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

koord-descheduler: add arbitration to migration controller #1651

Merged
merged 11 commits into from
Sep 26, 2023
Merged

koord-descheduler: add arbitration to migration controller #1651

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
bda37bf
koord-descheduler: add arbitration to migration controller
baowj-678 Sep 13, 2023
3ee4346
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 18, 2023
29cf06b
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 22, 2023
d7c6f09
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 22, 2023
cbb098d
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 22, 2023
3fc715a
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 23, 2023
b40624f
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 23, 2023
2aaae4b
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 25, 2023
0a30c7b
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 25, 2023
d86da54
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 26, 2023
5178116
koord-descheduler: add arbitration to migration controller.
baowj-678 Sep 26, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
71 changes: 22 additions & 49 deletions pkg/descheduler/controllers/migration/controller.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,13 +47,13 @@ import (
sev1alpha1 "github.com/koordinator-sh/koordinator/apis/scheduling/v1alpha1"
deschedulerconfig "github.com/koordinator-sh/koordinator/pkg/descheduler/apis/config"
"github.com/koordinator-sh/koordinator/pkg/descheduler/apis/config/validation"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/arbitrator"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/controllerfinder"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/evictor"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/reservation"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/util"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/names"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/options"
evictionsutil "github.com/koordinator-sh/koordinator/pkg/descheduler/evictions"
"github.com/koordinator-sh/koordinator/pkg/descheduler/framework"
utilclient "github.com/koordinator-sh/koordinator/pkg/util/client"
)
Expand Down Expand Up @@ -108,7 +108,23 @@ func New(args runtime.Object, handle framework.Handle) (framework.Plugin, error)
return nil, err
}

if err = c.Watch(&source.Kind{Type: &sev1alpha1.PodMigrationJob{}}, &handler.EnqueueRequestForObject{}, &predicate.Funcs{
// New Arbitrator
eahydra marked this conversation as resolved.
Show resolved Hide resolved
var eventHandler handler.EventHandler = &handler.EnqueueRequestForObject{}
if controllerArgs.ArbitrationArgs.Enabled {
eahydra marked this conversation as resolved.
Show resolved Hide resolved
a, err := arbitrator.New(controllerArgs.ArbitrationArgs, arbitrator.Options{
Client: r.Client,
EventRecorder: r.eventRecorder,
RetryableFilter: r.retryablePodFilter,
NonRetryableFilter: r.nonRetryablePodFilter,
Manager: options.Manager,
})
if err != nil {
klog.ErrorS(err, "failed to New Arbitrator")
eahydra marked this conversation as resolved.
Show resolved Hide resolved
}
eventHandler = arbitrator.NewHandler(a, r.Client)
}

if err = c.Watch(&source.Kind{Type: &sev1alpha1.PodMigrationJob{}}, eventHandler, &predicate.Funcs{
DeleteFunc: func(event event.DeleteEvent) bool {
job := event.Object.(*sev1alpha1.PodMigrationJob)
r.assumedCache.delete(job)
Expand Down Expand Up @@ -421,18 +437,11 @@ func (r *Reconciler) preparePendingJob(ctx context.Context, job *sev1alpha1.PodM
}

markPodPrepareMigrating(pod)
if !evictionsutil.HaveEvictAnnotation(job) {
if aborted, err := r.abortJobIfNonRetryablePodFilterFailed(ctx, pod, job); aborted || err != nil {
if err == nil {
err = fmt.Errorf("abort job since failed to non-retryable Pod filter")
}
return reconcile.Result{}, err
}
if requeue, err := r.requeueJobIfRetryablePodFilterFailed(ctx, pod, job); requeue || err != nil {
return reconcile.Result{RequeueAfter: defaultRequeueAfter}, err
}
}

// delete passed arbitration annotation
if job.Annotations != nil {
delete(job.Annotations, arbitrator.AnnotationPassedArbitration)
eahydra marked this conversation as resolved.
Show resolved Hide resolved
}
job.Status.Phase = sev1alpha1.PodMigrationJobRunning
err = r.Client.Status().Update(ctx, job)
return reconcile.Result{}, err
Expand Down Expand Up @@ -487,42 +496,6 @@ func (r *Reconciler) abortJobIfTimeout(ctx context.Context, job *sev1alpha1.PodM
return true, err
}

func (r *Reconciler) requeueJobIfRetryablePodFilterFailed(ctx context.Context, pod *corev1.Pod, job *sev1alpha1.PodMigrationJob) (bool, error) {
if r.retryablePodFilter == nil {
return false, nil
}

if pod != nil {
if !r.retryablePodFilter(pod) {
r.eventRecorder.Eventf(job, nil, corev1.EventTypeWarning, "Requeue", "Migrating", "Failed to retriable filter")
return true, nil
}
}

return false, nil
}

func (r *Reconciler) abortJobIfNonRetryablePodFilterFailed(ctx context.Context, pod *corev1.Pod, job *sev1alpha1.PodMigrationJob) (bool, error) {
if r.nonRetryablePodFilter == nil {
return false, nil
}

if pod != nil {
if !r.nonRetryablePodFilter(pod) {
job.Status.Phase = sev1alpha1.PodMigrationJobFailed
job.Status.Reason = sev1alpha1.PodMigrationJobReasonForbiddenMigratePod
job.Status.Message = fmt.Sprintf("Pod %q is forbidden to migrate because it does not meet the requirements", klog.KObj(pod))
err := r.Status().Update(ctx, job)
if err == nil {
r.eventRecorder.Eventf(job, nil, corev1.EventTypeWarning, sev1alpha1.PodMigrationJobReasonForbiddenMigratePod, "Migrating", job.Status.Message)
}
return true, err
}
}

return false, nil
}

func (r *Reconciler) abortJobByInvalidPodRef(ctx context.Context, job *sev1alpha1.PodMigrationJob) error {
job.Status.Phase = sev1alpha1.PodMigrationJobFailed
job.Status.Reason = "InvalidPodRef"
Expand Down
115 changes: 5 additions & 110 deletions pkg/descheduler/controllers/migration/controller_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@ import (
sev1alpha1 "github.com/koordinator-sh/koordinator/apis/scheduling/v1alpha1"
deschedulerconfig "github.com/koordinator-sh/koordinator/pkg/descheduler/apis/config"
"github.com/koordinator-sh/koordinator/pkg/descheduler/apis/config/v1alpha2"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/arbitrator"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/controllerfinder"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/reservation"
"github.com/koordinator-sh/koordinator/pkg/descheduler/controllers/migration/util"
Expand Down Expand Up @@ -1441,7 +1442,6 @@ func TestEvict(t *testing.T) {
Phase: corev1.PodRunning,
},
}
assert.True(t, reconciler.Filter(pod))

assert.True(t, reconciler.Evict(context.TODO(), pod, framework.EvictOptions{}))
var jobList sev1alpha1.PodMigrationJobList
Expand Down Expand Up @@ -1509,115 +1509,6 @@ func TestAbortJobIfReserveOnSameNode(t *testing.T) {
assert.Equal(t, sev1alpha1.PodMigrationJobReasonForbiddenMigratePod, job.Status.Reason)
}

func TestRequeueJobIfRetryablePodFilterFailed(t *testing.T) {
reconciler := newTestReconciler()
enter := false
reconciler.retryablePodFilter = func(pod *corev1.Pod) bool {
enter = true
assert.True(t, isPodPrepareMigrating(pod))
return false
}

job := &sev1alpha1.PodMigrationJob{
ObjectMeta: metav1.ObjectMeta{
Name: "test",
CreationTimestamp: metav1.Time{Time: time.Now()},
},
Spec: sev1alpha1.PodMigrationJobSpec{
PodRef: &corev1.ObjectReference{
Namespace: "default",
Name: "test-pod",
},
},
}
assert.Nil(t, reconciler.Client.Create(context.TODO(), job))
pod := &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{
Namespace: "default",
Name: "test-pod",
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: "apps/v1",
Controller: pointer.Bool(true),
Kind: "StatefulSet",
Name: "test",
UID: "2f96233d-a6b9-4981-b594-7c90c987aed9",
},
},
},
Spec: corev1.PodSpec{
SchedulerName: "koord-scheduler",
},
Status: corev1.PodStatus{
Phase: corev1.PodRunning,
},
}
assert.Nil(t, reconciler.Client.Create(context.TODO(), pod))

result, err := reconciler.doMigrate(context.TODO(), job)
assert.True(t, enter)
assert.NoError(t, err)
assert.True(t, result.RequeueAfter != 0)
assert.NoError(t, reconciler.Client.Get(context.TODO(), types.NamespacedName{Name: job.Name}, job))
assert.Equal(t, sev1alpha1.PodMigrationJobPhase(""), job.Status.Phase)
assert.Equal(t, "", job.Status.Reason)
}

func TestAbortJobIfNonRetryablePodFilterFailed(t *testing.T) {
reconciler := newTestReconciler()
enter := false
reconciler.nonRetryablePodFilter = func(pod *corev1.Pod) bool {
enter = true
assert.True(t, isPodPrepareMigrating(pod))
return false
}

job := &sev1alpha1.PodMigrationJob{
ObjectMeta: metav1.ObjectMeta{
Name: "test",
CreationTimestamp: metav1.Time{Time: time.Now()},
},
Spec: sev1alpha1.PodMigrationJobSpec{
PodRef: &corev1.ObjectReference{
Namespace: "default",
Name: "test-pod",
},
},
}
assert.Nil(t, reconciler.Client.Create(context.TODO(), job))
pod := &corev1.Pod{
ObjectMeta: metav1.ObjectMeta{
Namespace: "default",
Name: "test-pod",
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: "apps/v1",
Controller: pointer.Bool(true),
Kind: "StatefulSet",
Name: "test",
UID: "2f96233d-a6b9-4981-b594-7c90c987aed9",
},
},
},
Spec: corev1.PodSpec{
SchedulerName: "koord-scheduler",
},
Status: corev1.PodStatus{
Phase: corev1.PodRunning,
},
}
assert.Nil(t, reconciler.Client.Create(context.TODO(), pod))

result, err := reconciler.doMigrate(context.TODO(), job)
assert.True(t, enter)
assert.NotNil(t, err)
assert.Equal(t, reconcile.Result{}, result)

assert.NoError(t, reconciler.Client.Get(context.TODO(), types.NamespacedName{Name: job.Name}, job))
assert.Equal(t, sev1alpha1.PodMigrationJobFailed, job.Status.Phase)
assert.Equal(t, sev1alpha1.PodMigrationJobReasonForbiddenMigratePod, job.Status.Reason)
}

func TestFilterExistingMigrationJob(t *testing.T) {
reconciler := newTestReconciler()

Expand Down Expand Up @@ -1760,6 +1651,7 @@ func TestFilterMaxMigratingPerNode(t *testing.T) {
ObjectMeta: metav1.ObjectMeta{
Name: fmt.Sprintf("test-%d", i),
CreationTimestamp: metav1.Time{Time: time.Now()},
Annotations: map[string]string{arbitrator.AnnotationPassedArbitration: "true"},
},
Spec: sev1alpha1.PodMigrationJobSpec{
PodRef: &corev1.ObjectReference{
Expand Down Expand Up @@ -1909,6 +1801,7 @@ func TestFilterMaxMigratingPerNamespace(t *testing.T) {
ObjectMeta: metav1.ObjectMeta{
Name: fmt.Sprintf("test-%d", i),
CreationTimestamp: metav1.Time{Time: time.Now()},
Annotations: map[string]string{arbitrator.AnnotationPassedArbitration: "true"},
},
Spec: sev1alpha1.PodMigrationJobSpec{
PodRef: &corev1.ObjectReference{
Expand Down Expand Up @@ -2096,6 +1989,7 @@ func TestFilterMaxMigratingPerWorkload(t *testing.T) {
ObjectMeta: metav1.ObjectMeta{
Name: fmt.Sprintf("test-%d", i),
CreationTimestamp: metav1.Time{Time: time.Now()},
Annotations: map[string]string{arbitrator.AnnotationPassedArbitration: "true"},
},
Spec: sev1alpha1.PodMigrationJobSpec{
PodRef: &corev1.ObjectReference{
Expand Down Expand Up @@ -2326,6 +2220,7 @@ func TestFilterMaxUnavailablePerWorkload(t *testing.T) {
ObjectMeta: metav1.ObjectMeta{
Name: fmt.Sprintf("test-%d", i),
CreationTimestamp: metav1.Time{Time: time.Now()},
Annotations: map[string]string{arbitrator.AnnotationPassedArbitration: "true"},
},
Spec: sev1alpha1.PodMigrationJobSpec{
PodRef: &corev1.ObjectReference{
Expand Down
5 changes: 0 additions & 5 deletions pkg/descheduler/controllers/migration/evict.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,11 +39,6 @@ func (r *Reconciler) Evict(ctx context.Context, pod *corev1.Pod, evictOptions fr
return true
}

if !r.Filter(pod) {
klog.Errorf("Pod %q cannot be evicted since failed to filter", klog.KObj(pod))
return false
}

err := CreatePodMigrationJob(ctx, pod, evictOptions, r.Client, r.args)
return err == nil
}
Expand Down
Loading