Merge pull request #2574 from RenaudWasTaken/nvidia

Return a NoopManager if metricset does not container the accelerator value

accelerators/nvidia.go

@@ -24,6 +24,7 @@ import (
 	"sync"
 	"time"
 
+	"github.com/google/cadvisor/container"
 	info "github.com/google/cadvisor/info/v1"
 	"github.com/google/cadvisor/stats"
 
@@ -48,18 +49,23 @@ var sysFsPCIDevicesPath = "/sys/bus/pci/devices/"
 
 const nvidiaVendorID = "0x10de"
 
-func NewNvidiaManager() stats.Manager {
+func NewNvidiaManager(includedMetrics container.MetricSet) stats.Manager {
+	if !includedMetrics.Has(container.AcceleratorUsageMetrics) {
+		klog.V(2).Info("NVIDIA GPU metrics disabled")
+		return &stats.NoopManager{}
+	}
+
 	manager := &nvidiaManager{}
 	err := manager.setup()
 	if err != nil {
-		klog.Warningf("NVidia GPU metrics will not be available: %s", err)
+		klog.Warningf("NVIDIA GPU metrics will not be available: %s", err)
 		manager.Destroy()
 		return &stats.NoopManager{}
 	}
 	return manager
 }
 
-// setup initializes NVML if nvidia devices are present on the node.
+// setup initializes NVML if NVIDIA devices are present on the node.
 func (nm *nvidiaManager) setup() error {
 	if !detectDevices(nvidiaVendorID) {
 		return fmt.Errorf("no NVIDIA devices found")
@@ -104,21 +110,21 @@ var initializeNVML = func(nm *nvidiaManager) error {
 	nm.nvmlInitialized = true
 	numDevices, err := gonvml.DeviceCount()
 	if err != nil {
-		return fmt.Errorf("GPU metrics would not be available. Failed to get the number of nvidia devices: %v", err)
+		return fmt.Errorf("GPU metrics would not be available. Failed to get the number of NVIDIA devices: %v", err)
 	}
 	if numDevices == 0 {
 		return nil
 	}
-	klog.V(1).Infof("NVML initialized. Number of nvidia devices: %v", numDevices)
+	klog.V(1).Infof("NVML initialized. Number of NVIDIA devices: %v", numDevices)
 	nm.nvidiaDevices = make(map[int]gonvml.Device, numDevices)
 	for i := 0; i < int(numDevices); i++ {
 		device, err := gonvml.DeviceHandleByIndex(uint(i))
 		if err != nil {
-			return fmt.Errorf("Failed to get nvidia device handle %d: %v", i, err)
+			return fmt.Errorf("Failed to get NVIDIA device handle %d: %v", i, err)
 		}
 		minorNumber, err := device.MinorNumber()
 		if err != nil {
-			return fmt.Errorf("Failed to get nvidia device minor number: %v", err)
+			return fmt.Errorf("Failed to get NVIDIA device minor number: %v", err)
 		}
 		nm.nvidiaDevices[int(minorNumber)] = device
 	}
@@ -135,7 +141,7 @@ func (nm *nvidiaManager) Destroy() {
 	}
 }
 
-// GetCollector returns a collector that can fetch nvidia gpu metrics for nvidia devices
+// GetCollector returns a collector that can fetch NVIDIA gpu metrics for NVIDIA devices
 // present in the devices.list file in the given devicesCgroupPath.
 func (nm *nvidiaManager) GetCollector(devicesCgroupPath string) (stats.Collector, error) {
 	nc := &nvidiaCollector{}
@@ -165,7 +171,7 @@ func (nm *nvidiaManager) GetCollector(devicesCgroupPath string) (stats.Collector
 	for _, minor := range nvidiaMinorNumbers {
 		device, ok := nm.nvidiaDevices[minor]
 		if !ok {
-			return &stats.NoopCollector{}, fmt.Errorf("nvidia device minor number %d not found in cached devices", minor)
+			return &stats.NoopCollector{}, fmt.Errorf("NVIDIA device minor number %d not found in cached devices", minor)
 		}
 		nc.devices = append(nc.devices, device)
 	}

cmd/cadvisor.go

@@ -93,6 +93,7 @@ var (
 
 	// List of metrics that can be ignored.
 	ignoreWhitelist = container.MetricSet{
+		container.AcceleratorUsageMetrics:        struct{}{},
 		container.DiskUsageMetrics:               struct{}{},
 		container.DiskIOMetrics:                  struct{}{},
 		container.NetworkUsageMetrics:            struct{}{},
@@ -136,7 +137,7 @@ func (ml *metricSetValue) Set(value string) error {
 }
 
 func init() {
-	flag.Var(&ignoreMetrics, "disable_metrics", "comma-separated list of `metrics` to be disabled. Options are 'cpu_topology','disk', 'diskIO', 'network', 'tcp', 'udp', 'percpu', 'sched', 'process', 'hugetlb', 'referenced_memory'.")
+	flag.Var(&ignoreMetrics, "disable_metrics", "comma-separated list of `metrics` to be disabled. Options are 'accelerator', 'cpu_topology','disk', 'diskIO', 'network', 'tcp', 'udp', 'percpu', 'sched', 'process', 'hugetlb', 'referenced_memory'.")
 
 	// Default logging verbosity to V(2)
 	flag.Set("v", "2")

manager/manager.go

@@ -197,7 +197,7 @@ func New(memoryCache *memory.InMemoryCache, sysfs sysfs.SysFs, houskeepingConfig
 		containerWatchers:                     []watcher.ContainerWatcher{},
 		eventsChannel:                         eventsChannel,
 		collectorHTTPClient:                   collectorHTTPClient,
-		nvidiaManager:                         accelerators.NewNvidiaManager(),
+		nvidiaManager:                         accelerators.NewNvidiaManager(includedMetricsSet),
 		rawContainerCgroupPathPrefixWhiteList: rawContainerCgroupPathPrefixWhiteList,
 	}