Add fake runtimes and CRI changes for KEP-2371

Added new gRPC call 'ListPodSanboxMetrics' which would return additional
container stats currently supported by cAdvisor, but outside the scope
of /stats/summary api. Added new types to support metric exporting of
prometheus, including Metric and other subfields. Added fake runtime
changes associated with the CRI changes.

pkg/kubelet/cri/remote/fake/fake_runtime.go

@@ -336,3 +336,23 @@ func (f *RemoteRuntime) CheckpointContainer(ctx context.Context, req *kubeapi.Ch
 func (f *RemoteRuntime) GetContainerEvents(req *kubeapi.GetEventsRequest, ces kubeapi.RuntimeService_GetContainerEventsServer) error {
 	return nil
 }
+
+// ListMetricDescriptors gets the descriptors for the metrics that will be returned in ListPodSandboxMetrics.
+func (f *RemoteRuntime) ListMetricDescriptors(ctx context.Context, req *kubeapi.ListMetricDescriptorsRequest) (*kubeapi.ListMetricDescriptorsResponse, error) {
+	descs, err := f.RuntimeService.ListMetricDescriptors(ctx)
+	if err != nil {
+		return nil, err
+	}
+
+	return &kubeapi.ListMetricDescriptorsResponse{Descriptors: descs}, nil
+}
+
+// ListPodSandboxMetrics retrieves the metrics for all pod sandboxes.
+func (f *RemoteRuntime) ListPodSandboxMetrics(ctx context.Context, req *kubeapi.ListPodSandboxMetricsRequest) (*kubeapi.ListPodSandboxMetricsResponse, error) {
+	podMetrics, err := f.RuntimeService.ListPodSandboxMetrics(ctx)
+	if err != nil {
+		return nil, err
+	}
+
+	return &kubeapi.ListPodSandboxMetricsResponse{PodMetrics: podMetrics}, nil
+}

pkg/kubelet/cri/remote/remote_runtime.go

@@ -815,3 +815,33 @@ func (r *remoteRuntimeService) GetContainerEvents(containerEventsCh chan *runtim
 		}
 	}
 }
+
+// ListMetricDescriptors gets the descriptors for the metrics that will be returned in ListPodSandboxMetrics.
+func (r *remoteRuntimeService) ListMetricDescriptors(ctx context.Context) ([]*runtimeapi.MetricDescriptor, error) {
+	ctx, cancel := context.WithTimeout(ctx, r.timeout)
+	defer cancel()
+
+	resp, err := r.runtimeClient.ListMetricDescriptors(ctx, &runtimeapi.ListMetricDescriptorsRequest{})
+	if err != nil {
+		klog.ErrorS(err, "ListMetricDescriptors from runtime service failed")
+		return nil, err
+	}
+	klog.V(10).InfoS("[RemoteRuntimeService] ListMetricDescriptors Response", "stats", resp.GetDescriptors())
+
+	return resp.GetDescriptors(), nil
+}
+
+// ListPodSandboxMetrics retrieves the metrics for all pod sandboxes.
+func (r *remoteRuntimeService) ListPodSandboxMetrics(ctx context.Context) ([]*runtimeapi.PodSandboxMetrics, error) {
+	ctx, cancel := context.WithTimeout(ctx, r.timeout)
+	defer cancel()
+
+	resp, err := r.runtimeClient.ListPodSandboxMetrics(ctx, &runtimeapi.ListPodSandboxMetricsRequest{})
+	if err != nil {
+		klog.ErrorS(err, "ListPodSandboxMetrics from runtime service failed")
+		return nil, err
+	}
+	klog.V(10).InfoS("[RemoteRuntimeService] ListPodSandboxMetrics Response", "stats", resp.GetPodMetrics())
+
+	return resp.GetPodMetrics(), nil
+}

pkg/kubelet/kuberuntime/instrumented_services.go

@@ -343,3 +343,21 @@ func (in instrumentedRuntimeService) GetContainerEvents(containerEventsCh chan *
 	recordError(operation, err)
 	return err
 }
+
+func (in instrumentedRuntimeService) ListMetricDescriptors(ctx context.Context) ([]*runtimeapi.MetricDescriptor, error) {
+	const operation = "list_metric_descriptors"
+	defer recordOperation(operation, time.Now())
+
+	out, err := in.service.ListMetricDescriptors(ctx)
+	recordError(operation, err)
+	return out, err
+}
+
+func (in instrumentedRuntimeService) ListPodSandboxMetrics(ctx context.Context) ([]*runtimeapi.PodSandboxMetrics, error) {
+	const operation = "list_podsandbox_metrics"
+	defer recordOperation(operation, time.Now())
+
+	out, err := in.service.ListPodSandboxMetrics(ctx)
+	recordError(operation, err)
+	return out, err
+}