Startup policy implementation

api/jobset/v1alpha2/jobset_types.go

@@ -51,6 +51,8 @@ const (
 	JobSetFailed JobSetConditionType = "Failed"
 	// JobSetSuspended means the job is suspended
 	JobSetSuspended JobSetConditionType = "Suspended"
+	// JobSetStartupPolicyCompleted means the StartupPolicy was complete
+	JobSetStartupPolicyCompleted JobSetConditionType = "StartupPolicyCompleted"
 )
 
 // JobSetSpec defines the desired state of JobSet
@@ -79,6 +81,10 @@ type JobSetSpec struct {
 	// +kubebuilder:validation:XValidation:rule="self == oldSelf",message="Value is immutable"
 	FailurePolicy *FailurePolicy `json:"failurePolicy,omitempty"`
 
+	// StartupPolicy, if set, configures in what order jobs must be started
+	// +kubebuilder:validation:XValidation:rule="self == oldSelf",message="Value is immutable"
+	StartupPolicy *StartupPolicy `json:"startupPolicy,omitempty"`
+
 	// Suspend suspends all running child Jobs when set to true.
 	Suspend *bool `json:"suspend,omitempty"`
 }
@@ -191,6 +197,27 @@ type SuccessPolicy struct {
 	TargetReplicatedJobs []string `json:"targetReplicatedJobs,omitempty"`
 }
 
+type StartupPolicyOptions string
+
+const (
+	// This is the default setting
+	// AnyOrder means that we will start the replicated jobs
+	// without any specific order.
+	AnyOrder StartupPolicyOptions = "AnyOrder"
+	// InOrder starts the replicated jobs in order
+	// that they are listed.
+	InOrder StartupPolicyOptions = "InOrder"
+)
+
+type StartupPolicy struct {
+	// StartupPolicyOrder determines the startup order of the ReplicatedJobs.
+	// AnyOrder means to start replicated jobs in any order.
+	// InOrder means to start them as they are listed in the JobSet. A ReplicatedJob is started only
+	// when all the jobs of the previous one are ready.
+	// +kubebuilder:validation:Enum=AnyOrder;InOrder
+	StartupPolicyOrder StartupPolicyOptions `json:"startupPolicyOrder"`
+}
+
 func init() {
 	SchemeBuilder.Register(&JobSet{}, &JobSetList{})
 }

api/jobset/v1alpha2/jobset_webhook.go

@@ -51,6 +51,9 @@ func (js *JobSet) Default() {
 	if js.Spec.SuccessPolicy == nil {
 		js.Spec.SuccessPolicy = &SuccessPolicy{Operator: OperatorAll}
 	}
+	if js.Spec.StartupPolicy == nil {
+		js.Spec.StartupPolicy = &StartupPolicy{StartupPolicyOrder: AnyOrder}
+	}
 	for i := range js.Spec.ReplicatedJobs {
 		// Default job completion mode to indexed.
 		if js.Spec.ReplicatedJobs[i].Template.Spec.CompletionMode == nil {

api/jobset/v1alpha2/jobset_webhook_test.go

@@ -30,6 +30,7 @@ var TestPodTemplate = corev1.PodTemplateSpec{
 
 func TestJobSetDefaulting(t *testing.T) {
 	defaultSuccessPolicy := &SuccessPolicy{Operator: OperatorAll}
+	defaultStartupPolicy := &StartupPolicy{StartupPolicyOrder: AnyOrder}
 	testCases := []struct {
 		name string
 		js   *JobSet
@@ -40,6 +41,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			js: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -55,6 +57,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			want: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -90,6 +93,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			want: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -109,6 +113,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			js: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					ReplicatedJobs: []ReplicatedJob{
 						{
 							Template: batchv1.JobTemplateSpec{
@@ -124,6 +129,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			want: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -143,6 +149,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			js: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(false)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -159,6 +166,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			want: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(false)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -178,6 +186,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			js: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -196,6 +205,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			want: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -219,6 +229,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			js: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -239,6 +250,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			want: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: defaultSuccessPolicy,
+					StartupPolicy: defaultStartupPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
@@ -261,7 +273,8 @@ func TestJobSetDefaulting(t *testing.T) {
 			name: "success policy unset",
 			js: &JobSet{
 				Spec: JobSetSpec{
-					Network: &Network{EnableDNSHostnames: ptr.To(true)},
+					StartupPolicy: defaultStartupPolicy,
+					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{
 							Template: batchv1.JobTemplateSpec{
@@ -280,6 +293,7 @@ func TestJobSetDefaulting(t *testing.T) {
 			},
 			want: &JobSet{
 				Spec: JobSetSpec{
+					StartupPolicy: defaultStartupPolicy,
 					SuccessPolicy: defaultSuccessPolicy,
 					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
@@ -301,12 +315,64 @@ func TestJobSetDefaulting(t *testing.T) {
 		},
 		{
 			name: "success policy operator set, replicatedJobNames unset",
+			js: &JobSet{
+				Spec: JobSetSpec{
+					SuccessPolicy: &SuccessPolicy{
+						Operator: OperatorAny,
+					},
+					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
+					StartupPolicy: defaultStartupPolicy,
+					ReplicatedJobs: []ReplicatedJob{
+						{
+							Template: batchv1.JobTemplateSpec{
+								Spec: batchv1.JobSpec{
+									Template: corev1.PodTemplateSpec{
+										Spec: corev1.PodSpec{
+											RestartPolicy: corev1.RestartPolicyAlways,
+										},
+									},
+									CompletionMode: completionModePtr(batchv1.IndexedCompletion),
+								},
+							},
+						},
+					},
+				},
+			},
+			want: &JobSet{
+				Spec: JobSetSpec{
+					SuccessPolicy: &SuccessPolicy{
+						Operator: OperatorAny,
+					},
+					StartupPolicy: defaultStartupPolicy,
+					Network:       &Network{EnableDNSHostnames: ptr.To(true)},
+					ReplicatedJobs: []ReplicatedJob{
+						{
+							Template: batchv1.JobTemplateSpec{
+								Spec: batchv1.JobSpec{
+									Template: corev1.PodTemplateSpec{
+										Spec: corev1.PodSpec{
+											RestartPolicy: corev1.RestartPolicyAlways,
+										},
+									},
+									CompletionMode: completionModePtr(batchv1.IndexedCompletion),
+								},
+							},
+						},
+					},
+				},
+			},
+		},
+		{
+			name: "startup policy order InOrder set",
 			js: &JobSet{
 				Spec: JobSetSpec{
 					SuccessPolicy: &SuccessPolicy{
 						Operator: OperatorAny,
 					},
 					Network: &Network{EnableDNSHostnames: ptr.To(true)},
+					StartupPolicy: &StartupPolicy{
+						StartupPolicyOrder: InOrder,
+					},
 					ReplicatedJobs: []ReplicatedJob{
 						{
 							Template: batchv1.JobTemplateSpec{
@@ -328,6 +394,9 @@ func TestJobSetDefaulting(t *testing.T) {
 					SuccessPolicy: &SuccessPolicy{
 						Operator: OperatorAny,
 					},
+					StartupPolicy: &StartupPolicy{
+						StartupPolicyOrder: InOrder,
+					},
 					Network: &Network{EnableDNSHostnames: ptr.To(true)},
 					ReplicatedJobs: []ReplicatedJob{
 						{