shum16
diff --git a/‎cmd/cadvisor.go‎
Lines changed: 3 additions & 1 deletion b/‎cmd/cadvisor.go‎
Lines changed: 3 additions & 1 deletion
diff --git a/‎container/raw/factory.go‎
Lines changed: 2 additions & 2 deletions b/‎container/raw/factory.go‎
Lines changed: 2 additions & 2 deletions
diff --git a/‎docs/runtime_options.md‎
Lines changed: 8 additions & 0 deletions b/‎docs/runtime_options.md‎
Lines changed: 8 additions & 0 deletions
diff --git a/‎info/v1/machine.go‎
Lines changed: 4 additions & 0 deletions b/‎info/v1/machine.go‎
Lines changed: 4 additions & 0 deletions
diff --git a/‎machine/info.go‎
Lines changed: 1 addition & 0 deletions b/‎machine/info.go‎
Lines changed: 1 addition & 0 deletions
diff --git a/‎machine/machine.go‎
Lines changed: 16 additions & 0 deletions b/‎machine/machine.go‎
Lines changed: 16 additions & 0 deletions
diff --git a/‎machine/topology_test.go‎
Lines changed: 24 additions & 0 deletions b/‎machine/topology_test.go‎
Lines changed: 24 additions & 0 deletions
diff --git a/‎manager/container.go‎
Lines changed: 2 additions & 1 deletion b/‎manager/container.go‎
Lines changed: 2 additions & 1 deletion
diff --git a/‎manager/manager.go‎
Lines changed: 10 additions & 14 deletions b/‎manager/manager.go‎
Lines changed: 10 additions & 14 deletions
@@ -73,6 +73,8 @@ var rawCgroupPrefixWhiteList = flag.String("raw_cgroup_prefix_whitelist", "", "A
 
 var perfEvents = flag.String("perf_events_config", "", "Path to a JSON file containing configuration of perf events to measure. Empty value disabled perf events measuring.")
 
+var resctrlInterval = flag.Duration("resctrl_interval", 0, "Resctrl mon groups updating interval. Zero value disables updating mon groups.")
+
 var (
 	// Metrics to be ignored.
 	// Tcp metrics are ignored by default.
@@ -131,7 +133,7 @@ func main() {
 
 	collectorHttpClient := createCollectorHttpClient(*collectorCert, *collectorKey)
 
-	resourceManager, err := manager.New(memoryStorage, sysFs, manager.HousekeepingConfigFlags, includedMetrics, &collectorHttpClient, strings.Split(*rawCgroupPrefixWhiteList, ","), strings.Split(*envMetadataWhiteList, ","), *perfEvents)
+	resourceManager, err := manager.New(memoryStorage, sysFs, manager.HousekeepingConfigFlags, includedMetrics, &collectorHttpClient, strings.Split(*rawCgroupPrefixWhiteList, ","), strings.Split(*envMetadataWhiteList, ","), *perfEvents, *resctrlInterval)
 	if err != nil {
 		klog.Fatalf("Failed to create a manager: %s", err)
 	}
 
@@ -29,7 +29,7 @@ import (
 	"k8s.io/klog/v2"
 )
 
-var dockerOnly = flag.Bool("docker_only", false, "Only report docker containers in addition to root stats")
+var DockerOnly = flag.Bool("docker_only", false, "Only report docker containers in addition to root stats")
 var disableRootCgroupStats = flag.Bool("disable_root_cgroup_stats", false, "Disable collecting root Cgroup stats")
 
 type rawFactory struct {
@@ -69,7 +69,7 @@ func (f *rawFactory) CanHandleAndAccept(name string) (bool, bool, error) {
 	if name == "/" {
 		return true, true, nil
 	}
-	if *dockerOnly && f.rawPrefixWhiteList[0] == "" {
+	if *DockerOnly && f.rawPrefixWhiteList[0] == "" {
 		return true, false, nil
 	}
 	for _, prefix := range f.rawPrefixWhiteList {
 
@@ -422,6 +422,14 @@ should be a human readable string that will become a metric name.
 * `cas_count_read` will be measured as uncore non-grouped event on all Integrated Memory Controllers Performance Monitoring Units because of unset `type` field and
 `uncore_imc` prefix.
 
+## Resctrl
+To gain metrics, cAdvisor creates own monitoring groups with `cadvisor` prefix.
+
+Resctrl file system is not hierarchical like cgroups, so users should set `--docker_only` flag to avoid race conditions and unexpected behaviours.
+
+```
+--resctrl_interval=0: Resctrl mon groups updating interval. Zero value disables updating mon groups.
+``` 
 
 ## Storage driver specific instructions:
 
 
@@ -178,6 +178,9 @@ type MachineInfo struct {
 	// The time of this information point.
 	Timestamp time.Time `json:"timestamp"`
 
+	// Vendor id of CPU.
+	CPUVendorID string `json:"vendor_id"`
+
 	// The number of cores in this machine.
 	NumCores int `json:"num_cores"`
 
@@ -249,6 +252,7 @@ func (m *MachineInfo) Clone() *MachineInfo {
 		}
 	}
 	copy := MachineInfo{
+		CPUVendorID:      m.CPUVendorID,
 		Timestamp:        m.Timestamp,
 		NumCores:         m.NumCores,
 		NumPhysicalCores: m.NumPhysicalCores,
 
@@ -121,6 +121,7 @@ func Info(sysFs sysfs.SysFs, fsInfo fs.FsInfo, inHostNamespace bool) (*info.Mach
 
 	machineInfo := &info.MachineInfo{
 		Timestamp:        time.Now(),
+		CPUVendorID:      GetCPUVendorID(cpuinfo),
 		NumCores:         numCores,
 		NumPhysicalCores: GetPhysicalCores(cpuinfo),
 		NumSockets:       GetSockets(cpuinfo),
 
@@ -43,6 +43,7 @@ var (
 	cpuClockSpeedMHz     = regexp.MustCompile(`(?:cpu MHz|CPU MHz|clock)\s*:\s*([0-9]+\.[0-9]+)(?:MHz)?`)
 	memoryCapacityRegexp = regexp.MustCompile(`MemTotal:\s*([0-9]+) kB`)
 	swapCapacityRegexp   = regexp.MustCompile(`SwapTotal:\s*([0-9]+) kB`)
+	vendorIDRegexp       = regexp.MustCompile(`vendor_id\s*:\s*(\w+)`)
 
 	cpuBusPath         = "/sys/bus/cpu/devices/"
 	isMemoryController = regexp.MustCompile("mc[0-9]+")
@@ -54,6 +55,21 @@ var (
 const memTypeFileName = "dimm_mem_type"
 const sizeFileName = "size"
 
+// GetCPUVendorID returns "vendor_id" reading /proc/cpuinfo file.
+func GetCPUVendorID(procInfo []byte) string {
+	vendorID := ""
+
+	matches := vendorIDRegexp.FindSubmatch(procInfo)
+	if len(matches) != 2 {
+		klog.Warning("Cannot read vendor id correctly, set empty.")
+		return vendorID
+	}
+
+	vendorID = string(matches[1])
+
+	return vendorID
+}
+
 // GetPhysicalCores returns number of CPU cores reading /proc/cpuinfo file or if needed information from sysfs cpu path
 func GetPhysicalCores(procInfo []byte) int {
 	numCores := getUniqueMatchesCount(string(procInfo), coreRegExp)
 
@@ -462,3 +462,27 @@ func TestClockSpeedOnCpuLowerCase(t *testing.T) {
 	assert.NotNil(t, clockSpeed)
 	assert.Equal(t, uint64(1450*1000), clockSpeed)
 }
+
+func TestGetCPUVendorID(t *testing.T) {
+	var testCases = []struct {
+		file     string
+		expected string
+	}{
+		{
+			"./testdata/cpuinfo_onesocket_many_NUMAs",
+			"GenuineIntel",
+		},
+		{
+			"./testdata/cpuinfo_arm",
+			"",
+		},
+	}
+
+	for _, test := range testCases {
+		testcpuinfo, err := ioutil.ReadFile(test.file)
+		assert.Nil(t, err)
+		assert.NotNil(t, testcpuinfo)
+		cpuVendorID := GetCPUVendorID(testcpuinfo)
+		assert.Equal(t, test.expected, cpuVendorID)
+	}
+}
@@ -130,6 +130,7 @@ func (cd *containerData) Stop() error {
 	}
 	close(cd.stop)
 	cd.perfCollector.Destroy()
+	cd.resctrlCollector.Destroy()
 	return nil
 }
 
@@ -727,7 +728,7 @@ func (cd *containerData) updateStats() error {
 		return perfStatsErr
 	}
 	if resctrlStatsErr != nil {
-		klog.Errorf("error occurred while collecting resctrl stats for container %s: %s", cInfo.Name, err)
+		klog.Errorf("error occurred while collecting resctrl stats for container %s: %s", cInfo.Name, resctrlStatsErr)
 		return resctrlStatsErr
 	}
 	return customStatsErr
 
@@ -49,7 +49,6 @@ import (
 
 	"github.com/opencontainers/runc/libcontainer/cgroups"
 	"github.com/opencontainers/runc/libcontainer/cgroups/fs2"
-	"github.com/opencontainers/runc/libcontainer/intelrdt"
 
 	"k8s.io/klog/v2"
 	"k8s.io/utils/clock"
@@ -152,7 +151,7 @@ type HouskeepingConfig = struct {
 }
 
 // New takes a memory storage and returns a new manager.
-func New(memoryCache *memory.InMemoryCache, sysfs sysfs.SysFs, houskeepingConfig HouskeepingConfig, includedMetricsSet container.MetricSet, collectorHTTPClient *http.Client, rawContainerCgroupPathPrefixWhiteList, containerEnvMetadataWhiteList []string, perfEventsFile string) (Manager, error) {
+func New(memoryCache *memory.InMemoryCache, sysfs sysfs.SysFs, houskeepingConfig HouskeepingConfig, includedMetricsSet container.MetricSet, collectorHTTPClient *http.Client, rawContainerCgroupPathPrefixWhiteList, containerEnvMetadataWhiteList []string, perfEventsFile string, resctrlInterval time.Duration) (Manager, error) {
 	if memoryCache == nil {
 		return nil, fmt.Errorf("manager requires memory storage")
 	}
@@ -224,7 +223,7 @@ func New(memoryCache *memory.InMemoryCache, sysfs sysfs.SysFs, houskeepingConfig
 		return nil, err
 	}
 
-	newManager.resctrlManager, err = resctrl.NewManager(selfContainer)
+	newManager.resctrlManager, err = resctrl.NewManager(resctrlInterval, resctrl.Setup, machineInfo.CPUVendorID, inHostNamespace)
 	if err != nil {
 		klog.V(4).Infof("Cannot gather resctrl metrics: %v", err)
 	}
@@ -269,7 +268,7 @@ type manager struct {
 	collectorHTTPClient      *http.Client
 	nvidiaManager            stats.Manager
 	perfManager              stats.Manager
-	resctrlManager           stats.Manager
+	resctrlManager           resctrl.Manager
 	// List of raw container cgroup path prefix whitelist.
 	rawContainerCgroupPathPrefixWhiteList []string
 	// List of container env prefix whitelist, the matched container envs would be collected into metrics as extra labels.
@@ -336,7 +335,7 @@ func (m *manager) Start() error {
 
 func (m *manager) Stop() error {
 	defer m.nvidiaManager.Destroy()
-	defer m.destroyPerfCollectors()
+	defer m.destroyCollectors()
 	// Stop and wait on all quit channels.
 	for i, c := range m.quitChannels {
 		// Send the exit signal and wait on the thread to exit (by closing the channel).
@@ -354,9 +353,10 @@ func (m *manager) Stop() error {
 	return nil
 }
 
-func (m *manager) destroyPerfCollectors() {
+func (m *manager) destroyCollectors() {
 	for _, container := range m.containers {
 		container.perfCollector.Destroy()
+		container.resctrlCollector.Destroy()
 	}
 }
 
@@ -973,14 +973,11 @@ func (m *manager) createContainerLocked(containerName string, watchSource watche
 	}
 
 	if m.includedMetrics.Has(container.ResctrlMetrics) {
-		resctrlPath, err := intelrdt.GetIntelRdtPath(containerName)
+		cont.resctrlCollector, err = m.resctrlManager.GetCollector(containerName, func() ([]string, error) {
+			return cont.getContainerPids(m.inHostNamespace)
+		}, len(m.machineInfo.Topology))
 		if err != nil {
-			klog.V(4).Infof("Error getting resctrl path: %q", err)
-		} else {
-			cont.resctrlCollector, err = m.resctrlManager.GetCollector(resctrlPath)
-			if err != nil {
-				klog.V(4).Infof("resctrl metrics will not be available for container %s: %s", cont.info.Name, err)
-			}
+			klog.V(4).Infof("resctrl metrics will not be available for container %s: %s", cont.info.Name, err)
 		}
 	}
 
@@ -1022,7 +1019,6 @@ func (m *manager) createContainerLocked(containerName string, watchSource watche
 	if err != nil {
 		return err
 	}
-
 	// Start the container's housekeeping.
 	return cont.Start()
 }
Original file line number	Diff line number	Diff line change
`@@ -29,7 +29,7 @@ import (`
`29`	`29`	`"k8s.io/klog/v2"`
`30`	`30`	`)`
`31`	`31`
`32`		`-var dockerOnly = flag.Bool("docker_only", false, "Only report docker containers in addition to root stats")`
	`32`	`+var DockerOnly = flag.Bool("docker_only", false, "Only report docker containers in addition to root stats")`
`33`	`33`	`var disableRootCgroupStats = flag.Bool("disable_root_cgroup_stats", false, "Disable collecting root Cgroup stats")`
`34`	`34`
`35`	`35`	`type rawFactory struct {`
`@@ -69,7 +69,7 @@ func (f *rawFactory) CanHandleAndAccept(name string) (bool, bool, error) {`
`69`	`69`	`if name == "/" {`
`70`	`70`	`return true, true, nil`
`71`	`71`	`}`
`72`		`- if *dockerOnly && f.rawPrefixWhiteList[0] == "" {`
	`72`	`+ if *DockerOnly && f.rawPrefixWhiteList[0] == "" {`
`73`	`73`	`return true, false, nil`
`74`	`74`	`}`
`75`	`75`	`for _, prefix := range f.rawPrefixWhiteList {`
Original file line number	Diff line number	Diff line change
`@@ -130,6 +130,7 @@ func (cd *containerData) Stop() error {`
`130`	`130`	`}`
`131`	`131`	`close(cd.stop)`
`132`	`132`	`cd.perfCollector.Destroy()`
	`133`	`+ cd.resctrlCollector.Destroy()`
`133`	`134`	`return nil`
`134`	`135`	`}`
`135`	`136`
`@@ -727,7 +728,7 @@ func (cd *containerData) updateStats() error {`
`727`	`728`	`return perfStatsErr`
`728`	`729`	`}`
`729`	`730`	`if resctrlStatsErr != nil {`
`730`		`- klog.Errorf("error occurred while collecting resctrl stats for container %s: %s", cInfo.Name, err)`
	`731`	`+ klog.Errorf("error occurred while collecting resctrl stats for container %s: %s", cInfo.Name, resctrlStatsErr)`
`731`	`732`	`return resctrlStatsErr`
`732`	`733`	`}`
`733`	`734`	`return customStatsErr`