add publish not ready headless service to jobset

Makefile

@@ -297,4 +297,5 @@ golangci-lint: ## Download golangci-lint locally if necessary.
 GOTESTSUM = $(shell pwd)/bin/gotestsum
 .PHONY: gotestsum
 gotestsum: ## Download gotestsum locally if necessary.
-	@GOBIN=$(PROJECT_DIR)/bin GO111MODULE=on $(GO_CMD) install gotest.tools/gotestsum@v1.8.2
+	@GOBIN=$(PROJECT_DIR)/bin GO111MODULE=on $(GO_CMD) install gotest.tools/gotestsum@v1.8.2
+

api/jobset/v1alpha2/jobset_types.go

@@ -198,6 +198,12 @@ type Network struct {
 	// Defaults to <jobSet.name> if not set.
 	// +optional
 	Subdomain string `json:"subdomain,omitempty"`
+
+	// PublishNotReadyAddresses indicates that any agent
+	// which deals with endpoints for this Service should disregard any indications of ready/not-ready.
+	// Defaults to true if not set
+	// +optional
+	PublishNotReadyAddresses *bool `json:"publishNotReadyAddresses,omitempty"`
 }
 
 // Operator defines the target of a SuccessPolicy or FailurePolicy.

config/components/crd/bases/jobset.x-k8s.io_jobsets.yaml

@@ -85,6 +85,12 @@ spec:
                       Pods will be reachable using the fully qualified pod hostname:
                       <jobSet.name>-<spec.replicatedJob.name>-<job-index>-<pod-index>.<subdomain>
                     type: boolean
+                  publishNotReadyAddresses:
+                    description: |-
+                      PublishNotReadyAddresses indicates that any agent
+                      which deals with endpoints for this Service should disregard any indications of ready/not-ready.
+                      Defaults to true if not set
+                    type: boolean
                   subdomain:
                     description: |-
                       Subdomain is an explicit choice for a network subdomain name

hack/python-sdk/swagger.json

@@ -159,6 +159,10 @@
           "description": "EnableDNSHostnames allows pods to be reached via their hostnames. Pods will be reachable using the fully qualified pod hostname: \u003cjobSet.name\u003e-\u003cspec.replicatedJob.name\u003e-\u003cjob-index\u003e-\u003cpod-index\u003e.\u003csubdomain\u003e",
           "type": "boolean"
         },
+        "publishNotReadyAddresses": {
+          "description": "PublishNotReadyAddresses indicates that any agent which deals with endpoints for this Service should disregard any indications of ready/not-ready. Defaults to true if not set",
+          "type": "boolean"
+        },
         "subdomain": {
           "description": "Subdomain is an explicit choice for a network subdomain name When set, any replicated job in the set is added to this network. Defaults to \u003cjobSet.name\u003e if not set.",
           "type": "string"

pkg/controllers/jobset_controller.go

@@ -566,6 +566,7 @@ func (r *JobSetReconciler) createHeadlessSvcIfNecessary(ctx context.Context, js
 				Selector: map[string]string{
 					jobset.JobSetNameKey: js.Name,
 				},
+				PublishNotReadyAddresses: ptr.Deref(js.Spec.Network.PublishNotReadyAddresses, true),
 			},
 		}
 

pkg/util/testing/wrappers.go

@@ -121,12 +121,18 @@ func (j *JobSetWrapper) NetworkSubdomain(val string) *JobSetWrapper {
 	return j
 }
 
-// EnableDNSHostnames sets the value of ReplicatedJob.Network.EnableDNSHostnames.
+// EnableDNSHostnames sets the value of JobSet.Network.EnableDNSHostnames.
 func (j *JobSetWrapper) EnableDNSHostnames(val bool) *JobSetWrapper {
 	j.JobSet.Spec.Network.EnableDNSHostnames = ptr.To(val)
 	return j
 }
 
+// PublishNotReadyAddresses sets the value of JobSet.Network.PublishNotReadyAddresses.
+func (j *JobSetWrapper) PublishNotReadyAddresses(val bool) *JobSetWrapper {
+	j.JobSet.Spec.Network.PublishNotReadyAddresses = ptr.To(val)
+	return j
+}
+
 // TTLSecondsAfterFinished sets the value of JobSet.Spec.TTLSecondsAfterFinished
 func (j *JobSetWrapper) TTLSecondsAfterFinished(seconds int32) *JobSetWrapper {
 	j.Spec.TTLSecondsAfterFinished = &seconds

pkg/webhooks/jobset_webhook.go

@@ -104,20 +104,23 @@ func (j *jobSetWebhook) Default(ctx context.Context, obj runtime.Object) error {
 		if js.Spec.ReplicatedJobs[i].Template.Spec.CompletionMode == nil {
 			js.Spec.ReplicatedJobs[i].Template.Spec.CompletionMode = completionModePtr(batchv1.IndexedCompletion)
 		}
-		// Enable DNS hostnames by default.
-		if js.Spec.Network == nil {
-			js.Spec.Network = &jobset.Network{}
-		}
-		if js.Spec.Network.EnableDNSHostnames == nil {
-			js.Spec.Network.EnableDNSHostnames = ptr.To(true)
-		}
-
 		// Default pod restart policy to OnFailure.
 		if js.Spec.ReplicatedJobs[i].Template.Spec.Template.Spec.RestartPolicy == "" {
 			js.Spec.ReplicatedJobs[i].Template.Spec.Template.Spec.RestartPolicy = corev1.RestartPolicyOnFailure
 		}
 	}
 
+	// Enable DNS hostnames by default.
+	if js.Spec.Network == nil {
+		js.Spec.Network = &jobset.Network{}
+	}
+	if js.Spec.Network.EnableDNSHostnames == nil {
+		js.Spec.Network.EnableDNSHostnames = ptr.To(true)
+	}
+	if js.Spec.Network.PublishNotReadyAddresses == nil {
+		js.Spec.Network.PublishNotReadyAddresses = ptr.To(true)
+	}
+
 	if js.Spec.ManagedBy == nil {
 		js.Spec.ManagedBy = ptr.To(jobset.JobSetControllerName)
 	}