mirror of
https://github.com/ceph/ceph-csi.git
synced 2025-06-13 18:43:34 +00:00
rebase: update K8s packages to v0.32.1
Update K8s packages in go.mod to v0.32.1 Signed-off-by: Praveen M <m.praveen@ibm.com>
This commit is contained in:
224
vendor/k8s.io/kubernetes/pkg/scheduler/metrics/metric_recorder.go
generated
vendored
Normal file
224
vendor/k8s.io/kubernetes/pkg/scheduler/metrics/metric_recorder.go
generated
vendored
Normal file
@ -0,0 +1,224 @@
|
||||
/*
|
||||
Copyright 2019 The Kubernetes Authors.
|
||||
|
||||
Licensed under the Apache License, Version 2.0 (the "License");
|
||||
you may not use this file except in compliance with the License.
|
||||
You may obtain a copy of the License at
|
||||
|
||||
http://www.apache.org/licenses/LICENSE-2.0
|
||||
|
||||
Unless required by applicable law or agreed to in writing, software
|
||||
distributed under the License is distributed on an "AS IS" BASIS,
|
||||
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
See the License for the specific language governing permissions and
|
||||
limitations under the License.
|
||||
*/
|
||||
|
||||
package metrics
|
||||
|
||||
import (
|
||||
"time"
|
||||
|
||||
"k8s.io/component-base/metrics"
|
||||
)
|
||||
|
||||
// MetricRecorder represents a metric recorder which takes action when the
|
||||
// metric Inc(), Dec() and Clear()
|
||||
type MetricRecorder interface {
|
||||
Inc()
|
||||
Dec()
|
||||
Clear()
|
||||
}
|
||||
|
||||
var _ MetricRecorder = &PendingPodsRecorder{}
|
||||
|
||||
// PendingPodsRecorder is an implementation of MetricRecorder
|
||||
type PendingPodsRecorder struct {
|
||||
recorder metrics.GaugeMetric
|
||||
}
|
||||
|
||||
// NewActivePodsRecorder returns ActivePods in a Prometheus metric fashion
|
||||
func NewActivePodsRecorder() *PendingPodsRecorder {
|
||||
return &PendingPodsRecorder{
|
||||
recorder: ActivePods(),
|
||||
}
|
||||
}
|
||||
|
||||
// NewUnschedulablePodsRecorder returns UnschedulablePods in a Prometheus metric fashion
|
||||
func NewUnschedulablePodsRecorder() *PendingPodsRecorder {
|
||||
return &PendingPodsRecorder{
|
||||
recorder: UnschedulablePods(),
|
||||
}
|
||||
}
|
||||
|
||||
// NewBackoffPodsRecorder returns BackoffPods in a Prometheus metric fashion
|
||||
func NewBackoffPodsRecorder() *PendingPodsRecorder {
|
||||
return &PendingPodsRecorder{
|
||||
recorder: BackoffPods(),
|
||||
}
|
||||
}
|
||||
|
||||
// NewGatedPodsRecorder returns GatedPods in a Prometheus metric fashion
|
||||
func NewGatedPodsRecorder() *PendingPodsRecorder {
|
||||
return &PendingPodsRecorder{
|
||||
recorder: GatedPods(),
|
||||
}
|
||||
}
|
||||
|
||||
// Inc increases a metric counter by 1, in an atomic way
|
||||
func (r *PendingPodsRecorder) Inc() {
|
||||
r.recorder.Inc()
|
||||
}
|
||||
|
||||
// Dec decreases a metric counter by 1, in an atomic way
|
||||
func (r *PendingPodsRecorder) Dec() {
|
||||
r.recorder.Dec()
|
||||
}
|
||||
|
||||
// Clear set a metric counter to 0, in an atomic way
|
||||
func (r *PendingPodsRecorder) Clear() {
|
||||
r.recorder.Set(float64(0))
|
||||
}
|
||||
|
||||
// histogramVecMetric is the data structure passed in the buffer channel between the main framework thread
|
||||
// and the metricsRecorder goroutine.
|
||||
type histogramVecMetric struct {
|
||||
metric *metrics.HistogramVec
|
||||
labelValues []string
|
||||
value float64
|
||||
}
|
||||
|
||||
type gaugeVecMetric struct {
|
||||
metric *metrics.GaugeVec
|
||||
labelValues []string
|
||||
valueToAdd float64
|
||||
}
|
||||
|
||||
type gaugeVecMetricKey struct {
|
||||
metricName string
|
||||
labelValue string
|
||||
}
|
||||
|
||||
// MetricAsyncRecorder records metric in a separate goroutine to avoid overhead in the critical path.
|
||||
type MetricAsyncRecorder struct {
|
||||
// bufferCh is a channel that serves as a metrics buffer before the metricsRecorder goroutine reports it.
|
||||
bufferCh chan *histogramVecMetric
|
||||
// if bufferSize is reached, incoming metrics will be discarded.
|
||||
bufferSize int
|
||||
// how often the recorder runs to flush the metrics.
|
||||
interval time.Duration
|
||||
|
||||
// aggregatedInflightEventMetric is only to record InFlightEvents metric asynchronously.
|
||||
// It's a map from gaugeVecMetricKey to the aggregated value
|
||||
// and the aggregated value is flushed to Prometheus every time the interval is reached.
|
||||
// Note that we don't lock the map deliberately because we assume the queue takes lock before updating the in-flight events.
|
||||
aggregatedInflightEventMetric map[gaugeVecMetricKey]int
|
||||
aggregatedInflightEventMetricLastFlushTime time.Time
|
||||
aggregatedInflightEventMetricBufferCh chan *gaugeVecMetric
|
||||
|
||||
// stopCh is used to stop the goroutine which periodically flushes metrics.
|
||||
stopCh <-chan 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{}
|
||||
}
|
||||
|
||||
func NewMetricsAsyncRecorder(bufferSize int, interval time.Duration, stopCh <-chan struct{}) *MetricAsyncRecorder {
|
||||
recorder := &MetricAsyncRecorder{
|
||||
bufferCh: make(chan *histogramVecMetric, bufferSize),
|
||||
bufferSize: bufferSize,
|
||||
interval: interval,
|
||||
stopCh: stopCh,
|
||||
aggregatedInflightEventMetric: make(map[gaugeVecMetricKey]int),
|
||||
aggregatedInflightEventMetricLastFlushTime: time.Now(),
|
||||
aggregatedInflightEventMetricBufferCh: make(chan *gaugeVecMetric, bufferSize),
|
||||
IsStoppedCh: make(chan struct{}),
|
||||
}
|
||||
go recorder.run()
|
||||
return recorder
|
||||
}
|
||||
|
||||
// ObservePluginDurationAsync observes the plugin_execution_duration_seconds metric.
|
||||
// The metric will be flushed to Prometheus asynchronously.
|
||||
func (r *MetricAsyncRecorder) ObservePluginDurationAsync(extensionPoint, pluginName, status string, value float64) {
|
||||
r.observeMetricAsync(PluginExecutionDuration, value, pluginName, extensionPoint, status)
|
||||
}
|
||||
|
||||
// ObserveQueueingHintDurationAsync observes the queueing_hint_execution_duration_seconds metric.
|
||||
// The metric will be flushed to Prometheus asynchronously.
|
||||
func (r *MetricAsyncRecorder) ObserveQueueingHintDurationAsync(pluginName, event, hint string, value float64) {
|
||||
r.observeMetricAsync(queueingHintExecutionDuration, value, pluginName, event, hint)
|
||||
}
|
||||
|
||||
// ObserveInFlightEventsAsync observes the in_flight_events metric.
|
||||
//
|
||||
// Note that this function is not goroutine-safe;
|
||||
// we don't lock the map deliberately for the performance reason and we assume the queue (i.e., the caller) takes lock before updating the in-flight events.
|
||||
func (r *MetricAsyncRecorder) ObserveInFlightEventsAsync(eventLabel string, valueToAdd float64, forceFlush bool) {
|
||||
r.aggregatedInflightEventMetric[gaugeVecMetricKey{metricName: InFlightEvents.Name, labelValue: eventLabel}] += int(valueToAdd)
|
||||
|
||||
// Only flush the metric to the channel if the interval is reached.
|
||||
// The values are flushed to Prometheus in the run() function, which runs once the interval time.
|
||||
// Note: we implement this flushing here, not in FlushMetrics, because, if we did so, we would need to implement a lock for the map, which we want to avoid.
|
||||
if forceFlush || time.Since(r.aggregatedInflightEventMetricLastFlushTime) > r.interval {
|
||||
for key, value := range r.aggregatedInflightEventMetric {
|
||||
newMetric := &gaugeVecMetric{
|
||||
metric: InFlightEvents,
|
||||
labelValues: []string{key.labelValue},
|
||||
valueToAdd: float64(value),
|
||||
}
|
||||
select {
|
||||
case r.aggregatedInflightEventMetricBufferCh <- newMetric:
|
||||
default:
|
||||
}
|
||||
}
|
||||
r.aggregatedInflightEventMetricLastFlushTime = time.Now()
|
||||
// reset
|
||||
r.aggregatedInflightEventMetric = make(map[gaugeVecMetricKey]int)
|
||||
}
|
||||
}
|
||||
|
||||
func (r *MetricAsyncRecorder) observeMetricAsync(m *metrics.HistogramVec, value float64, labelsValues ...string) {
|
||||
newMetric := &histogramVecMetric{
|
||||
metric: m,
|
||||
labelValues: labelsValues,
|
||||
value: value,
|
||||
}
|
||||
select {
|
||||
case r.bufferCh <- newMetric:
|
||||
default:
|
||||
}
|
||||
}
|
||||
|
||||
// run flushes buffered metrics into Prometheus every second.
|
||||
func (r *MetricAsyncRecorder) run() {
|
||||
for {
|
||||
select {
|
||||
case <-r.stopCh:
|
||||
close(r.IsStoppedCh)
|
||||
return
|
||||
default:
|
||||
}
|
||||
r.FlushMetrics()
|
||||
time.Sleep(r.interval)
|
||||
}
|
||||
}
|
||||
|
||||
// FlushMetrics tries to clean up the bufferCh by reading at most bufferSize metrics.
|
||||
func (r *MetricAsyncRecorder) FlushMetrics() {
|
||||
for i := 0; i < r.bufferSize; i++ {
|
||||
select {
|
||||
case m := <-r.bufferCh:
|
||||
m.metric.WithLabelValues(m.labelValues...).Observe(m.value)
|
||||
default:
|
||||
// no more value
|
||||
}
|
||||
|
||||
select {
|
||||
case m := <-r.aggregatedInflightEventMetricBufferCh:
|
||||
m.metric.WithLabelValues(m.labelValues...).Add(m.valueToAdd)
|
||||
default:
|
||||
// no more value
|
||||
}
|
||||
}
|
||||
}
|
416
vendor/k8s.io/kubernetes/pkg/scheduler/metrics/metrics.go
generated
vendored
Normal file
416
vendor/k8s.io/kubernetes/pkg/scheduler/metrics/metrics.go
generated
vendored
Normal file
@ -0,0 +1,416 @@
|
||||
/*
|
||||
Copyright 2015 The Kubernetes Authors.
|
||||
|
||||
Licensed under the Apache License, Version 2.0 (the "License");
|
||||
you may not use this file except in compliance with the License.
|
||||
You may obtain a copy of the License at
|
||||
|
||||
http://www.apache.org/licenses/LICENSE-2.0
|
||||
|
||||
Unless required by applicable law or agreed to in writing, software
|
||||
distributed under the License is distributed on an "AS IS" BASIS,
|
||||
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
See the License for the specific language governing permissions and
|
||||
limitations under the License.
|
||||
*/
|
||||
|
||||
package metrics
|
||||
|
||||
import (
|
||||
"sync"
|
||||
"time"
|
||||
|
||||
utilfeature "k8s.io/apiserver/pkg/util/feature"
|
||||
"k8s.io/component-base/metrics"
|
||||
"k8s.io/component-base/metrics/legacyregistry"
|
||||
"k8s.io/kubernetes/pkg/features"
|
||||
volumebindingmetrics "k8s.io/kubernetes/pkg/scheduler/framework/plugins/volumebinding/metrics"
|
||||
)
|
||||
|
||||
const (
|
||||
// SchedulerSubsystem - subsystem name used by scheduler.
|
||||
SchedulerSubsystem = "scheduler"
|
||||
)
|
||||
|
||||
// Below are possible values for the work and operation label.
|
||||
const (
|
||||
// PrioritizingExtender - prioritizing extender work/operation label value.
|
||||
PrioritizingExtender = "prioritizing_extender"
|
||||
// Binding - binding work/operation label value.
|
||||
Binding = "binding"
|
||||
)
|
||||
|
||||
const (
|
||||
GoroutineResultSuccess = "success"
|
||||
GoroutineResultError = "error"
|
||||
)
|
||||
|
||||
// ExtentionPoints is a list of possible values for the extension_point label.
|
||||
var ExtentionPoints = []string{
|
||||
PreFilter,
|
||||
Filter,
|
||||
PreFilterExtensionAddPod,
|
||||
PreFilterExtensionRemovePod,
|
||||
PostFilter,
|
||||
PreScore,
|
||||
Score,
|
||||
ScoreExtensionNormalize,
|
||||
PreBind,
|
||||
Bind,
|
||||
PostBind,
|
||||
Reserve,
|
||||
Unreserve,
|
||||
Permit,
|
||||
}
|
||||
|
||||
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"
|
||||
)
|
||||
|
||||
const (
|
||||
QueueingHintResultQueue = "Queue"
|
||||
QueueingHintResultQueueSkip = "QueueSkip"
|
||||
QueueingHintResultError = "Error"
|
||||
)
|
||||
|
||||
const (
|
||||
PodPoppedInFlightEvent = "PodPopped"
|
||||
)
|
||||
|
||||
// All the histogram based metrics have 1ms as size for the smallest bucket.
|
||||
var (
|
||||
scheduleAttempts *metrics.CounterVec
|
||||
EventHandlingLatency *metrics.HistogramVec
|
||||
schedulingLatency *metrics.HistogramVec
|
||||
SchedulingAlgorithmLatency *metrics.Histogram
|
||||
PreemptionVictims *metrics.Histogram
|
||||
PreemptionAttempts *metrics.Counter
|
||||
pendingPods *metrics.GaugeVec
|
||||
InFlightEvents *metrics.GaugeVec
|
||||
Goroutines *metrics.GaugeVec
|
||||
|
||||
// PodSchedulingDuration is deprecated as of Kubernetes v1.28, and will be removed
|
||||
// in v1.31. Please use PodSchedulingSLIDuration instead.
|
||||
PodSchedulingDuration *metrics.HistogramVec
|
||||
PodSchedulingSLIDuration *metrics.HistogramVec
|
||||
PodSchedulingAttempts *metrics.Histogram
|
||||
FrameworkExtensionPointDuration *metrics.HistogramVec
|
||||
PluginExecutionDuration *metrics.HistogramVec
|
||||
|
||||
PermitWaitDuration *metrics.HistogramVec
|
||||
CacheSize *metrics.GaugeVec
|
||||
unschedulableReasons *metrics.GaugeVec
|
||||
PluginEvaluationTotal *metrics.CounterVec
|
||||
|
||||
// The below two are only available when the QHint feature gate is enabled.
|
||||
queueingHintExecutionDuration *metrics.HistogramVec
|
||||
SchedulerQueueIncomingPods *metrics.CounterVec
|
||||
|
||||
// The below two are only available when the async-preemption feature gate is enabled.
|
||||
PreemptionGoroutinesDuration *metrics.HistogramVec
|
||||
PreemptionGoroutinesExecutionTotal *metrics.CounterVec
|
||||
|
||||
// metricsList is a list of all metrics that should be registered always, regardless of any feature gate's value.
|
||||
metricsList []metrics.Registerable
|
||||
)
|
||||
|
||||
var registerMetrics sync.Once
|
||||
|
||||
// Register all metrics.
|
||||
func Register() {
|
||||
// Register the metrics.
|
||||
registerMetrics.Do(func() {
|
||||
InitMetrics()
|
||||
RegisterMetrics(metricsList...)
|
||||
volumebindingmetrics.RegisterVolumeSchedulingMetrics()
|
||||
|
||||
if utilfeature.DefaultFeatureGate.Enabled(features.SchedulerQueueingHints) {
|
||||
RegisterMetrics(queueingHintExecutionDuration, InFlightEvents)
|
||||
}
|
||||
if utilfeature.DefaultFeatureGate.Enabled(features.SchedulerAsyncPreemption) {
|
||||
RegisterMetrics(PreemptionGoroutinesDuration, PreemptionGoroutinesExecutionTotal)
|
||||
}
|
||||
})
|
||||
}
|
||||
|
||||
func InitMetrics() {
|
||||
scheduleAttempts = metrics.NewCounterVec(
|
||||
&metrics.CounterOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "schedule_attempts_total",
|
||||
Help: "Number of attempts to schedule pods, by the result. 'unschedulable' means a pod could not be scheduled, while 'error' means an internal scheduler problem.",
|
||||
StabilityLevel: metrics.STABLE,
|
||||
}, []string{"result", "profile"})
|
||||
|
||||
EventHandlingLatency = metrics.NewHistogramVec(
|
||||
&metrics.HistogramOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "event_handling_duration_seconds",
|
||||
Help: "Event handling latency in seconds.",
|
||||
// Start with 0.1ms with the last bucket being [~200ms, Inf)
|
||||
Buckets: metrics.ExponentialBuckets(0.0001, 2, 12),
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
}, []string{"event"})
|
||||
|
||||
schedulingLatency = metrics.NewHistogramVec(
|
||||
&metrics.HistogramOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "scheduling_attempt_duration_seconds",
|
||||
Help: "Scheduling attempt latency in seconds (scheduling algorithm + binding)",
|
||||
Buckets: metrics.ExponentialBuckets(0.001, 2, 15),
|
||||
StabilityLevel: metrics.STABLE,
|
||||
}, []string{"result", "profile"})
|
||||
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",
|
||||
// we think #victims>64 is pretty rare, therefore [64, +Inf) is considered a single bucket.
|
||||
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,
|
||||
})
|
||||
pendingPods = metrics.NewGaugeVec(
|
||||
&metrics.GaugeOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "pending_pods",
|
||||
Help: "Number of pending pods, by the queue type. 'active' means number of pods in activeQ; 'backoff' means number of pods in backoffQ; 'unschedulable' means number of pods in unschedulablePods that the scheduler attempted to schedule and failed; 'gated' is the number of unschedulable pods that the scheduler never attempted to schedule because they are gated.",
|
||||
StabilityLevel: metrics.STABLE,
|
||||
}, []string{"queue"})
|
||||
InFlightEvents = metrics.NewGaugeVec(
|
||||
&metrics.GaugeOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "inflight_events",
|
||||
Help: "Number of events currently tracked in the scheduling queue.",
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
}, []string{"event"})
|
||||
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 is deprecated as of Kubernetes v1.28, and will be removed
|
||||
// in v1.31. Please use PodSchedulingSLIDuration instead.
|
||||
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.",
|
||||
// Start with 10ms with the last bucket being [~88m, Inf).
|
||||
Buckets: metrics.ExponentialBuckets(0.01, 2, 20),
|
||||
StabilityLevel: metrics.STABLE,
|
||||
DeprecatedVersion: "1.29.0",
|
||||
},
|
||||
[]string{"attempts"})
|
||||
|
||||
PodSchedulingSLIDuration = metrics.NewHistogramVec(
|
||||
&metrics.HistogramOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "pod_scheduling_sli_duration_seconds",
|
||||
Help: "E2e latency for a pod being scheduled, from the time the pod enters the scheduling queue and might involve multiple scheduling attempts.",
|
||||
// Start with 10ms with the last bucket being [~88m, Inf).
|
||||
Buckets: metrics.ExponentialBuckets(0.01, 2, 20),
|
||||
StabilityLevel: metrics.BETA,
|
||||
},
|
||||
[]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.",
|
||||
// Start with 0.1ms with the last bucket being [~200ms, Inf)
|
||||
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.",
|
||||
// Start with 0.01ms with the last bucket being [~22ms, Inf). We use a small factor (1.5)
|
||||
// so that we have better granularity since plugin latency is very sensitive.
|
||||
Buckets: metrics.ExponentialBuckets(0.00001, 1.5, 20),
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
},
|
||||
[]string{"plugin", "extension_point", "status"})
|
||||
|
||||
// This is only available when the QHint feature gate is enabled.
|
||||
queueingHintExecutionDuration = metrics.NewHistogramVec(
|
||||
&metrics.HistogramOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "queueing_hint_execution_duration_seconds",
|
||||
Help: "Duration for running a queueing hint function of a plugin.",
|
||||
// Start with 0.01ms with the last bucket being [~22ms, Inf). We use a small factor (1.5)
|
||||
// so that we have better granularity since plugin latency is very sensitive.
|
||||
Buckets: metrics.ExponentialBuckets(0.00001, 1.5, 20),
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
},
|
||||
[]string{"plugin", "event", "hint"})
|
||||
|
||||
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"})
|
||||
|
||||
unschedulableReasons = metrics.NewGaugeVec(
|
||||
&metrics.GaugeOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "unschedulable_pods",
|
||||
Help: "The number of unschedulable pods broken down by plugin name. A pod will increment the gauge for all plugins that caused it to not schedule and so this metric have meaning only when broken down by plugin.",
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
}, []string{"plugin", "profile"})
|
||||
|
||||
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, Filter, PreScore, and Score).",
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
}, []string{"plugin", "extension_point", "profile"})
|
||||
|
||||
PreemptionGoroutinesDuration = metrics.NewHistogramVec(
|
||||
&metrics.HistogramOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "preemption_goroutines_duration_seconds",
|
||||
Help: "Duration in seconds for running goroutines for the preemption.",
|
||||
Buckets: metrics.ExponentialBuckets(0.01, 2, 20),
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
},
|
||||
[]string{"result"})
|
||||
|
||||
PreemptionGoroutinesExecutionTotal = metrics.NewCounterVec(
|
||||
&metrics.CounterOpts{
|
||||
Subsystem: SchedulerSubsystem,
|
||||
Name: "preemption_goroutines_execution_total",
|
||||
Help: "Number of preemption goroutines executed.",
|
||||
StabilityLevel: metrics.ALPHA,
|
||||
},
|
||||
[]string{"result"})
|
||||
|
||||
metricsList = []metrics.Registerable{
|
||||
scheduleAttempts,
|
||||
schedulingLatency,
|
||||
SchedulingAlgorithmLatency,
|
||||
EventHandlingLatency,
|
||||
PreemptionVictims,
|
||||
PreemptionAttempts,
|
||||
pendingPods,
|
||||
PodSchedulingDuration,
|
||||
PodSchedulingSLIDuration,
|
||||
PodSchedulingAttempts,
|
||||
FrameworkExtensionPointDuration,
|
||||
PluginExecutionDuration,
|
||||
SchedulerQueueIncomingPods,
|
||||
Goroutines,
|
||||
PermitWaitDuration,
|
||||
CacheSize,
|
||||
unschedulableReasons,
|
||||
PluginEvaluationTotal,
|
||||
}
|
||||
}
|
||||
|
||||
// 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 RegisterMetrics(extraMetrics ...metrics.Registerable) {
|
||||
for _, metric := range extraMetrics {
|
||||
legacyregistry.MustRegister(metric)
|
||||
}
|
||||
}
|
||||
|
||||
// GetGather returns the gatherer. It used by test case outside current package.
|
||||
func GetGather() metrics.Gatherer {
|
||||
return legacyregistry.DefaultGatherer
|
||||
}
|
||||
|
||||
// ActivePods returns the pending pods metrics with the label active
|
||||
func ActivePods() metrics.GaugeMetric {
|
||||
return pendingPods.With(metrics.Labels{"queue": "active"})
|
||||
}
|
||||
|
||||
// BackoffPods returns the pending pods metrics with the label backoff
|
||||
func BackoffPods() metrics.GaugeMetric {
|
||||
return pendingPods.With(metrics.Labels{"queue": "backoff"})
|
||||
}
|
||||
|
||||
// UnschedulablePods returns the pending pods metrics with the label unschedulable
|
||||
func UnschedulablePods() metrics.GaugeMetric {
|
||||
return pendingPods.With(metrics.Labels{"queue": "unschedulable"})
|
||||
}
|
||||
|
||||
// GatedPods returns the pending pods metrics with the label gated
|
||||
func GatedPods() metrics.GaugeMetric {
|
||||
return pendingPods.With(metrics.Labels{"queue": "gated"})
|
||||
}
|
||||
|
||||
// SinceInSeconds gets the time since the specified start in seconds.
|
||||
func SinceInSeconds(start time.Time) float64 {
|
||||
return time.Since(start).Seconds()
|
||||
}
|
||||
|
||||
func UnschedulableReason(plugin string, profile string) metrics.GaugeMetric {
|
||||
return unschedulableReasons.With(metrics.Labels{"plugin": plugin, "profile": profile})
|
||||
}
|
48
vendor/k8s.io/kubernetes/pkg/scheduler/metrics/profile_metrics.go
generated
vendored
Normal file
48
vendor/k8s.io/kubernetes/pkg/scheduler/metrics/profile_metrics.go
generated
vendored
Normal file
@ -0,0 +1,48 @@
|
||||
/*
|
||||
Copyright 2020 The Kubernetes Authors.
|
||||
|
||||
Licensed under the Apache License, Version 2.0 (the "License");
|
||||
you may not use this file except in compliance with the License.
|
||||
You may obtain a copy of the License at
|
||||
|
||||
http://www.apache.org/licenses/LICENSE-2.0
|
||||
|
||||
Unless required by applicable law or agreed to in writing, software
|
||||
distributed under the License is distributed on an "AS IS" BASIS,
|
||||
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
See the License for the specific language governing permissions and
|
||||
limitations under the License.
|
||||
*/
|
||||
|
||||
package metrics
|
||||
|
||||
// This file contains helpers for metrics that are associated to a profile.
|
||||
|
||||
var (
|
||||
ScheduledResult = "scheduled"
|
||||
UnschedulableResult = "unschedulable"
|
||||
ErrorResult = "error"
|
||||
)
|
||||
|
||||
// PodScheduled can records a successful scheduling attempt and the duration
|
||||
// since `start`.
|
||||
func PodScheduled(profile string, duration float64) {
|
||||
observeScheduleAttemptAndLatency(ScheduledResult, profile, duration)
|
||||
}
|
||||
|
||||
// PodUnschedulable can records a scheduling attempt for an unschedulable pod
|
||||
// and the duration since `start`.
|
||||
func PodUnschedulable(profile string, duration float64) {
|
||||
observeScheduleAttemptAndLatency(UnschedulableResult, profile, duration)
|
||||
}
|
||||
|
||||
// PodScheduleError can records a scheduling attempt that had an error and the
|
||||
// duration since `start`.
|
||||
func PodScheduleError(profile string, duration float64) {
|
||||
observeScheduleAttemptAndLatency(ErrorResult, profile, duration)
|
||||
}
|
||||
|
||||
func observeScheduleAttemptAndLatency(result, profile string, duration float64) {
|
||||
schedulingLatency.WithLabelValues(result, profile).Observe(duration)
|
||||
scheduleAttempts.WithLabelValues(result, profile).Inc()
|
||||
}
|
Reference in New Issue
Block a user