Modify XGBoostJob example for the new Controller

examples/v1beta1/xgboost-lightgbm.yaml

@@ -2,7 +2,7 @@ apiVersion: kubeflow.org/v1beta1
 kind: Experiment
 metadata:
   namespace: kubeflow
-  name: xgboost-lightgbm
+  name: xgboostjob-lightgbm
 spec:
   objective:
     type: maximize
@@ -35,11 +35,7 @@ spec:
         max: "60"
         step: "1"
   trialTemplate:
-    primaryPodLabels:
-      job-role: master
-    primaryContainerName: xgboostjob
-    successCondition: status.conditions.#(type=="Succeeded")#|#(status=="True")#
-    failureCondition: status.conditions.#(type=="Failed")#|#(status=="True")#
+    primaryContainerName: xgboost
     trialParameters:
       - name: learningRate
         description: Learning rate for the training model
@@ -48,8 +44,7 @@ spec:
         description: Number of leaves for one tree
         reference: num-leaves
     trialSpec:
-      # TODO (andreyvelich): Change to kubeflow.org/v1 once all-in-one operator is finished.
-      apiVersion: xgboostjob.kubeflow.org/v1
+      apiVersion: kubeflow.org/v1
       kind: XGBoostJob
       spec:
         xgbReplicaSpecs:
@@ -59,7 +54,7 @@ spec:
             template:
               spec:
                 containers:
-                  - name: xgboostjob
+                  - name: xgboost
                     image: docker.io/kubeflowkatib/xgboost-lightgbm:1.0
                     ports:
                       - containerPort: 9991
@@ -93,7 +88,7 @@ spec:
             template:
               spec:
                 containers:
-                  - name: xgboostjob
+                  - name: xgboost
                     image: docker.io/kubeflowkatib/xgboost-lightgbm:1.0
                     ports:
                       - containerPort: 9991

manifests/v1beta1/components/controller/controller.yaml

@@ -29,8 +29,7 @@ spec:
             - "--trial-resources=TFJob.v1.kubeflow.org"
             - "--trial-resources=PyTorchJob.v1.kubeflow.org"
             - "--trial-resources=MPIJob.v1.kubeflow.org"
-            # TODO (andreyvelich): Change to v1.kubeflow.org once all-in-one operator is finished.
-            - "--trial-resources=XGBoostJob.v1.xgboostjob.kubeflow.org"
+            - "--trial-resources=XGBoostJob.v1.kubeflow.org"
           ports:
             - containerPort: 8443
               name: webhook

manifests/v1beta1/components/controller/rbac.yaml

@@ -53,12 +53,6 @@ rules:
       - tfjobs
       - pytorchjobs
       - mpijobs
-    verbs:
-      - "*"
-  # TODO (andreyvelich): Move to "apiGroup: kubeflow.org" once all-in-one operator is finished.
-  - apiGroups:
-      - xgboostjob.kubeflow.org
-    resources:
       - xgboostjobs
     verbs:
       - "*"

pkg/apis/controller/experiments/v1beta1/experiment_defaults.go

@@ -96,7 +96,7 @@ func (e *Experiment) setDefaultObjective() {
 func (e *Experiment) setDefaultTrialTemplate() {
 	t := e.Spec.TrialTemplate
 
-	// Set default values for Job, TFJob and PyTorchJob if TrialSpec is not nil
+	// Set default values for Job and Kubeflow Training Job if TrialSpec is not nil
 	if t != nil && t.TrialSource.TrialSpec != nil {
 		jobKind := t.TrialSource.TrialSpec.GetKind()
 		if jobKind == consts.JobKindJob {
@@ -106,7 +106,7 @@ func (e *Experiment) setDefaultTrialTemplate() {
 			if t.FailureCondition == "" {
 				t.FailureCondition = DefaultJobFailureCondition
 			}
-		} else if jobKind == consts.JobKindTF || jobKind == consts.JobKindPyTorch {
+		} else if jobKind == consts.JobKindTF || jobKind == consts.JobKindPyTorch || jobKind == consts.JobKindXGBoost {
 			if t.SuccessCondition == "" {
 				t.SuccessCondition = DefaultKubeflowJobSuccessCondition
 			}

pkg/controller.v1beta1/consts/const.go

@@ -134,6 +134,8 @@ const (
 	JobKindTF = "TFJob"
 	// JobKindPyTorch is the kind of PyTorchJob.
 	JobKindPyTorch = "PyTorchJob"
+	// JobKindXGBoost is the kind of XGBoostJob.
+	JobKindXGBoost = "XGBoostJob"
 
 	// AnnotationIstioSidecarInjectName is the annotation of Istio Sidecar
 	AnnotationIstioSidecarInjectName = "sidecar.istio.io/inject"