metrics

Documentation

Index

• Constants
• Variables
• func ActivePods() metrics.GaugeMetric
• func BackoffPods() metrics.GaugeMetric
• func GatedPods() metrics.GaugeMetric
• func GetGather() metrics.Gatherer
• func PodScheduleError(profile string, duration float64)
• func PodScheduled(profile string, duration float64)
• func PodUnschedulable(profile string, duration float64)
• func Register()
• func RegisterMetrics(extraMetrics ...metrics.Registerable)
• func SinceInSeconds(start time.Time) float64
• func UnschedulablePods() metrics.GaugeMetric
• func UnschedulableReason(plugin string, profile string) metrics.GaugeMetric
• type MetricAsyncRecorder
  • func NewMetricsAsyncRecorder(bufferSize int, interval time.Duration, stopCh <-chan struct{}) *MetricAsyncRecorder
  • func (r *MetricAsyncRecorder) FlushMetrics()
  • func (r *MetricAsyncRecorder) ObservePluginDurationAsync(extensionPoint, pluginName, status string, value float64)
• type MetricRecorder
• type PendingPodsRecorder
  • func NewActivePodsRecorder() *PendingPodsRecorder
  • func NewBackoffPodsRecorder() *PendingPodsRecorder
  • func NewGatedPodsRecorder() *PendingPodsRecorder
  • func NewUnschedulablePodsRecorder() *PendingPodsRecorder
  • func (r *PendingPodsRecorder) Clear()
  • func (r *PendingPodsRecorder) Dec()
  • func (r *PendingPodsRecorder) Inc()

Constants

const (
	// SchedulerSubsystem - subsystem name used by scheduler
	SchedulerSubsystem = "scheduler"

	// PrioritizingExtender - prioritizing extender work label value
	PrioritizingExtender = "prioritizing_extender"
	// Binding - binding work label value
	Binding = "binding"
)

Variables

var (
	SchedulingAlgorithmLatency = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "scheduling_algorithm_duration_seconds",
			Help:           "Scheduling algorithm latency in seconds",
			Buckets:        metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel: metrics.ALPHA,
		},
	)
	PreemptionVictims = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem: SchedulerSubsystem,
			Name:      "preemption_victims",
			Help:      "Number of selected preemption victims",

			Buckets:        metrics.ExponentialBuckets(1, 2, 7),
			StabilityLevel: metrics.STABLE,
		})
	PreemptionAttempts = metrics.NewCounter(
		&metrics.CounterOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "preemption_attempts_total",
			Help:           "Total preemption attempts in the cluster till now",
			StabilityLevel: metrics.STABLE,
		})

	// SchedulerGoroutines isn't called in some parts where goroutines start.
	// Goroutines metric replaces SchedulerGoroutines metric. Goroutine metric tracks all goroutines.
	SchedulerGoroutines = metrics.NewGaugeVec(
		&metrics.GaugeOpts{
			Subsystem:         SchedulerSubsystem,
			DeprecatedVersion: "1.26.0",
			Name:              "scheduler_goroutines",
			Help:              "Number of running goroutines split by the work they do such as binding. This metric is replaced by the \"goroutines\" metric.",
			StabilityLevel:    metrics.ALPHA,
		}, []string{"work"})
	Goroutines = metrics.NewGaugeVec(
		&metrics.GaugeOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "goroutines",
			Help:           "Number of running goroutines split by the work they do such as binding.",
			StabilityLevel: metrics.ALPHA,
		}, []string{"operation"})
	PodSchedulingDuration = metrics.NewHistogramVec(
		&metrics.HistogramOpts{
			Subsystem: SchedulerSubsystem,
			Name:      "pod_scheduling_duration_seconds",
			Help:      "E2e latency for a pod being scheduled which may include multiple scheduling attempts.",

			Buckets:        metrics.ExponentialBuckets(0.01, 2, 20),
			StabilityLevel: metrics.STABLE,
		},
		[]string{"attempts"})

	PodSchedulingAttempts = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "pod_scheduling_attempts",
			Help:           "Number of attempts to successfully schedule a pod.",
			Buckets:        metrics.ExponentialBuckets(1, 2, 5),
			StabilityLevel: metrics.STABLE,
		})

	FrameworkExtensionPointDuration = metrics.NewHistogramVec(
		&metrics.HistogramOpts{
			Subsystem: SchedulerSubsystem,
			Name:      "framework_extension_point_duration_seconds",
			Help:      "Latency for running all plugins of a specific extension point.",

			Buckets:        metrics.ExponentialBuckets(0.0001, 2, 12),
			StabilityLevel: metrics.STABLE,
		},
		[]string{"extension_point", "status", "profile"})

	PluginExecutionDuration = metrics.NewHistogramVec(
		&metrics.HistogramOpts{
			Subsystem: SchedulerSubsystem,
			Name:      "plugin_execution_duration_seconds",
			Help:      "Duration for running a plugin at a specific extension point.",

			Buckets:        metrics.ExponentialBuckets(0.00001, 1.5, 20),
			StabilityLevel: metrics.ALPHA,
		},
		[]string{"plugin", "extension_point", "status"})

	SchedulerQueueIncomingPods = metrics.NewCounterVec(
		&metrics.CounterOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "queue_incoming_pods_total",
			Help:           "Number of pods added to scheduling queues by event and queue type.",
			StabilityLevel: metrics.STABLE,
		}, []string{"queue", "event"})

	PermitWaitDuration = metrics.NewHistogramVec(
		&metrics.HistogramOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "permit_wait_duration_seconds",
			Help:           "Duration of waiting on permit.",
			Buckets:        metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel: metrics.ALPHA,
		},
		[]string{"result"})

	CacheSize = metrics.NewGaugeVec(
		&metrics.GaugeOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "scheduler_cache_size",
			Help:           "Number of nodes, pods, and assumed (bound) pods in the scheduler cache.",
			StabilityLevel: metrics.ALPHA,
		}, []string{"type"})
)

All the histogram based metrics have 1ms as size for the smallest bucket.

Functions

func ActivePods

func ActivePods() metrics.GaugeMetric

ActivePods returns the pending pods metrics with the label active

func BackoffPods

func BackoffPods() metrics.GaugeMetric

BackoffPods returns the pending pods metrics with the label backoff

func GatedPods

func GatedPods() metrics.GaugeMetric

GatedPods returns the pending pods metrics with the label gated

func GetGather

func GetGather() metrics.Gatherer

GetGather returns the gatherer. It used by test case outside current package.

func PodScheduleError

func PodScheduleError(profile string, duration float64)

PodScheduleError can records a scheduling attempt that had an error and the duration since `start`.

func PodScheduled

func PodScheduled(profile string, duration float64)

PodScheduled can records a successful scheduling attempt and the duration since `start`.

func PodUnschedulable

func PodUnschedulable(profile string, duration float64)

PodUnschedulable can records a scheduling attempt for an unschedulable pod and the duration since `start`.

func Register

func Register()

Register all metrics.

func RegisterMetrics

func RegisterMetrics(extraMetrics ...metrics.Registerable)

RegisterMetrics registers a list of metrics. This function is exported because it is intended to be used by out-of-tree plugins to register their custom metrics.

func SinceInSeconds

func SinceInSeconds(start time.Time) float64

SinceInSeconds gets the time since the specified start in seconds.

func UnschedulablePods

func UnschedulablePods() metrics.GaugeMetric

UnschedulablePods returns the pending pods metrics with the label unschedulable

func UnschedulableReason

func UnschedulableReason(plugin string, profile string) metrics.GaugeMetric

Types

type MetricAsyncRecorder

type MetricAsyncRecorder struct {

	// IsStoppedCh indicates whether the goroutine is stopped. It's used in tests only to make sure
	// the metric flushing goroutine is stopped so that tests can collect metrics for verification.
	IsStoppedCh chan struct{}
	// contains filtered or unexported fields
}

MetricAsyncRecorder records metric in a separate goroutine to avoid overhead in the critical path.

func NewMetricsAsyncRecorder

func NewMetricsAsyncRecorder(bufferSize int, interval time.Duration, stopCh <-chan struct{}) *MetricAsyncRecorder

func (*MetricAsyncRecorder) FlushMetrics

func (r *MetricAsyncRecorder) FlushMetrics()

FlushMetrics tries to clean up the bufferCh by reading at most bufferSize metrics.

func (*MetricAsyncRecorder) ObservePluginDurationAsync

func (r *MetricAsyncRecorder) ObservePluginDurationAsync(extensionPoint, pluginName, status string, value float64)

ObservePluginDurationAsync observes the plugin_execution_duration_seconds metric. The metric will be flushed to Prometheus asynchronously.

type MetricRecorder

type MetricRecorder interface {
	Inc()
	Dec()
	Clear()
}

MetricRecorder represents a metric recorder which takes action when the metric Inc(), Dec() and Clear()

type PendingPodsRecorder

type PendingPodsRecorder struct {
	// contains filtered or unexported fields
}

PendingPodsRecorder is an implementation of MetricRecorder

func NewActivePodsRecorder

func NewActivePodsRecorder() *PendingPodsRecorder

NewActivePodsRecorder returns ActivePods in a Prometheus metric fashion

func NewBackoffPodsRecorder

func NewBackoffPodsRecorder() *PendingPodsRecorder

NewBackoffPodsRecorder returns BackoffPods in a Prometheus metric fashion

func NewGatedPodsRecorder

func NewGatedPodsRecorder() *PendingPodsRecorder

NewGatedPodsRecorder returns GatedPods in a Prometheus metric fashion

func NewUnschedulablePodsRecorder

func NewUnschedulablePodsRecorder() *PendingPodsRecorder

NewUnschedulablePodsRecorder returns UnschedulablePods in a Prometheus metric fashion

func (*PendingPodsRecorder) Clear

func (r *PendingPodsRecorder) Clear()

Clear set a metric counter to 0, in an atomic way

func (*PendingPodsRecorder) Dec

func (r *PendingPodsRecorder) Dec()

Dec decreases a metric counter by 1, in an atomic way

func (*PendingPodsRecorder) Inc

func (r *PendingPodsRecorder) Inc()

Inc increases a metric counter by 1, in an atomic way