Add restart strategy

api/jobset/v1alpha2/jobset_types.go

@@ -307,13 +307,31 @@ type FailurePolicy struct {
 	// A restart is achieved by recreating all active child jobs.
 	MaxRestarts int32 `json:"maxRestarts,omitempty"`
 
+	// RestartStrategy defines the strategy to use when restarting the JobSet.
+	// Defaults to Recreate.
+	// +optional
+	// +kubebuilder:default=Recreate
+	RestartStrategy JobSetRestartStrategy `json:"restartStrategy,omitempty"`
+
 	// List of failure policy rules for this JobSet.
 	// For a given Job failure, the rules will be evaluated in order,
 	// and only the first matching rule will be executed.
 	// If no matching rule is found, the RestartJobSet action is applied.
 	Rules []FailurePolicyRule `json:"rules,omitempty"`
 }
 
+// +kubebuilder:validation:Enum=Recreate;BlockingRecreate
+type JobSetRestartStrategy string
+
+const (
+	// Recreate Jobs on a Job-by-Job basis.
+	Recreate JobSetRestartStrategy = "Recreate"
+
+	// BlockingRecreate ensures that all Jobs (and Pods) from a previous iteration are deleted before
+	// creating new Jobs.
+	BlockingRecreate JobSetRestartStrategy = "BlockingRecreate"
+)
+
 type SuccessPolicy struct {
 	// Operator determines either All or Any of the selected jobs should succeed to consider the JobSet successful
 	// +kubebuilder:validation:Enum=All;Any

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

@@ -97,6 +97,15 @@ spec:
                       A restart is achieved by recreating all active child jobs.
                     format: int32
                     type: integer
+                  restartStrategy:
+                    default: Recreate
+                    description: |-
+                      RestartStrategy defines the strategy to use when restarting the JobSet.
+                      Defaults to Recreate.
+                    enum:
+                    - Recreate
+                    - BlockingRecreate
+                    type: string
                   rules:
                     description: |-
                       List of failure policy rules for this JobSet.

go.mod

@@ -1,6 +1,6 @@
 module sigs.k8s.io/jobset
 
-go 1.23
+go 1.23.0
 
 require (
 	github.com/google/go-cmp v0.6.0

hack/python-sdk/swagger.json

@@ -39,6 +39,10 @@
           "type": "integer",
           "format": "int32"
         },
+        "restartStrategy": {
+          "description": "RestartStrategy defines the strategy to use when restarting the JobSet. Defaults to Recreate.",
+          "type": "string"
+        },
         "rules": {
           "description": "List of failure policy rules for this JobSet. For a given Job failure, the rules will be evaluated in order, and only the first matching rule will be executed. If no matching rule is found, the RestartJobSet action is applied.",
           "type": "array",

pkg/controllers/jobset_controller.go

@@ -63,8 +63,9 @@ type childJobs struct {
 	successful []*batchv1.Job
 	failed     []*batchv1.Job
 
-	// Jobs marked for deletion are mutually exclusive with the set of jobs in active, successful, and failed.
-	delete []*batchv1.Job
+	// Jobs from a previous restart (marked for deletion) are mutually exclusive
+	// with the set of jobs in active, successful, and failed.
+	previous []*batchv1.Job
 }
 
 // statusUpdateOpts tracks if a JobSet status update should be performed at the end of the reconciliation
@@ -169,8 +170,8 @@ func (r *JobSetReconciler) reconcile(ctx context.Context, js *jobset.JobSet, upd
 		return ctrl.Result{}, nil
 	}
 
-	// Delete any jobs marked for deletion.
-	if err := r.deleteJobs(ctx, ownedJobs.delete); err != nil {
+	// Delete all jobs from a previous restart that are marked for deletion.
+	if err := r.deleteJobs(ctx, ownedJobs.previous); err != nil {
 		log.Error(err, "deleting jobs")
 		return ctrl.Result{}, err
 	}
@@ -281,11 +282,11 @@ func (r *JobSetReconciler) getChildJobs(ctx context.Context, js *jobset.JobSet)
 		jobRestarts, err := strconv.Atoi(job.Labels[constants.RestartsKey])
 		if err != nil {
 			log.Error(err, fmt.Sprintf("invalid value for label %s, must be integer", constants.RestartsKey))
-			ownedJobs.delete = append(ownedJobs.delete, &childJobList.Items[i])
+			ownedJobs.previous = append(ownedJobs.previous, &childJobList.Items[i])
 			return nil, err
 		}
 		if int32(jobRestarts) < js.Status.Restarts {
-			ownedJobs.delete = append(ownedJobs.delete, &childJobList.Items[i])
+			ownedJobs.previous = append(ownedJobs.previous, &childJobList.Items[i])
 			continue
 		}
 
@@ -637,6 +638,13 @@ func executeSuccessPolicy(js *jobset.JobSet, ownedJobs *childJobs, updateStatusO
 
 func constructJobsFromTemplate(js *jobset.JobSet, rjob *jobset.ReplicatedJob, ownedJobs *childJobs) []*batchv1.Job {
 	var jobs []*batchv1.Job
+	// If the JobSet is using the BlockingRecreate failure policy, we should not create any new jobs until
+	// all the jobs slated for deletion (i.e. from the last restart index) have been deleted.
+	useBlockingRecreate := js.Spec.FailurePolicy != nil && js.Spec.FailurePolicy.RestartStrategy == jobset.BlockingRecreate
+	if len(ownedJobs.previous) > 0 && useBlockingRecreate {
+		return jobs
+	}
+
 	for jobIdx := 0; jobIdx < int(rjob.Replicas); jobIdx++ {
 		jobName := placement.GenJobName(js.Name, rjob.Name, jobIdx)
 		if create := shouldCreateJob(jobName, ownedJobs); !create {
@@ -700,7 +708,7 @@ func shouldCreateJob(jobName string, ownedJobs *childJobs) bool {
 	// TODO: maybe we can use a job map here so we can do O(1) lookups
 	// to check if the job already exists, rather than a linear scan
 	// through all the jobs owned by the jobset.
-	for _, job := range collections.Concat(ownedJobs.active, ownedJobs.successful, ownedJobs.failed, ownedJobs.delete) {
+	for _, job := range collections.Concat(ownedJobs.active, ownedJobs.successful, ownedJobs.failed, ownedJobs.previous) {
 		if jobName == job.Name {
 			return false
 		}

pkg/controllers/jobset_controller_test.go

@@ -260,7 +260,7 @@ func TestConstructJobsFromTemplate(t *testing.T) {
 					Replicas(2).
 					Obj()).Obj(),
 			ownedJobs: &childJobs{
-				delete: []*batchv1.Job{
+				previous: []*batchv1.Job{
 					testutils.MakeJob("test-jobset-replicated-job-0", ns).Obj(),
 				},
 			},
@@ -275,6 +275,23 @@ func TestConstructJobsFromTemplate(t *testing.T) {
 					Suspend(false).Obj(),
 			},
 		},
+		{
+			name: "job creation blocked until all previous jobs no longer exist",
+			js: testutils.MakeJobSet(jobSetName, ns).
+				FailurePolicy(&jobset.FailurePolicy{
+					RestartStrategy: jobset.BlockingRecreate,
+				}).
+				ReplicatedJob(testutils.MakeReplicatedJob(replicatedJobName).
+					Job(testutils.MakeJobTemplate(jobName, ns).Obj()).
+					Replicas(2).
+					Obj()).Obj(),
+			ownedJobs: &childJobs{
+				previous: []*batchv1.Job{
+					testutils.MakeJob("test-jobset-replicated-job-0", ns).Obj(),
+				},
+			},
+			want: nil,
+		},
 		{
 			name: "multiple replicated jobs",
 			js: testutils.MakeJobSet(jobSetName, ns).

sdk/python/docs/JobsetV1alpha2FailurePolicy.md

@@ -4,6 +4,7 @@
 Name | Type | Description | Notes
 ------------ | ------------- | ------------- | -------------
 **max_restarts** | **int** | MaxRestarts defines the limit on the number of JobSet restarts. A restart is achieved by recreating all active child jobs. | [optional] 
+**restart_strategy** | **str** | RestartStrategy defines the strategy to use when restarting the JobSet. Defaults to Recreate. | [optional] 
 **rules** | [**list[JobsetV1alpha2FailurePolicyRule]**](JobsetV1alpha2FailurePolicyRule.md) | List of failure policy rules for this JobSet. For a given Job failure, the rules will be evaluated in order, and only the first matching rule will be executed. If no matching rule is found, the RestartJobSet action is applied. | [optional] 
 
 [[Back to Model list]](../README.md#documentation-for-models) [[Back to API list]](../README.md#documentation-for-api-endpoints) [[Back to README]](../README.md)

sdk/python/jobset/models/jobset_v1alpha2_failure_policy.py

@@ -34,26 +34,31 @@ class JobsetV1alpha2FailurePolicy(object):
     """
     openapi_types = {
         'max_restarts': 'int',
+        'restart_strategy': 'str',
         'rules': 'list[JobsetV1alpha2FailurePolicyRule]'
     }
 
     attribute_map = {
         'max_restarts': 'maxRestarts',
+        'restart_strategy': 'restartStrategy',
         'rules': 'rules'
     }
 
-    def __init__(self, max_restarts=None, rules=None, local_vars_configuration=None):  # noqa: E501
+    def __init__(self, max_restarts=None, restart_strategy=None, rules=None, local_vars_configuration=None):  # noqa: E501
         """JobsetV1alpha2FailurePolicy - a model defined in OpenAPI"""  # noqa: E501
         if local_vars_configuration is None:
             local_vars_configuration = Configuration()
         self.local_vars_configuration = local_vars_configuration
 
         self._max_restarts = None
+        self._restart_strategy = None
         self._rules = None
         self.discriminator = None
 
         if max_restarts is not None:
             self.max_restarts = max_restarts
+        if restart_strategy is not None:
+            self.restart_strategy = restart_strategy
         if rules is not None:
             self.rules = rules
 
@@ -80,6 +85,29 @@ def max_restarts(self, max_restarts):
 
         self._max_restarts = max_restarts
 
+    @property
+    def restart_strategy(self):
+        """Gets the restart_strategy of this JobsetV1alpha2FailurePolicy.  # noqa: E501
+
+        RestartStrategy defines the strategy to use when restarting the JobSet. Defaults to Recreate.  # noqa: E501
+
+        :return: The restart_strategy of this JobsetV1alpha2FailurePolicy.  # noqa: E501
+        :rtype: str
+        """
+        return self._restart_strategy
+
+    @restart_strategy.setter
+    def restart_strategy(self, restart_strategy):
+        """Sets the restart_strategy of this JobsetV1alpha2FailurePolicy.
+
+        RestartStrategy defines the strategy to use when restarting the JobSet. Defaults to Recreate.  # noqa: E501
+
+        :param restart_strategy: The restart_strategy of this JobsetV1alpha2FailurePolicy.  # noqa: E501
+        :type: str
+        """
+
+        self._restart_strategy = restart_strategy
+
     @property
     def rules(self):
         """Gets the rules of this JobsetV1alpha2FailurePolicy.  # noqa: E501

sdk/python/test/test_jobset_v1alpha2_failure_policy.py

@@ -39,6 +39,7 @@ def make_instance(self, include_optional):
         if include_optional :
             return JobsetV1alpha2FailurePolicy(
                 max_restarts = 56, 
+                restart_strategy = '0', 
                 rules = [
                     jobset.models.jobset_v1alpha2_failure_policy_rule.JobsetV1alpha2FailurePolicyRule(
                         action = '0', 

sdk/python/test/test_jobset_v1alpha2_job_set.py

@@ -48,6 +48,7 @@ def make_instance(self, include_optional):
                         replicated_job = '0', ), 
                     failure_policy = jobset.models.jobset_v1alpha2_failure_policy.JobsetV1alpha2FailurePolicy(
                         max_restarts = 56, 
+                        restart_strategy = '0', 
                         rules = [
                             jobset.models.jobset_v1alpha2_failure_policy_rule.JobsetV1alpha2FailurePolicyRule(
                                 action = '0', 

sdk/python/test/test_jobset_v1alpha2_job_set_list.py

@@ -51,6 +51,7 @@ def make_instance(self, include_optional):
                                 replicated_job = '0', ), 
                             failure_policy = jobset.models.jobset_v1alpha2_failure_policy.JobsetV1alpha2FailurePolicy(
                                 max_restarts = 56, 
+                                restart_strategy = '0', 
                                 rules = [
                                     jobset.models.jobset_v1alpha2_failure_policy_rule.JobsetV1alpha2FailurePolicyRule(
                                         action = '0', 
@@ -113,6 +114,7 @@ def make_instance(self, include_optional):
                                 replicated_job = '0', ), 
                             failure_policy = jobset.models.jobset_v1alpha2_failure_policy.JobsetV1alpha2FailurePolicy(
                                 max_restarts = 56, 
+                                restart_strategy = '0', 
                                 rules = [
                                     jobset.models.jobset_v1alpha2_failure_policy_rule.JobsetV1alpha2FailurePolicyRule(
                                         action = '0', 

sdk/python/test/test_jobset_v1alpha2_job_set_spec.py

@@ -44,6 +44,7 @@ def make_instance(self, include_optional):
                     replicated_job = '0', ), 
                 failure_policy = jobset.models.jobset_v1alpha2_failure_policy.JobsetV1alpha2FailurePolicy(
                     max_restarts = 56, 
+                    restart_strategy = '0', 
                     rules = [
                         jobset.models.jobset_v1alpha2_failure_policy_rule.JobsetV1alpha2FailurePolicyRule(
                             action = '0', 

site/content/en/docs/tasks/_index.md

@@ -23,7 +23,7 @@ Here we have some simple examples demonstrating core JobSet features.
 Success Policy allows one to specify when to mark a JobSet as success.  
 This example showcases an example of using the success policy to mark the JobSet as successful if the worker replicated job completes.
 
-- [Failure Policy with Max Restarts](https://github.com/kubernetes-sigs/jobset/blob/release-0.5/examples/simple/max-restarts.yaml) demonstrates an example of utilizing `failurePolicy`. Failure Policy allows one to control how many restarts a JobSet can do before declaring the JobSet as failed.
+- [Failure Policy](https://github.com/kubernetes-sigs/jobset/blob/release-0.5/examples/simple/failure-policy.yaml) demonstrates an example of utilizing `failurePolicy`. Failure Policy allows one to control how many restarts a JobSet can do before declaring the JobSet as failed. The strategy used when restarting can also be specified (i.e. whether to first delete all Jobs, or recreate on a one-by-one basis).
 
 - [Exclusive Job Placement](https://github.com/kubernetes-sigs/jobset/blob/release-0.5/examples/simple/exclusive-placement.yaml) demonstrates how you can configure a JobSet to have a 1:1 mapping between each child Job and a particular topology domain, such as a datacenter rack or zone. This means that all the pods belonging to a child job will be colocated in the same topology domain, while pods from other jobs will not be allowed to run within this domain. This gives the child job exclusive access to computer resources in this domain.
 

site/static/examples/simple/max-restarts.yaml → site/static/examples/simple/failure-policy.yaml

@@ -1,10 +1,13 @@
 apiVersion: jobset.x-k8s.io/v1alpha2
 kind: JobSet
 metadata:
-  name: max-restarts
+  name: failure-policy
 spec:
-  # On failure, restart all jobs up to 3 times.
   failurePolicy:
+    # Wait for all Jobs to be fully deleted before recreating any.
+    # Defaults to "Recreate" which restarts Jobs individually.
+    restartStrategy: BlockingRecreate
+    # On failure, restart all jobs up to 3 times.
     maxRestarts: 3
   replicatedJobs:
   - name: leader