Merge pull request #84120 from smarterclayton/preempt

kubelet: Record preemptions similarly to evictions

pkg/kubelet/metrics/metrics.go

@@ -18,11 +18,12 @@ package metrics
 
 import (
 	"fmt"
-	"k8s.io/component-base/metrics"
-	"k8s.io/component-base/metrics/legacyregistry"
 	"sync"
 	"time"
 
+	"k8s.io/component-base/metrics"
+	"k8s.io/component-base/metrics/legacyregistry"
+
 	corev1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/types"
 	utilfeature "k8s.io/apiserver/pkg/util/feature"
@@ -45,6 +46,7 @@ const (
 	PLEGRelistIntervalKey                = "pleg_relist_interval_seconds"
 	EvictionsKey                         = "evictions"
 	EvictionStatsAgeKey                  = "eviction_stats_age_seconds"
+	PreemptionsKey                       = "preemptions"
 	DeprecatedPodWorkerLatencyKey        = "pod_worker_latency_microseconds"
 	DeprecatedPodStartLatencyKey         = "pod_start_latency_microseconds"
 	DeprecatedCgroupManagerOperationsKey = "cgroup_manager_latency_microseconds"
@@ -242,6 +244,18 @@ var (
 		},
 		[]string{"eviction_signal"},
 	)
+	// Preemptions is a Counter that tracks the cumulative number of pod preemptions initiated by the kubelet.
+	// Broken down by preemption signal. A preemption is only recorded for one resource, the sum of all signals
+	// is the number of preemptions on the given node.
+	Preemptions = metrics.NewCounterVec(
+		&metrics.CounterOpts{
+			Subsystem:      KubeletSubsystem,
+			Name:           PreemptionsKey,
+			Help:           "Cumulative number of pod preemptions by preemption resource",
+			StabilityLevel: metrics.ALPHA,
+		},
+		[]string{"preemption_signal"},
+	)
 	// DevicePluginRegistrationCount is a Counter that tracks the cumulative number of device plugin registrations.
 	// Broken down by resource name.
 	DevicePluginRegistrationCount = metrics.NewCounterVec(
@@ -502,6 +516,7 @@ func Register(containerCache kubecontainer.RuntimeCache, collectors ...metrics.C
 		legacyregistry.MustRegister(RuntimeOperationsErrors)
 		legacyregistry.MustRegister(Evictions)
 		legacyregistry.MustRegister(EvictionStatsAge)
+		legacyregistry.MustRegister(Preemptions)
 		legacyregistry.MustRegister(DevicePluginRegistrationCount)
 		legacyregistry.MustRegister(DevicePluginAllocationDuration)
 		legacyregistry.MustRegister(DeprecatedPodWorkerLatency)

pkg/kubelet/preemption/BUILD

@@ -16,6 +16,7 @@ go_library(
         "//pkg/kubelet/events:go_default_library",
         "//pkg/kubelet/eviction:go_default_library",
         "//pkg/kubelet/lifecycle:go_default_library",
+        "//pkg/kubelet/metrics:go_default_library",
         "//pkg/kubelet/types:go_default_library",
         "//pkg/kubelet/util/format:go_default_library",
         "//pkg/scheduler/algorithm/predicates:go_default_library",

pkg/kubelet/preemption/preemption.go

@@ -28,6 +28,7 @@ import (
 	"k8s.io/kubernetes/pkg/kubelet/events"
 	"k8s.io/kubernetes/pkg/kubelet/eviction"
 	"k8s.io/kubernetes/pkg/kubelet/lifecycle"
+	"k8s.io/kubernetes/pkg/kubelet/metrics"
 	kubetypes "k8s.io/kubernetes/pkg/kubelet/types"
 	"k8s.io/kubernetes/pkg/kubelet/util/format"
 	"k8s.io/kubernetes/pkg/scheduler/algorithm/predicates"
@@ -111,6 +112,11 @@ func (c *CriticalPodAdmissionHandler) evictPodsToFreeRequests(admitPod *v1.Pod,
 			// In future syncPod loops, the kubelet will retry the pod deletion steps that it was stuck on.
 			continue
 		}
+		if len(insufficientResources) > 0 {
+			metrics.Preemptions.WithLabelValues(insufficientResources[0].resourceName.String()).Inc()
+		} else {
+			metrics.Preemptions.WithLabelValues("").Inc()
+		}
 		klog.Infof("preemption: pod %s evicted successfully", format.Pod(pod))
 	}
 	return nil