Skip to content

feat: Optimize Pod QoS sort for coscheduling #884

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
+70 −2
Open

feat: Optimize Pod QoS sort for coscheduling #884

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.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 15 additions & 2 deletions pkg/coscheduling/coscheduling.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,6 +28,7 @@ import (
"k8s.io/client-go/tools/cache"
corev1helpers "k8s.io/component-helpers/scheduling/corev1"
"k8s.io/klog/v2"
v1qos "k8s.io/kubernetes/pkg/apis/core/v1/helper/qos"
"k8s.io/kubernetes/pkg/scheduler/framework"
"sigs.k8s.io/controller-runtime/pkg/client"

Expand Down Expand Up @@ -124,14 +125,26 @@ func (cs *Coscheduling) Name() string {

// Less is used to sort pods in the scheduling queue in the following order.
// 1. Compare the priorities of Pods.
// 2. Compare the initialization timestamps of PodGroups or Pods.
// 3. Compare the keys of PodGroups/Pods: <namespace>/<podname>.
// 2. Compare the Qos of Pods.
Copy link
Contributor

Choose a reason for hiding this comment

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

This would change the behavior, and comparing QoS is not the intent of this plugin.

// 3. Compare the initialization timestamps of PodGroups or Pods.
// 4. Compare the keys of PodGroups/Pods: <namespace>/<podname>.
func (cs *Coscheduling) Less(podInfo1, podInfo2 *framework.QueuedPodInfo) bool {
prio1 := corev1helpers.PodPriority(podInfo1.Pod)
prio2 := corev1helpers.PodPriority(podInfo2.Pod)
if prio1 != prio2 {
return prio1 > prio2
}

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 to obtain Qos type judgment in coscheduling? 🤔
If I understand correctly, pod have a priority defined by their own priority.
In QoS, it is used as a basis for judging whether to evict, and there is no need to sort QoS.

Copy link
Author

Choose a reason for hiding this comment

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

Hmmm.... Before comparing CreateTime, it may make more sense to compare the QosClass set by the user.

Or, Compare CreateTime first and then compare QosClass?

Copy link
Member

Choose a reason for hiding this comment

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

I understand createTime seems to be enough, because this change will not solve the issue: #874

if v1qos.GetPodQOS(podInfo1.Pod) != v1qos.GetPodQOS(podInfo2.Pod) {
if v1qos.GetPodQOS(podInfo1.Pod) == v1.PodQOSBestEffort {
return false
}
if v1qos.GetPodQOS(podInfo1.Pod) == v1.PodQOSBurstable && v1qos.GetPodQOS(podInfo2.Pod) == v1.PodQOSGuaranteed {
return false
}
return true
}

creationTime1 := cs.pgMgr.GetCreationTimestamp(context.TODO(), podInfo1.Pod, *podInfo1.InitialAttemptTimestamp)
creationTime2 := cs.pgMgr.GetCreationTimestamp(context.TODO(), podInfo2.Pod, *podInfo2.InitialAttemptTimestamp)
if creationTime1.Equal(creationTime2) {
Expand Down
55 changes: 55 additions & 0 deletions pkg/coscheduling/coscheduling_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,6 +187,11 @@ func TestLess(t *testing.T) {
lowPriority, highPriority := int32(10), int32(100)
now := time.Now()

capacity := map[v1.ResourceName]string{
v1.ResourceCPU: "25m",
v1.ResourceMemory: "100Mi",
}

tests := []struct {
name string
p1 *framework.QueuedPodInfo
Expand Down Expand Up @@ -400,6 +405,56 @@ func TestLess(t *testing.T) {
},
want: true,
},
{
name: "equal priority, p1.Qos greater than p2.Qos and p2 belong to pg2",
p1: &framework.QueuedPodInfo{
PodInfo: tu.MustNewPodInfo(t, st.MakePod().Name("p1").Namespace("ns1").Priority(highPriority).
Containers(
[]v1.Container{
st.MakeContainer().Name("p1").ResourceRequests(capacity).ResourceLimits(capacity).Obj(),
},
).Obj()),
InitialAttemptTimestamp: ptrTime(now.Add(time.Second * 1)),
},
p2: &framework.QueuedPodInfo{
PodInfo: tu.MustNewPodInfo(t, st.MakePod().Name("p2").Namespace("ns2").Priority(highPriority).
Label(v1alpha1.PodGroupLabel, "pg2").Containers(
[]v1.Container{
st.MakeContainer().Name("p1").ResourceRequests(capacity).Obj(),
},
).Obj()),
InitialAttemptTimestamp: ptrTime(now.Add(time.Second * 1)),
},
pgs: []*v1alpha1.PodGroup{
tu.MakePodGroup().Name("pg2").Namespace("ns2").Time(now.Add(time.Second * 2)).Obj(),
},
want: true,
},
{
name: "equal priority, p1.Qos greater than p2.Qos and p1 belongs pg1 and p2 belong to pg2",
p1: &framework.QueuedPodInfo{
PodInfo: tu.MustNewPodInfo(t, st.MakePod().Name("p1").Namespace("ns1").Priority(highPriority).
Label(v1alpha1.PodGroupLabel, "pg1").Containers(
[]v1.Container{
st.MakeContainer().Name("p1").ResourceRequests(capacity).ResourceLimits(capacity).Obj(),
},
).Obj()),
InitialAttemptTimestamp: ptrTime(now.Add(time.Second * 1)),
},
p2: &framework.QueuedPodInfo{
PodInfo: tu.MustNewPodInfo(t, st.MakePod().Name("p2").Namespace("ns2").Priority(highPriority).
Label(v1alpha1.PodGroupLabel, "pg2").Containers(
[]v1.Container{
st.MakeContainer().Name("p2").Obj(),
},
).Obj()),
InitialAttemptTimestamp: ptrTime(now.Add(time.Second * 1)),
},
pgs: []*v1alpha1.PodGroup{
tu.MakePodGroup().Name("pg1").Namespace("ns1").Time(now.Add(time.Second * 1)).Obj(),
tu.MakePodGroup().Name("pg2").Namespace("ns2").Time(now.Add(time.Second * 2)).Obj()},
want: true,
},
}

for _, tt := range tests {
Expand Down