Merge branch 'slice-main' into evict_if_not_tolerated

Author: pajakd

slice/cmd/main.go

@@ -27,7 +27,6 @@ import (
 
 	"k8s.io/apimachinery/pkg/runtime"
 	utilruntime "k8s.io/apimachinery/pkg/util/runtime"
-	utilfeature "k8s.io/apiserver/pkg/util/feature"
 	clientgoscheme "k8s.io/client-go/kubernetes/scheme"
 	ctrl "sigs.k8s.io/controller-runtime"
 	"sigs.k8s.io/controller-runtime/pkg/certwatcher"
@@ -78,7 +77,6 @@ func main() {
 	var enableHTTP2 bool
 	var sliceHealthNodeAffinityMode string
 	var tlsOpts []func(*tls.Config)
-	var featureGates string
 	flag.StringVar(&metricsAddr, "metrics-bind-address", "0", "The address the metrics endpoint binds to. "+
 		"Use :8443 for HTTPS or :8080 for HTTP, or leave as 0 to disable the metrics service.")
 	flag.StringVar(&probeAddr, "health-probe-bind-address", ":8081", "The address the probe endpoint binds to.")
@@ -101,20 +99,12 @@ func main() {
 		"If set, HTTP/2 will be enabled for the metrics and webhook servers")
 	flag.StringVar(&sliceHealthNodeAffinityMode, "default-slice-health-node-affinity", "HEALTHY",
 		"Default slice health node affinity. Possible values are HEALTHY or HEALTHY_AND_DEGRADED.")
-	flag.StringVar(&featureGates, "feature-gates", "", "A set of key=value pairs that describe feature gates for alpha/experimental features.")
 	opts := zap.Options{
 		Development: true,
 	}
 	opts.BindFlags(flag.CommandLine)
 	flag.Parse()
 
-	if featureGates != "" {
-		if err := utilfeature.DefaultMutableFeatureGate.Set(featureGates); err != nil {
-			setupLog.Error(err, "Unable to set flag gates for known features")
-			os.Exit(1)
-		}
-	}
-
 	var sliceHealthValues []string
 	switch sliceHealthNodeAffinityMode {
 	case "HEALTHY":
@@ -294,10 +284,6 @@ func setupControllers(mgr ctrl.Manager, certsReady chan struct{}, activationTime
 		setupLog.Error(err, "Unable to create webhook", "webhook", "Job")
 		os.Exit(1)
 	}
-	if err := webhooks.SetupPodWebhookWithManager(mgr); err != nil {
-		setupLog.Error(err, "Unable to create webhook", "webhook", "Pod")
-		os.Exit(1)
-	}
 
 	if failedCtrl, err := controller.SetupControllers(mgr, controller.Options{
 		ActivationTimeout: activationTimeout, RetryDelayOnSliceFailure: retryDelay}); err != nil {

slice/config/default/webhook_selector_patch.yaml

@@ -2,8 +2,5 @@ resources:
 - manifests.yaml
 - service.yaml
 
-patches:
-- path: webhook_selector_patch.yaml
-
 configurations:
 - kustomizeconfig.yaml

slice/config/webhook/kustomization.yaml

@@ -42,22 +42,3 @@ webhooks:
     resources:
     - jobsets
   sideEffects: None
-- admissionReviewVersions:
-  - v1
-  clientConfig:
-    service:
-      name: webhook-service
-      namespace: system
-      path: /mutate--v1-pod
-  failurePolicy: Fail
-  name: mpod.accelerator.gke.io
-  rules:
-  - apiGroups:
-    - ""
-    apiVersions:
-    - v1
-    operations:
-    - CREATE
-    resources:
-    - pods
-  sideEffects: None

slice/config/webhook/manifests.yaml

@@ -50,7 +50,6 @@ import (
 
 	"tpu-slice-controller/api/v1beta1"
 	"tpu-slice-controller/internal/core"
-	"tpu-slice-controller/internal/features"
 	"tpu-slice-controller/internal/topology"
 	"tpu-slice-controller/internal/util/api"
 	"tpu-slice-controller/internal/util/node"

slice/config/webhook/webhook_selector_patch.yaml

@@ -1613,12 +1613,10 @@ func TestWorkloadReconciler(t *testing.T) {
 						[]kueue.PodSetUpdate{
 							{
 								Name:         "ps1",
-								Labels:       map[string]string{core.PodWebhookLabelKey: "true"},
 								NodeSelector: map[string]string{"cloud.google.com/gke-tpu-topology": "4x4x4"},
 							},
 							{
 								Name:         "ps2",
-								Labels:       map[string]string{core.PodWebhookLabelKey: "true"},
 								NodeSelector: map[string]string{"cloud.google.com/gke-tpu-topology": "4x4x4"},
 							},
 						})).
@@ -1661,12 +1659,10 @@ func TestWorkloadReconciler(t *testing.T) {
 						[]kueue.PodSetUpdate{
 							{
 								Name:         "ps1",
-								Labels:       map[string]string{core.PodWebhookLabelKey: "true"},
 								NodeSelector: map[string]string{"cloud.google.com/gke-tpu-topology": "4x4x4"},
 							},
 							{
 								Name:         "ps2",
-								Labels:       map[string]string{core.PodWebhookLabelKey: "true"},
 								NodeSelector: map[string]string{"cloud.google.com/gke-tpu-topology": "4x4x4"},
 							},
 						})).
@@ -1823,12 +1819,10 @@ func TestWorkloadReconciler(t *testing.T) {
 						[]kueue.PodSetUpdate{
 							{
 								Name:         "ps1",
-								Labels:       map[string]string{core.PodWebhookLabelKey: "true"},
 								NodeSelector: map[string]string{"cloud.google.com/gke-tpu-topology": "4x4x4"},
 							},
 							{
 								Name:         "ps2",
-								Labels:       map[string]string{core.PodWebhookLabelKey: "true"},
 								NodeSelector: map[string]string{"cloud.google.com/gke-tpu-topology": "4x4x4"},
 							},
 						})).

slice/internal/controller/workload_controller.go

@@ -24,12 +24,10 @@ import (
 
 const (
 	TPUSliceTopologyAnnotation = "cloud.google.com/gke-tpu-slice-topology"
-	TPUSliceNodeLabel          = "cloud.google.com/gke-tpu-slice"
 	TPUTopologyAnnotation      = "cloud.google.com/gke-tpu-topology"
 	TPUAcceleratorLabel        = "cloud.google.com/gke-tpu-accelerator"
 	TPUBlockLabel              = "cloud.google.com/gce-topology-block"
 	TPUSubBlockLabel           = "cloud.google.com/gke-tpu-partition-4x4x4-id"
-	PodWebhookLabelKey         = "cloud.google.com/gke-tpu-slice-pod"
 	TPUResourceName            = "google.com/tpu"
 	TPUsPerCube                = 64
 

slice/internal/controller/workload_controller_test.go

@@ -55,7 +55,7 @@ func BaseSSAReplicatedJob(name string) jobset.ReplicatedJob {
 	}
 }
 
-func AddNodeAffinity(template *corev1.PodTemplateSpec, key string, operator corev1.NodeSelectorOperator, values []string) {
+func AddNodeAffinity(template *corev1.PodTemplateSpec, key string, values []string) {
 	if template.Spec.Affinity == nil {
 		template.Spec.Affinity = &corev1.Affinity{}
 	}
@@ -69,7 +69,7 @@ func AddNodeAffinity(template *corev1.PodTemplateSpec, key string, operator core
 	nodeSelector := template.Spec.Affinity.NodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution
 	requirement := corev1.NodeSelectorRequirement{
 		Key:      key,
-		Operator: operator,
+		Operator: corev1.NodeSelectorOpIn,
 		Values:   values,
 	}
 

slice/internal/core/constants.go

@@ -129,6 +129,6 @@ func (j *JobWrapper) Args(args ...string) *JobWrapper {
 }
 
 func (j *JobWrapper) NodeAffinity(key string, values []string) *JobWrapper {
-	core.AddNodeAffinity(&j.Spec.Template, key, corev1.NodeSelectorOpIn, values)
+	core.AddNodeAffinity(&j.Spec.Template, key, values)
 	return j
 }