Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add a field SchedulerName to TFJob for specifying a scheduler #408

Merged
merged 1 commit into from
Mar 7, 2018
Merged

Add a field SchedulerName to TFJob for specifying a scheduler #408

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
3 changes: 3 additions & 0 deletions pkg/apis/tensorflow/v1alpha1/types.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,9 @@ type TFJobSpec struct {

// TerminationPolicy specifies the condition that the tfjob should be considered finished.
TerminationPolicy *TerminationPolicySpec `json:"terminationPolicy,omitempty"`

// SchedulerName specifies the name of scheduler which should handle the TFJob
SchedulerName string `json:"schedulerName,omitempty"`
}

type TerminationPolicySpec struct {
Expand Down
2 changes: 2 additions & 0 deletions pkg/trainer/replicas.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,6 +171,8 @@ func (s *TFReplicaSet) CreatePodWithIndex(index int32) (*v1.Pod, error) {
Spec: *s.Spec.Template.Spec.DeepCopy(),
}

pod.Spec.SchedulerName = s.Job.SchedulerName()

// Configure the TFCONFIG environment variable.
tfConfig := TFConfig{
Cluster: s.Job.ClusterSpec(),
Expand Down
8 changes: 8 additions & 0 deletions pkg/trainer/replicas_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@ import (
"encoding/json"
"fmt"
"reflect"
"strings"
"testing"
"time"

Expand All @@ -44,6 +45,8 @@ var (
func TestTFReplicaSet(t *testing.T) {
clientSet := fake.NewSimpleClientset()

testSchedulerName := "test-scheduler"

jobSpec := &tfv1alpha1.TFJob{
ObjectMeta: meta_v1.ObjectMeta{
Name: "some-job",
Expand All @@ -67,6 +70,7 @@ func TestTFReplicaSet(t *testing.T) {
TFReplicaType: tfv1alpha1.PS,
},
},
SchedulerName: testSchedulerName,
},
}

Expand Down Expand Up @@ -169,6 +173,10 @@ func TestTFReplicaSet(t *testing.T) {
t.Fatalf("Expected 1 environment variable got %v", len(c.Env))
}

if strings.Compare(p.Spec.SchedulerName, testSchedulerName) != 0 {
t.Fatalf("p.Spec.Template.Spec.SchedulerName; Got %v; want %v", p.Spec.SchedulerName, testSchedulerName)
}

actualTFConfig := &TFConfig{}
if err := json.Unmarshal([]byte(c.Env[0].Value), actualTFConfig); err != nil {
t.Fatalf("Could not unmarshal TFConfig %v", err)
Expand Down
4 changes: 4 additions & 0 deletions pkg/trainer/training.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -412,3 +412,7 @@ func (j *TrainingJob) name() string {
func (j *TrainingJob) fullname() string {
return j.job.ObjectMeta.GetNamespace() + ":" + j.job.ObjectMeta.GetName()
}

func (j *TrainingJob) SchedulerName() string {
return j.job.Spec.SchedulerName
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we need a special Getter for SchedulerName?
We can just get it directly form j.job.Spec.SchedulerName.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Because TrainingJob.job is a private member so TrainingJob should provide the accessor method.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, the current structure is a little complex:

  • We have TFJob for API object Spec/Status
  • Then create TrainingJob and TFReplicaSet which keep many cache of TFJob

And PodSpec already have a field SchedulerName, can we reuse it?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

PodSpec.SchedulerName isn't suitable for the purpose. See the discussion start with this message #398 (comment)

}