Merge remote-tracking branch 'upstream/master' into tfjobs_iap

OWNERS

@@ -6,8 +6,10 @@ approvers:
   - wbuchwalter
 reviewers:
   - ankushagarwal
+  - cheyang
+  - codeflitting
+  - ddysher
   - Jimexist
   - lluunn
-  - mitake
-  - ddysher
   - willb
+  - yph152

cmd/tf-operator.v2/app/options/options.go

@@ -20,11 +20,12 @@ import (
 
 // ServerOption is the main context object for the controller manager.
 type ServerOption struct {
-	Kubeconfig    string
-	MasterURL     string
-	Threadiness   int
-	PrintVersion  bool
-	JSONLogFormat bool
+	Kubeconfig           string
+	MasterURL            string
+	Threadiness          int
+	PrintVersion         bool
+	JSONLogFormat        bool
+	EnableGangScheduling bool
 }
 
 // NewServerOption creates a new CMServer with a default config.
@@ -46,4 +47,5 @@ func (s *ServerOption) AddFlags(fs *flag.FlagSet) {
 
 	fs.BoolVar(&s.JSONLogFormat, "json-log-format", false,
 		"Set true to use json style log format. Set false to use plaintext style log format")
+	fs.BoolVar(&s.EnableGangScheduling, "enable-gang-scheduling", false, "Set true to enable gang scheduling by kube-arbitrator.")
 }

cmd/tf-operator.v2/app/server.go

@@ -98,7 +98,7 @@ func Run(opt *options.ServerOption) error {
 	unstructuredInformer := controller.NewUnstructuredTFJobInformer(kcfg)
 
 	// Create tf controller.
-	tc := controller.NewTFJobController(unstructuredInformer, kubeClientSet, tfJobClientSet, kubeInformerFactory, tfJobInformerFactory)
+	tc := controller.NewTFJobController(unstructuredInformer, kubeClientSet, tfJobClientSet, kubeInformerFactory, tfJobInformerFactory, *opt)
 
 	// Start informer goroutines.
 	go kubeInformerFactory.Start(stopCh)

dashboard/backend/handler/api_handler.go

@@ -3,6 +3,7 @@
 package handler
 
 import (
+	"fmt"
 	"net/http"
 
 	"github.com/emicklei/go-restful"
@@ -159,7 +160,7 @@ func (apiHandler *APIHandler) handleGetTFJobDetail(request *restful.Request, res
 
 	// Get associated pods
 	pods, err := apiHandler.cManager.ClientSet.CoreV1().Pods(namespace).List(metav1.ListOptions{
-	// LabelSelector: fmt.Sprintf("kubeflow.org=,runtime_id=%s", job.Spec.RuntimeId),
+		LabelSelector: fmt.Sprintf("group_name=kubeflow.org,tf_job_name=%s", name),
 	})
 	if err != nil {
 		log.Warningf("failed to list pods for TFJob %v under namespace %v: %v", name, namespace, err)

developer_guide.md

@@ -119,7 +119,7 @@ Or
 
 ```bash
 # If you are using v1alpha2
-cd ./test/e2e/dist-mnist
+cd ./examples/v1alpha2/dist-mnist
 docker build -f Dockerfile -t kubeflow/tf-dist-mnist-test:1.0 .
 kubectl create -f ./tf-job-mnist.yaml
 ```

hack/update-codegen.sh

@@ -43,5 +43,5 @@ ${GOPATH}/bin/defaulter-gen  --input-dirs github.com/kubeflow/tf-operator/pkg/ap
 cd - > /dev/null
 
 echo "Generating OpenAPI specification for v1alpha2"
-${GOPATH}/bin/openapi-gen --input-dirs github.com/kubeflow/tf-operator/pkg/apis/tensorflow/v1alpha2,k8s.io/api/core/v1 --output-package github.com/kubeflow/tf-operator/pkg/apis/tensorflow/v1alpha2 --go-header-file hack/boilerplate/boilerplate.go.txt "$@"
+${GOPATH}/bin/openapi-gen --input-dirs github.com/kubeflow/tf-operator/pkg/apis/tensorflow/v1alpha2,k8s.io/api/core/v1,k8s.io/apimachinery/pkg/apis/meta/v1,k8s.io/apimachinery/pkg/api/resource,k8s.io/apimachinery/pkg/runtime,k8s.io/apimachinery/pkg/util/intstr,k8s.io/apimachinery/pkg/version --output-package github.com/kubeflow/tf-operator/pkg/apis/tensorflow/v1alpha2 --go-header-file hack/boilerplate/boilerplate.go.txt "$@"
 cd - > /dev/null

pkg/apis/tensorflow/v1alpha2/defaults.go

@@ -88,7 +88,7 @@ func setTypeNameToCamelCase(tfJob *TFJob, typ TFReplicaType) {
 
 // SetDefaults_TFJob sets any unspecified values to defaults.
 func SetDefaults_TFJob(tfjob *TFJob) {
-	// Set default cleanpod policy to All.
+	// Set default cleanpod policy to Running.
 	if tfjob.Spec.CleanPodPolicy == nil {
 		running := CleanPodPolicyRunning
 		tfjob.Spec.CleanPodPolicy = &running

pkg/apis/tensorflow/v1alpha2/install.go

@@ -0,0 +1,22 @@
+package v1alpha2
+
+import (
+	"k8s.io/apimachinery/pkg/apimachinery/announced"
+	"k8s.io/apimachinery/pkg/apimachinery/registered"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/util/sets"
+)
+
+// Install registers the API group and adds type to a schema
+func Install(groupFactoryRegister announced.APIGroupFactoryRegistry, registry *registered.APIRegistrationManager, scheme *runtime.Scheme) error {
+	return announced.NewGroupMetaFactory(
+		&announced.GroupMetaFactoryArgs{
+			GroupName:              GroupName,
+			RootScopedKinds:        sets.NewString(),
+			VersionPreferenceOrder: []string{SchemeGroupVersion.Version},
+		},
+		announced.VersionToSchemeFunc{
+			SchemeGroupVersion.Version: AddToScheme,
+		},
+	).Announce(groupFactoryRegister).RegisterAndEnable(registry, scheme)
+}