metrics

Documentation

Index

• Constants
• Variables
• func RecordBESuppressCores(suppressType string, value float64)
• func RecordBESuppressLSUsedCPU(value float64)
• func RecordCollectNodeCPUInfoStatus(err error)
• func RecordContainerCPI(status *corev1.ContainerStatus, pod *corev1.Pod, cycles, instructions float64)
• func RecordContainerPSI(status *corev1.ContainerStatus, pod *corev1.Pod, psi *util.PSIByResource)
• func RecordContainerResourceLimits(resourceName string, status *corev1.ContainerStatus, pod *corev1.Pod, ...)
• func RecordContainerResourceRequests(resourceName string, status *corev1.ContainerStatus, pod *corev1.Pod, ...)
• func RecordContainerScaledCFSBurstUS(podNS, podName, containerID, containerName string, value float64)
• func RecordContainerScaledCFSQuotaUS(podNS, podName, containerID, containerName string, value float64)
• func RecordKoordletStartTime(nodeName string, value float64)
• func RecordNodeResourceAllocatable(resourceName string, value float64)
• func RecordNodeUsedCPU(value float64)
• func RecordPodEviction(reasonType string)
• func RecordPodPSI(pod *corev1.Pod, psi *util.PSIByResource)
• func Register(node *corev1.Node)
• func ResetCPUBurstCollector()
• func ResetContainerCPI()
• func ResetContainerPSI()
• func ResetContainerResourceLimits()
• func ResetContainerResourceRequests()
• func ResetPodPSI()
• type PSIRecord

Constants

const (
	CPIField = "cpi_field"

	Cycles       = "cycles"
	Instructions = "instructions"
)

const (
	KoordletSubsystem = "koordlet"

	NodeKey = "node"

	StatusKey     = "status"
	StatusSucceed = "succeeded"
	StatusFailed  = "failed"

	EvictionReasonKey = "reason"
	BESuppressTypeKey = "type"

	ContainerID   = "container_id"
	ContainerName = "container_name"

	PodUID       = "pod_uid"
	PodName      = "pod_name"
	PodNamespace = "pod_namespace"

	ResourceKey = "resource"
)

const (
	PSIDegree       = "psi_degree"
	PSIPrecision    = "psi_precision"
	PSIResourceType = "psi_resource_type"

	CPUFullSupported = "cpu_full_supported"
)

const (
	ResourceTypeCPU = "cpu"
	ResourceTypeMem = "mem"
	ResourceTypeIO  = "io"

	Precision10  = "avg10"
	Precision60  = "avg60"
	Precision300 = "avg300"

	DegreeSome = "some"
	DegreeFull = "full"
)

Variables

var (
	KoordletStartTime = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "start_time",
		Help:      "the start time of koordlet",
	}, []string{NodeKey})

	CollectNodeCPUInfoStatus = prometheus.NewCounterVec(prometheus.CounterOpts{
		Subsystem: KoordletSubsystem,
		Name:      "collect_node_cpu_info_status",
		Help:      "the count of CollectNodeCPUInfo status",
	}, []string{NodeKey, StatusKey})

	PodEviction = prometheus.NewCounterVec(prometheus.CounterOpts{
		Subsystem: KoordletSubsystem,
		Name:      "pod_eviction",
		Help:      "Number of eviction launched by koordlet",
	}, []string{NodeKey, EvictionReasonKey})

	NodeUsedCPU = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "node_used_cpu_cores",
		Help:      "Number of cpu cores used by node in realtime",
	}, []string{NodeKey})

	CommonCollectors = []prometheus.Collector{
		KoordletStartTime,
		CollectNodeCPUInfoStatus,
		PodEviction,
		NodeUsedCPU,
	}
)

var (
	ContainerCPI = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "container_cpi",
		Help:      "Container cpi collected by koordlet",
	}, []string{NodeKey, ContainerID, ContainerName, PodUID, PodName, PodNamespace, CPIField})

	CPICollectors = []prometheus.Collector{
		ContainerCPI,
	}
)

var (
	ContainerScaledCFSBurstUS = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "container_scaled_cfs_burst_us",
		Help:      "The maximum accumulated run-time(in microseconds) in container-level set by koordlet",
	}, []string{NodeKey, PodNamespace, PodName, ContainerID, ContainerName})

	ContainerScaledCFSQuotaUS = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "container_scaled_cfs_quota_us",
		Help:      "Run-time replenished within a period (in microseconds) in container-level set by koordlet",
	}, []string{NodeKey, PodNamespace, PodName, ContainerID, ContainerName})

	CPUBurstCollector = []prometheus.Collector{
		ContainerScaledCFSBurstUS,
		ContainerScaledCFSQuotaUS,
	}
)

var (
	BESuppressCPU = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "be_suppress_cpu_cores",
		Help:      "Number of cores suppress by koordlet",
	}, []string{NodeKey, BESuppressTypeKey})

	BESuppressLSUsedCPU = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "be_suppress_ls_used_cpu_cores",
		Help:      "Number of cpu cores used by LS. We consider non-BE pods and podMeta-missing pods as LS.",
	}, []string{NodeKey})

	CPUSuppressCollector = []prometheus.Collector{
		BESuppressCPU,
		BESuppressLSUsedCPU,
	}
)

var (
	NodeName string
	Node     *corev1.Node
)

var (
	ContainerPSI = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "container_psi",
		Help:      "Container psi collected by koordlet",
	}, []string{NodeKey, ContainerID, ContainerName, PodUID, PodName, PodNamespace, PSIResourceType, PSIPrecision, PSIDegree, CPUFullSupported})

	PodPSI = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "pod_psi",
		Help:      "Pod psi collected by koordlet",
	}, []string{NodeKey, PodUID, PodName, PodNamespace, PSIResourceType, PSIPrecision, PSIDegree, CPUFullSupported})

	PSICollectors = []prometheus.Collector{
		ContainerPSI,
		PodPSI,
	}
)

var (
	NodeResourceAllocatable = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "node_resource_allocatable",
		Help:      "the node allocatable of resources updated by koordinator",
	}, []string{NodeKey, ResourceKey})

	ContainerResourceRequests = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "container_resource_requests",
		Help:      "the container requests of resources updated by koordinator",
	}, []string{NodeKey, ResourceKey, PodUID, PodName, PodNamespace, ContainerID, ContainerName})

	ContainerResourceLimits = prometheus.NewGaugeVec(prometheus.GaugeOpts{
		Subsystem: KoordletSubsystem,
		Name:      "container_resource_limits",
		Help:      "the container limits of resources updated by koordinator",
	}, []string{NodeKey, ResourceKey, PodUID, PodName, PodNamespace, ContainerID, ContainerName})

	ResourceSummaryCollectors = []prometheus.Collector{
		NodeResourceAllocatable,
		ContainerResourceRequests,
		ContainerResourceLimits,
	}
)

Functions

func RecordBESuppressCores

func RecordBESuppressCores(suppressType string, value float64)

func RecordBESuppressLSUsedCPU

func RecordBESuppressLSUsedCPU(value float64)

func RecordCollectNodeCPUInfoStatus

func RecordCollectNodeCPUInfoStatus(err error)

func RecordContainerCPI

func RecordContainerCPI(status *corev1.ContainerStatus, pod *corev1.Pod, cycles, instructions float64)

func RecordContainerPSI

func RecordContainerPSI(status *corev1.ContainerStatus, pod *corev1.Pod, psi *util.PSIByResource)

func RecordContainerResourceLimits

func RecordContainerResourceLimits(resourceName string, status *corev1.ContainerStatus, pod *corev1.Pod, value float64)

func RecordContainerResourceRequests

func RecordContainerResourceRequests(resourceName string, status *corev1.ContainerStatus, pod *corev1.Pod, value float64)

func RecordContainerScaledCFSBurstUS

func RecordContainerScaledCFSBurstUS(podNS, podName, containerID, containerName string, value float64)

func RecordContainerScaledCFSQuotaUS

func RecordContainerScaledCFSQuotaUS(podNS, podName, containerID, containerName string, value float64)

func RecordKoordletStartTime

func RecordKoordletStartTime(nodeName string, value float64)

func RecordNodeResourceAllocatable

func RecordNodeResourceAllocatable(resourceName string, value float64)

func RecordNodeUsedCPU

func RecordNodeUsedCPU(value float64)

func RecordPodEviction

func RecordPodEviction(reasonType string)

func RecordPodPSI

func RecordPodPSI(pod *corev1.Pod, psi *util.PSIByResource)

func Register

func Register(node *corev1.Node)

Register registers the metrics with the node object

func ResetCPUBurstCollector

func ResetCPUBurstCollector()

func ResetContainerCPI

func ResetContainerCPI()

func ResetContainerPSI

func ResetContainerPSI()

func ResetContainerResourceLimits

func ResetContainerResourceLimits()

func ResetContainerResourceRequests

func ResetContainerResourceRequests()

func ResetPodPSI

func ResetPodPSI()

Types

type PSIRecord

type PSIRecord struct {
	ResourceType     string
	Precision        string
	Degree           string
	Value            float64
	CPUFullSupported bool
}