Ensure that PyTorch CRD exists

Gopkg.toml

@@ -52,6 +52,10 @@ required = [
   name = "k8s.io/apimachinery"
   branch = "release-1.10"
 
+[[constraint]]
+  name = "k8s.io/apiextensions-apiserver"
+  branch = "release-1.10"
+
 [[constraint]]
   name = "k8s.io/code-generator"
   # We can not use master since the generated package name is changed from tensorflow to kubeflow.

cmd/pytorch-operator.v1beta1/app/options/options.go

@@ -54,5 +54,5 @@ func (s *ServerOption) AddFlags(fs *flag.FlagSet) {
 
 	fs.BoolVar(&s.JSONLogFormat, "json-log-format", true,
 		"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.")
+	fs.BoolVar(&s.EnableGangScheduling, "enable-gang-scheduling", false, "Set true to enable gang scheduling by kube-batch.")
 }

cmd/pytorch-operator.v1beta1/app/server.go

@@ -21,6 +21,7 @@ import (
 
 	log "github.com/sirupsen/logrus"
 	"k8s.io/api/core/v1"
+	crdclient "k8s.io/apiextensions-apiserver/pkg/client/clientset/clientset"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	kubeinformers "k8s.io/client-go/informers"
 	kubeclientset "k8s.io/client-go/kubernetes"
@@ -154,6 +155,15 @@ func Run(opt *options.ServerOption) error {
 }
 
 func createClientSets(config *restclientset.Config) (kubeclientset.Interface, kubeclientset.Interface, jobclientset.Interface, error) {
+
+	crdClient, err := crdclient.NewForConfig(config)
+
+	if err != nil {
+		return nil, nil, nil, err
+	}
+
+	checkCRDExists(crdClient, v1beta1.PytorchCRD)
+
 	kubeClientSet, err := kubeclientset.NewForConfig(restclientset.AddUserAgent(config, "pytorch-operator"))
 	if err != nil {
 		return nil, nil, nil, err
@@ -171,3 +181,12 @@ func createClientSets(config *restclientset.Config) (kubeclientset.Interface, ku
 
 	return kubeClientSet, leaderElectionClientSet, jobClientSet, nil
 }
+
+func checkCRDExists(clientset crdclient.Interface, crdName string) {
+	_, err := clientset.ApiextensionsV1beta1().CustomResourceDefinitions().Get(crdName, metav1.GetOptions{})
+
+	if err != nil {
+		log.Error(err)
+		os.Exit(1)
+	}
+}

pkg/apis/pytorch/v1beta1/register.go

@@ -39,6 +39,8 @@ const (
 	Plural = "pytorchjobs"
 	// Singular is the singular for pytorchJob.
 	Singular = "pytorchjob"
+	// PytorchCRD is the CRD name for PytorchJob.
+	PytorchCRD = "pytorchjobs.kubeflow.org"
 )
 
 var (

pkg/controller.v1beta1/pytorch/job.go

@@ -156,7 +156,7 @@ func (pc *PyTorchController) cleanupPyTorchJob(job *v1beta1.PyTorchJob) error {
 	return nil
 }
 
-// deletePyTorchJob delets the given PyTorchJob.
+// deletePyTorchJob deletes the given PyTorchJob.
 func (pc *PyTorchController) deletePyTorchJob(job *v1beta1.PyTorchJob) error {
 	return pc.jobClientSet.KubeflowV1beta1().PyTorchJobs(job.Namespace).Delete(job.Name, &metav1.DeleteOptions{})
 }

pkg/controller.v1beta1/pytorch/pod.go

@@ -35,7 +35,7 @@ import (
 
 const (
 	// podTemplateRestartPolicyReason is the warning reason when the restart
-	// policy is setted in pod template.
+	// policy is set in pod template.
 	podTemplateRestartPolicyReason = "SettedPodTemplateRestartPolicy"
 	exitedWithCodeReason           = "ExitedWithCode"
 )