memorymanager:metrics: add metrics

As part of the memory manager GA graduation effort, we should add
metrics in order to iprove observability.

The metrics also mentioned in the PR https://github.com/kubernetes/enhancements/pull/4251 (which was not merged yet)

Signed-off-by: Talor Itzhak <titzhak@redhat.com>

pkg/kubelet/cm/memorymanager/policy_static.go

@@ -34,6 +34,7 @@ import (
 	"k8s.io/kubernetes/pkg/kubelet/cm/memorymanager/state"
 	"k8s.io/kubernetes/pkg/kubelet/cm/topologymanager"
 	"k8s.io/kubernetes/pkg/kubelet/cm/topologymanager/bitmask"
+	"k8s.io/kubernetes/pkg/kubelet/metrics"
 	"k8s.io/kubernetes/pkg/kubelet/types"
 )
 
@@ -95,7 +96,7 @@ func (p *staticPolicy) Start(s state.State) error {
 }
 
 // Allocate call is idempotent
-func (p *staticPolicy) Allocate(s state.State, pod *v1.Pod, container *v1.Container) error {
+func (p *staticPolicy) Allocate(s state.State, pod *v1.Pod, container *v1.Container) (rerr error) {
 	// allocate the memory only for guaranteed pods
 	if v1qos.GetPodQOS(pod) != v1.PodQOSGuaranteed {
 		return nil
@@ -103,6 +104,13 @@ func (p *staticPolicy) Allocate(s state.State, pod *v1.Pod, container *v1.Contai
 
 	podUID := string(pod.UID)
 	klog.InfoS("Allocate", "pod", klog.KObj(pod), "containerName", container.Name)
+	// container belongs in an exclusively allocated pool
+	metrics.MemoryManagerPinningRequestTotal.Inc()
+	defer func() {
+		if rerr != nil {
+			metrics.MemoryManagerPinningErrorsTotal.Inc()
+		}
+	}()
 	if blocks := s.GetMemoryBlocks(podUID, container.Name); blocks != nil {
 		p.updatePodReusableMemory(pod, container, blocks)
 

pkg/kubelet/metrics/metrics.go

@@ -108,6 +108,10 @@ const (
 	CPUManagerPinningRequestsTotalKey = "cpu_manager_pinning_requests_total"
 	CPUManagerPinningErrorsTotalKey   = "cpu_manager_pinning_errors_total"
 
+	// Metrics to track the Memory manager behavior
+	MemoryManagerPinningRequestsTotalKey = "memory_manager_pinning_requests_total"
+	MemoryManagerPinningErrorsTotalKey   = "memory_manager_pinning_errors_total"
+
 	// Metrics to track the Topology manager behavior
 	TopologyManagerAdmissionRequestsTotalKey = "topology_manager_admission_requests_total"
 	TopologyManagerAdmissionErrorsTotalKey   = "topology_manager_admission_errors_total"
@@ -719,6 +723,25 @@ var (
 		},
 	)
 
+	// MemoryManagerPinningRequestTotal tracks the number of times the pod spec required the memory manager to pin memory pages
+	MemoryManagerPinningRequestTotal = metrics.NewCounter(
+		&metrics.CounterOpts{
+			Subsystem:      KubeletSubsystem,
+			Name:           MemoryManagerPinningRequestsTotalKey,
+			Help:           "The number of memory pages allocations which required pinning.",
+			StabilityLevel: metrics.ALPHA,
+		})
+
+	// MemoryManagerPinningErrorsTotal tracks the number of times the pod spec required the memory manager to pin memory pages, but the allocation failed
+	MemoryManagerPinningErrorsTotal = metrics.NewCounter(
+		&metrics.CounterOpts{
+			Subsystem:      KubeletSubsystem,
+			Name:           MemoryManagerPinningErrorsTotalKey,
+			Help:           "The number of memory pages allocations which required pinning that failed.",
+			StabilityLevel: metrics.ALPHA,
+		},
+	)
+
 	// TopologyManagerAdmissionRequestsTotal tracks the number of times the pod spec will cause the topology manager to admit a pod
 	TopologyManagerAdmissionRequestsTotal = metrics.NewCounter(
 		&metrics.CounterOpts{
@@ -887,6 +910,10 @@ func Register(collectors ...metrics.StableCollector) {
 		legacyregistry.MustRegister(RunPodSandboxErrors)
 		legacyregistry.MustRegister(CPUManagerPinningRequestsTotal)
 		legacyregistry.MustRegister(CPUManagerPinningErrorsTotal)
+		if utilfeature.DefaultFeatureGate.Enabled(features.MemoryManager) {
+			legacyregistry.MustRegister(MemoryManagerPinningRequestTotal)
+			legacyregistry.MustRegister(MemoryManagerPinningErrorsTotal)
+		}
 		legacyregistry.MustRegister(TopologyManagerAdmissionRequestsTotal)
 		legacyregistry.MustRegister(TopologyManagerAdmissionErrorsTotal)
 		legacyregistry.MustRegister(TopologyManagerAdmissionDuration)