-                          operator: In
-                          values:
-                          - unit-test-cluster
                 containers:
                 - env:
                   - name: MY_POD_IP
@@ -149,19 +130,3 @@ spec:
                       memory: 5G
                       nvidia.com/gpu: 7
       replicas: 1
-    - generictemplate:
-        apiVersion: route.openshift.io/v1
-        kind: Route
-        metadata:
-          labels:
-            odh-ray-cluster-service: unit-test-cluster-head-svc
-          name: ray-dashboard-unit-test-cluster
-          namespace: ns
-        spec:
-          port:
-            targetPort: dashboard
-          to:
-            kind: Service
-            name: unit-test-cluster-head-svc
-      replicas: 1
-    Items: []

tests/unit_test.py

@@ -1109,23 +1109,6 @@ def get_ray_obj(group, version, namespace, plural, cls=None):
                         "template": {
                             "metadata": {},
                             "spec": {
-                                "affinity": {
-                                    "nodeAffinity": {
-                                        "requiredDuringSchedulingIgnoredDuringExecution": {
-                                            "nodeSelectorTerms": [
-                                                {
-                                                    "matchExpressions": [
-                                                        {
-                                                            "key": "quicktest",
-                                                            "operator": "In",
-                                                            "values": ["quicktest"],
-                                                        }
-                                                    ]
-                                                }
-                                            ]
-                                        }
-                                    }
-                                },
                                 "containers": [
                                     {
                                         "env": [
@@ -1264,23 +1247,6 @@ def get_ray_obj(group, version, namespace, plural, cls=None):
                                     "labels": {"key": "value"},
                                 },
                                 "spec": {
-                                    "affinity": {
-                                        "nodeAffinity": {
-                                            "requiredDuringSchedulingIgnoredDuringExecution": {
-                                                "nodeSelectorTerms": [
-                                                    {
-                                                        "matchExpressions": [
-                                                            {
-                                                                "key": "quicktest",
-                                                                "operator": "In",
-                                                                "values": ["quicktest"],
-                                                            }
-                                                        ]
-                                                    }
-                                                ]
-                                            }
-                                        }
-                                    },
                                     "containers": [
                                         {
                                             "env": [
@@ -1678,7 +1644,7 @@ def get_aw_obj(group, version, namespace, plural):
                 "kind": "AppWrapper",
                 "metadata": {
                     "annotations": {
-                        "kubectl.kubernetes.io/last-applied-configuration": '{"apiVersion":"codeflare.dev/v1beta1","kind":"AppWrapper","metadata":{"annotations":{},"name":"quicktest1","namespace":"ns"},"spec":{"priority":9,"resources":{"GenericItems":[{"custompodresources":[{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}},{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}],"generictemplate":{"apiVersion":"ray.io/v1","kind":"RayCluster","metadata":{"labels":{"appwrapper.codeflare.dev":"quicktest1","controller-tools.k8s.io":"1.0"},"name":"quicktest1","namespace":"ns"},"spec":{"autoscalerOptions":{"idleTimeoutSeconds":60,"imagePullPolicy":"Always","resources":{"limits":{"cpu":"500m","memory":"512Mi"},"requests":{"cpu":"500m","memory":"512Mi"}},"upscalingMode":"Default"},"enableInTreeAutoscaling":false,"headGroupSpec":{"rayStartParams":{"block":"true","dashboard-host":"0.0.0.0","num-gpus":"0"},"serviceType":"ClusterIP","template":{"spec":{"containers":[{"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","imagePullPolicy":"Always","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"ray-head","ports":[{"containerPort":6379,"name":"gcs"},{"containerPort":8265,"name":"dashboard"},{"containerPort":10001,"name":"client"}],"resources":{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}}}]}}},"rayVersion":"1.12.0","workerGroupSpecs":[{"groupName":"small-group-quicktest","maxReplicas":1,"minReplicas":1,"rayStartParams":{"block":"true","num-gpus":"0"},"replicas":1,"template":{"metadata":{"annotations":{"key":"value"},"labels":{"key":"value"}},"spec":{"containers":[{"env":[{"name":"MY_POD_IP","valueFrom":{"fieldRef":{"fieldPath":"status.podIP"}}}],"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"machine-learning","resources":{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}}],}}}]}},"replicas":1},{"generictemplate":{"apiVersion":"route.openshift.io/v1","kind":"Route","metadata":{"labels":{"odh-ray-cluster-service":"quicktest-head-svc"},"name":"ray-dashboard-quicktest","namespace":"default"},"spec":{"port":{"targetPort":"dashboard"},"to":{"kind":"Service","name":"quicktest-head-svc"}}},"replica":1}],"Items":[]}}}\n'
+                        "kubectl.kubernetes.io/last-applied-configuration": '{"apiVersion":"codeflare.dev/v1beta1","kind":"AppWrapper","metadata":{"annotations":{},"name":"quicktest1","namespace":"ns"},"spec":{"resources":{"GenericItems":[{"custompodresources":[{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}},{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}],"generictemplate":{"apiVersion":"ray.io/v1","kind":"RayCluster","metadata":{"labels":{"appwrapper.codeflare.dev":"quicktest1","controller-tools.k8s.io":"1.0"},"name":"quicktest1","namespace":"ns"},"spec":{"autoscalerOptions":{"idleTimeoutSeconds":60,"imagePullPolicy":"Always","resources":{"limits":{"cpu":"500m","memory":"512Mi"},"requests":{"cpu":"500m","memory":"512Mi"}},"upscalingMode":"Default"},"enableInTreeAutoscaling":false,"headGroupSpec":{"rayStartParams":{"block":"true","dashboard-host":"0.0.0.0","num-gpus":"0"},"serviceType":"ClusterIP","template":{"spec":{"containers":[{"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","imagePullPolicy":"Always","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"ray-head","ports":[{"containerPort":6379,"name":"gcs"},{"containerPort":8265,"name":"dashboard"},{"containerPort":10001,"name":"client"}],"resources":{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}}}]}}},"rayVersion":"1.12.0","workerGroupSpecs":[{"groupName":"small-group-quicktest","maxReplicas":1,"minReplicas":1,"rayStartParams":{"block":"true","num-gpus":"0"},"replicas":1,"template":{"metadata":{"annotations":{"key":"value"},"labels":{"key":"value"}},"spec":{"containers":[{"env":[{"name":"MY_POD_IP","valueFrom":{"fieldRef":{"fieldPath":"status.podIP"}}}],"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"machine-learning","resources":{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}}],}}}]}},"replicas":1},{"generictemplate":{"apiVersion":"route.openshift.io/v1","kind":"Route","metadata":{"labels":{"odh-ray-cluster-service":"quicktest-head-svc"},"name":"ray-dashboard-quicktest","namespace":"default"},"spec":{"port":{"targetPort":"dashboard"},"to":{"kind":"Service","name":"quicktest-head-svc"}}},"replica":1}],"Items":[]}}}\n'
                     },
                     "creationTimestamp": "2023-02-22T16:26:07Z",
                     "generation": 4,
@@ -1738,7 +1704,6 @@ def get_aw_obj(group, version, namespace, plural):
                     "uid": "6334fc1b-471e-4876-8e7b-0b2277679235",
                 },
                 "spec": {
-                    "priority": 9,
                     "resources": {
                         "GenericItems": [
                             {
@@ -2005,7 +1970,7 @@ def get_aw_obj(group, version, namespace, plural):
                 "kind": "AppWrapper",
                 "metadata": {
                     "annotations": {
-                        "kubectl.kubernetes.io/last-applied-configuration": '{"apiVersion":"codeflare.dev/v1beta1","kind":"AppWrapper","metadata":{"annotations":{},"name":"quicktest2","namespace":"ns"},"spec":{"priority":9,"resources":{"GenericItems":[{"custompodresources":[{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}},{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}],"generictemplate":{"apiVersion":"ray.io/v1","kind":"RayCluster","metadata":{"labels":{"appwrapper.codeflare.dev":"quicktest2","controller-tools.k8s.io":"1.0"},"name":"quicktest2","namespace":"ns"},"spec":{"autoscalerOptions":{"idleTimeoutSeconds":60,"imagePullPolicy":"Always","resources":{"limits":{"cpu":"500m","memory":"512Mi"},"requests":{"cpu":"500m","memory":"512Mi"}},"upscalingMode":"Default"},"enableInTreeAutoscaling":false,"headGroupSpec":{"rayStartParams":{"block":"true","dashboard-host":"0.0.0.0","num-gpus":"0"},"serviceType":"ClusterIP","template":{"spec":{"containers":[{"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","imagePullPolicy":"Always","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"ray-head","ports":[{"containerPort":6379,"name":"gcs"},{"containerPort":8265,"name":"dashboard"},{"containerPort":10001,"name":"client"}],"resources":{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}}}]}}},"rayVersion":"1.12.0","workerGroupSpecs":[{"groupName":"small-group-quicktest","maxReplicas":1,"minReplicas":1,"rayStartParams":{"block":"true","num-gpus":"0"},"replicas":1,"template":{"metadata":{"annotations":{"key":"value"},"labels":{"key":"value"}},"spec":{"containers":[{"env":[{"name":"MY_POD_IP","valueFrom":{"fieldRef":{"fieldPath":"status.podIP"}}}],"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"machine-learning","resources":{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}}],}}}]}},"replicas":1},{"generictemplate":{"apiVersion":"route.openshift.io/v1","kind":"Route","metadata":{"labels":{"odh-ray-cluster-service":"quicktest-head-svc"},"name":"ray-dashboard-quicktest","namespace":"default"},"spec":{"port":{"targetPort":"dashboard"},"to":{"kind":"Service","name":"quicktest-head-svc"}}},"replica":1}],"Items":[]}}}\n'
+                        "kubectl.kubernetes.io/last-applied-configuration": '{"apiVersion":"codeflare.dev/v1beta1","kind":"AppWrapper","metadata":{"annotations":{},"name":"quicktest2","namespace":"ns"},"spec":{"resources":{"GenericItems":[{"custompodresources":[{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}},{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"replicas":1,"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}],"generictemplate":{"apiVersion":"ray.io/v1","kind":"RayCluster","metadata":{"labels":{"appwrapper.codeflare.dev":"quicktest2","controller-tools.k8s.io":"1.0"},"name":"quicktest2","namespace":"ns"},"spec":{"autoscalerOptions":{"idleTimeoutSeconds":60,"imagePullPolicy":"Always","resources":{"limits":{"cpu":"500m","memory":"512Mi"},"requests":{"cpu":"500m","memory":"512Mi"}},"upscalingMode":"Default"},"enableInTreeAutoscaling":false,"headGroupSpec":{"rayStartParams":{"block":"true","dashboard-host":"0.0.0.0","num-gpus":"0"},"serviceType":"ClusterIP","template":{"spec":{"containers":[{"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","imagePullPolicy":"Always","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"ray-head","ports":[{"containerPort":6379,"name":"gcs"},{"containerPort":8265,"name":"dashboard"},{"containerPort":10001,"name":"client"}],"resources":{"limits":{"cpu":2,"memory":"8G","nvidia.com/gpu":0},"requests":{"cpu":2,"memory":"8G","nvidia.com/gpu":0}}}]}}},"rayVersion":"1.12.0","workerGroupSpecs":[{"groupName":"small-group-quicktest","maxReplicas":1,"minReplicas":1,"rayStartParams":{"block":"true","num-gpus":"0"},"replicas":1,"template":{"metadata":{"annotations":{"key":"value"},"labels":{"key":"value"}},"spec":{"containers":[{"env":[{"name":"MY_POD_IP","valueFrom":{"fieldRef":{"fieldPath":"status.podIP"}}}],"image":"ghcr.io/foundation-model-stack/base:ray2.1.0-py38-gpu-pytorch1.12.0cu116-20221213-193103","lifecycle":{"preStop":{"exec":{"command":["/bin/sh","-c","ray stop"]}}},"name":"machine-learning","resources":{"limits":{"cpu":1,"memory":"2G","nvidia.com/gpu":0},"requests":{"cpu":1,"memory":"2G","nvidia.com/gpu":0}}}],}}}]}},"replicas":1},{"generictemplate":{"apiVersion":"route.openshift.io/v1","kind":"Route","metadata":{"labels":{"odh-ray-cluster-service":"quicktest-head-svc"},"name":"ray-dashboard-quicktest","namespace":"default"},"spec":{"port":{"targetPort":"dashboard"},"to":{"kind":"Service","name":"quicktest-head-svc"}}},"replica":1}],"Items":[]}}}\n'
                     },
                     "creationTimestamp": "2023-02-22T16:26:07Z",
                     "generation": 4,
@@ -2065,7 +2030,6 @@ def get_aw_obj(group, version, namespace, plural):
                     "uid": "6334fc1b-471e-4876-8e7b-0b2277679235",
                 },
                 "spec": {
-                    "priority": 9,
                     "resources": {
                         "GenericItems": [
                             {