cleanup(scheduler): move metricRecorder to metrics package

[sanposhiho]

What type of PR is this?

/kind cleanup

What this PR does / why we need it:

#115083 wants to record plugin_execution_duration_seconds metric for PreEnqueue in the scheduling queue.
Currently, all extension points record plugin_execution_duration_seconds via metricsRecorder which records the metrics in a separate goroutine. Then, it's better to utilize this for PreEnqueue as well for better performance.

The scheduling queue package ("k8s.io/kubernetes/pkg/scheduler/internal/queue") is outside of the framework package and the metricsRecorder in the frameworkImpl struct isn't callable from the scheduling queue.

This PR just moves it to pkg/scheduler/metrics/metric_recorder.go with renaming it to MetricsAsyncRecorder so that we can initialize MetricsAsyncRecorder outside the framework and use the same MetricsAsyncRecorder object both in the framework and the scheduling queue (will be done in another PR).

Which issue(s) this PR fixes:

Related to #115083

Special notes for your reviewer:

Does this PR introduce a user-facing change?

NONE

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.:

[k8s-ci-robot]

This issue is currently awaiting triage.

If a SIG or subproject determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[Huang-Wei]

is it intentional to change the type of status from Status to string?

[sanposhiho]

Yes. We cannot use k8s.io/kubernetes/pkg/scheduler/framework here due to the import cycle.

# When we import k8s.io/kubernetes/pkg/scheduler/framework here,
# and build the scheduler.

package k8s.io/kubernetes/pkg/scheduler/metrics
	imports k8s.io/kubernetes/pkg/scheduler/framework
	imports k8s.io/kubernetes/pkg/scheduler/framework/parallelize
	imports k8s.io/kubernetes/pkg/scheduler/metrics: import cycle not allowed

[Huang-Wei]

/lgtm
/approve

[k8s-ci-robot]

LGTM label has been added.

Git tree hash: c3e6307bdb5f86d612868cd1ce7c3f537334d52e

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: Huang-Wei, sanposhiho

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• pkg/scheduler/OWNERS [Huang-Wei]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[kerthcet]

/hold
can we use metrics directly here?

[sanposhiho]

Good catch. Fixed.

[kerthcet]

What do you mean use the same MetricsAsyncRecorder object both in the framework and the scheduling queue, they're different recorder. @sanposhiho

[sanposhiho]

I'm not sure I understood correctly what you mean but I'd like to do (in the next PR) is:

• store the same MetricsAsyncRecorder object in PriorityQueue struct.
• run priorityqueue.metricsRecorder.ObservePluginDurationAsync("PreEnqueue", pl.Name(), status.Code().String(), metrics.SinceInSeconds(startTime)) in runPreEnqueuePlugins.

[sanposhiho]

Why the "same" MetricsAsyncRecorder object? -> Each MetricsAsyncRecorder object has one goroutine to flush metrics periodically. To reduce the num of goroutines working in the scheduler, it's better to use the same MetricsAsyncRecorder object in both the framework and the priorityqueue.

[kerthcet]

Got the idea, I thought the same record was

kubernetes/pkg/scheduler/internal/queue/scheduling_queue.go

Line 298 in 1e84987

activeQ: heap.NewWithRecorder(podInfoKeyFunc, comp, metrics.NewActivePodsRecorder()),

/lgtm
/hold cancel
/retest

[k8s-ci-robot]

LGTM label has been added.

Git tree hash: a1086fa25a7c36e54d9981960677cd682a6f8b3c