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
- type MetricRecorder
- type PendingPodsRecorder
Constants ¶
const ( // PrioritizingExtender - prioritizing extender work/operation label value. PrioritizingExtender = "prioritizing_extender" // Binding - binding work/operation label value. Binding = "binding" )
Below are possible values for the work and operation label.
const ( PreFilter = "PreFilter" Filter = "Filter" PreFilterExtensionAddPod = "PreFilterExtensionAddPod" PreFilterExtensionRemovePod = "PreFilterExtensionRemovePod" PostFilter = "PostFilter" PreScore = "PreScore" Score = "Score" ScoreExtensionNormalize = "ScoreExtensionNormalize" PreBind = "PreBind" Bind = "Bind" PostBind = "PostBind" Reserve = "Reserve" Unreserve = "Unreserve" Permit = "Permit" )
Below are possible values for the extension_point label.
const (
// SchedulerSubsystem - subsystem name used by scheduler.
SchedulerSubsystem = "scheduler"
)
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, }) 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"}) PluginEvaluationTotal = metrics.NewCounterVec( &metrics.CounterOpts{ Subsystem: SchedulerSubsystem, Name: "plugin_evaluation_total", Help: "Number of attempts to schedule pods by each plugin and the extension point (available only in PreFilter and Filter.).", StabilityLevel: metrics.ALPHA, }, []string{"plugin", "extension_point", "profile"}) )
All the histogram based metrics have 1ms as size for the smallest bucket.
var ( ScheduledResult = "scheduled" UnschedulableResult = "unschedulable" ErrorResult = "error" )
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 PodScheduleError ¶
PodScheduleError can records a scheduling attempt that had an error and the duration since `start`.
func PodScheduled ¶
PodScheduled can records a successful scheduling attempt and the duration since `start`.
func PodUnschedulable ¶
PodUnschedulable can records a scheduling attempt for an unschedulable pod and the duration since `start`.
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 ¶
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
Directories
¶
Path | Synopsis |
---|---|
Package resources provides a metrics collector that reports the resource consumption (requests and limits) of the pods in the cluster as the scheduler and kubelet would interpret it.
|
Package resources provides a metrics collector that reports the resource consumption (requests and limits) of the pods in the cluster as the scheduler and kubelet would interpret it. |