Optimize PrometheusCollector scrape efficiency.

cmd/internal/http/handlers.go

@@ -24,6 +24,7 @@ import (
 	"github.com/google/cadvisor/cmd/internal/pages"
 	"github.com/google/cadvisor/cmd/internal/pages/static"
 	"github.com/google/cadvisor/container"
+	v2 "github.com/google/cadvisor/info/v2"
 	"github.com/google/cadvisor/manager"
 	"github.com/google/cadvisor/metrics"
 	"github.com/google/cadvisor/validate"
@@ -92,31 +93,42 @@ func RegisterHandlers(mux httpmux.Mux, containerManager manager.Manager, httpAut
 	return nil
 }
 
-// RegisterPrometheusHandler creates a new PrometheusCollector and configures
+// RegisterPrometheusHandler creates a new ContainerCollector and configures
 // the provided HTTP mux to handle the given Prometheus endpoint.
 func RegisterPrometheusHandler(mux httpmux.Mux, resourceManager manager.Manager, prometheusEndpoint string,
 	f metrics.ContainerLabelsFunc, includedMetrics container.MetricSet) {
-	goCollector := prometheus.NewGoCollector()
-	processCollector := prometheus.NewProcessCollector(prometheus.ProcessCollectorOpts{})
-	machineCollector := metrics.NewPrometheusMachineCollector(resourceManager, includedMetrics)
+
+	r := prometheus.NewRegistry()
+	r.MustRegister(
+		prometheus.NewGoCollector(),
+		prometheus.NewProcessCollector(prometheus.ProcessCollectorOpts{}),
+	)
+	container := metrics.NewContainerCollector(resourceManager, f, includedMetrics, clock.RealClock{})
+	machine := metrics.NewMachineCollector(resourceManager, includedMetrics)
+
+	nameTypeCache := metrics.NewCachedGatherer(container.Collect, machine.Collect)
+	nameDockerCache := metrics.NewCachedGatherer(container.Collect, machine.Collect)
+	nameTypeGatherer := prometheus.NewMultiTRegistry(prometheus.ToTransactionalGatherer(r), nameTypeCache)
+	nameDockerGatherer := prometheus.NewMultiTRegistry(prometheus.ToTransactionalGatherer(r), nameDockerCache)
 
 	mux.Handle(prometheusEndpoint, http.HandlerFunc(func(w http.ResponseWriter, req *http.Request) {
+		// TODO(bwplotka): Why we ask for full object if we override half of fields?
 		opts, err := api.GetRequestOptions(req)
 		if err != nil {
 			http.Error(w, "No metrics gathered, last error:\n\n"+err.Error(), http.StatusInternalServerError)
 			return
 		}
-		opts.Count = 1        // we only want the latest datapoint
-		opts.Recursive = true // get all child containers
-
-		r := prometheus.NewRegistry()
-		r.MustRegister(
-			metrics.NewPrometheusCollector(resourceManager, f, includedMetrics, clock.RealClock{}, opts),
-			machineCollector,
-			goCollector,
-			processCollector,
-		)
-		promhttp.HandlerFor(r, promhttp.HandlerOpts{ErrorHandling: promhttp.ContinueOnError}).ServeHTTP(w, req)
+		opts.Count = 1        // We only want the latest datapoint.
+		opts.Recursive = true // Get all child containers.
+
+		// Present different metrics depending on option.
+		if opts.IdType == v2.TypeDocker {
+			nameDockerCache.UpdateOnMaxAge(opts)
+			promhttp.HandlerForTransactional(nameDockerGatherer, promhttp.HandlerOpts{ErrorHandling: promhttp.ContinueOnError}).ServeHTTP(w, req)
+			return
+		}
+		nameTypeCache.UpdateOnMaxAge(opts)
+		promhttp.HandlerForTransactional(nameTypeGatherer, promhttp.HandlerOpts{ErrorHandling: promhttp.ContinueOnError}).ServeHTTP(w, req)
 	}))
 }
 

container/libcontainer/handler.go

@@ -72,7 +72,7 @@ func NewHandler(cgroupManager cgroups.Manager, rootFs string, pid int, includedM
 	}
 }
 
-// Get cgroup and networking stats of the specified container
+// GetStats gets cgroup and networking stats of the specified container.
 func (h *Handler) GetStats() (*info.ContainerStats, error) {
 	ignoreStatsError := false
 	if cgroups.IsCgroup2UnifiedMode() {

go.mod

@@ -3,10 +3,11 @@ module github.com/google/cadvisor
 go 1.16
 
 require (
-	cloud.google.com/go v0.54.0
+	cloud.google.com/go v0.65.0
 	github.com/Microsoft/go-winio v0.4.15
 	github.com/aws/aws-sdk-go v1.35.24
 	github.com/blang/semver v3.5.1+incompatible
+	github.com/cespare/xxhash/v2 v2.1.2
 	github.com/containerd/containerd v1.4.12
 	github.com/containerd/ttrpc v1.0.2
 	github.com/containerd/typeurl v1.0.2
@@ -16,6 +17,7 @@ require (
 	github.com/docker/go-units v0.4.0
 	github.com/euank/go-kmsg-parser v2.0.0+incompatible
 	github.com/gogo/protobuf v1.3.2
+	github.com/golang/protobuf v1.5.2
 	github.com/gorilla/mux v1.8.0 // indirect
 	github.com/karrick/godirwalk v1.16.1
 	github.com/mindprince/gonvml v0.0.0-20190828220739-9ebdce4bb989
@@ -28,12 +30,12 @@ require (
 	github.com/opencontainers/runc v1.1.0
 	github.com/opencontainers/runtime-spec v1.0.3-0.20210326190908-1c3f411f0417
 	github.com/pkg/errors v0.9.1
-	github.com/prometheus/client_golang v1.7.1
+	github.com/prometheus/client_golang v1.12.2-0.20220223112252-1f81b3e9130f
 	github.com/prometheus/client_model v0.2.0
-	github.com/prometheus/common v0.10.0
+	github.com/prometheus/common v0.32.1
 	github.com/stretchr/testify v1.7.0
-	golang.org/x/net v0.0.0-20210226172049-e18ecbb05110
-	golang.org/x/sys v0.0.0-20211116061358-0a5406a5449c
+	golang.org/x/net v0.0.0-20210525063256-abc453219eb5
+	golang.org/x/sys v0.0.0-20220114195835-da31bd327af9
 	google.golang.org/grpc v1.33.2
 	k8s.io/klog/v2 v2.4.0
 	k8s.io/utils v0.0.0-20211116205334-6203023598ed