metrics

Documentation

Index

• Constants
• Variables
• func ActivePods() metrics.GaugeMetric
• func BackoffPods() metrics.GaugeMetric
• func GetGather() metrics.Gatherer
• func Register()
• func Reset()
• func SinceInSeconds(start time.Time) float64
• func UnschedulablePods() metrics.GaugeMetric
• type MetricRecorder
• type PendingPodsRecorder
  • func NewActivePodsRecorder() *PendingPodsRecorder
  • func NewBackoffPodsRecorder() *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"
	// DeprecatedSchedulingDurationName - scheduler duration metric name which is deprecated
	DeprecatedSchedulingDurationName = "scheduling_duration_seconds"

	// OperationLabel - operation label name
	OperationLabel = "operation"

	// PredicateEvaluation - predicate evaluation operation label value
	PredicateEvaluation = "predicate_evaluation"
	// PriorityEvaluation - priority evaluation operation label value
	PriorityEvaluation = "priority_evaluation"
	// PreemptionEvaluation - preemption evaluation operation label value (occurs in case of scheduling fitError).
	PreemptionEvaluation = "preemption_evaluation"
	// Binding - binding operation label value
	Binding = "binding"
)

Variables

var (

	// PodScheduleSuccesses counts how many pods were scheduled.
	// This metric will be initialized again in Register() to assure the metric is not no-op metric.
	PodScheduleSuccesses = scheduleAttempts.With(metrics.Labels{"result": "scheduled"})
	// PodScheduleFailures counts how many pods could not be scheduled.
	// This metric will be initialized again in Register() to assure the metric is not no-op metric.
	PodScheduleFailures = scheduleAttempts.With(metrics.Labels{"result": "unschedulable"})
	// PodScheduleErrors counts how many pods could not be scheduled due to a scheduler error.
	// This metric will be initialized again in Register() to assure the metric is not no-op metric.
	PodScheduleErrors            = scheduleAttempts.With(metrics.Labels{"result": "error"})
	DeprecatedSchedulingDuration = metrics.NewSummaryVec(
		&metrics.SummaryOpts{
			Subsystem: SchedulerSubsystem,
			Name:      DeprecatedSchedulingDurationName,
			Help:      "Scheduling latency in seconds split by sub-parts of the scheduling operation",

			MaxAge:            5 * time.Hour,
			StabilityLevel:    metrics.ALPHA,
			DeprecatedVersion: "1.19.0",
		},
		[]string{OperationLabel},
	)
	E2eSchedulingLatency = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "e2e_scheduling_duration_seconds",
			Help:           "E2e scheduling latency in seconds (scheduling algorithm + binding)",
			Buckets:        metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel: metrics.ALPHA,
		},
	)
	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,
		},
	)
	DeprecatedSchedulingAlgorithmPredicateEvaluationSecondsDuration = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:         SchedulerSubsystem,
			Name:              "scheduling_algorithm_predicate_evaluation_seconds",
			Help:              "Scheduling algorithm predicate evaluation duration in seconds",
			Buckets:           metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel:    metrics.ALPHA,
			DeprecatedVersion: "1.19.0",
		},
	)
	DeprecatedSchedulingAlgorithmPriorityEvaluationSecondsDuration = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:         SchedulerSubsystem,
			Name:              "scheduling_algorithm_priority_evaluation_seconds",
			Help:              "Scheduling algorithm priority evaluation duration in seconds",
			Buckets:           metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel:    metrics.ALPHA,
			DeprecatedVersion: "1.19.0",
		},
	)
	SchedulingAlgorithmPreemptionEvaluationDuration = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "scheduling_algorithm_preemption_evaluation_seconds",
			Help:           "Scheduling algorithm preemption evaluation duration in seconds",
			Buckets:        metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel: metrics.ALPHA,
		},
	)
	BindingLatency = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "binding_duration_seconds",
			Help:           "Binding latency in seconds",
			Buckets:        metrics.ExponentialBuckets(0.001, 2, 15),
			StabilityLevel: metrics.ALPHA,
		},
	)
	PreemptionVictims = metrics.NewHistogram(
		&metrics.HistogramOpts{
			Subsystem: SchedulerSubsystem,
			Name:      "pod_preemption_victims",
			Help:      "Number of selected preemption victims",

			Buckets:        metrics.LinearBuckets(5, 5, 10),
			StabilityLevel: metrics.ALPHA,
		})
	PreemptionAttempts = metrics.NewCounter(
		&metrics.CounterOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "total_preemption_attempts",
			Help:           "Total preemption attempts in the cluster till now",
			StabilityLevel: metrics.ALPHA,
		})

	SchedulerGoroutines = metrics.NewGaugeVec(
		&metrics.GaugeOpts{
			Subsystem:      SchedulerSubsystem,
			Name:           "scheduler_goroutines",
			Help:           "Number of running goroutines split by the work they do such as binding.",
			StabilityLevel: metrics.ALPHA,
		}, []string{"work"})

	PodSchedulingDuration = metrics.NewHistogram(
		&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(0.001, 2, 15),
			StabilityLevel: metrics.ALPHA,
		})

	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.ALPHA,
		})

	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.ALPHA,
		},
		[]string{"extension_point", "status"})

	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.ALPHA,
		}, []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 GetGather

func GetGather() metrics.Gatherer

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

func Register

func Register()

Register all metrics.

func Reset

func Reset()

Reset resets 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

Types

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 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