metrics

Documentation

Index

• Constants
• Variables
• func GetAnnotationsFromRuntimeInfo(info interface{}) map[string]string
• func GetQueryLabelsFromPrometheus(query string) map[string]string
• func InitMetrics()
• func StartMetricsService(port int, queueFunc ListQueueFunc, jobFunc ListJobFunc) string
• type Info
• type JobMetricCollector
  • func NewJobMetricsCollector(manager TimePointManager, listJob ListJobFunc) *JobMetricCollector
  • func (j *JobMetricCollector) Collect(metrics chan<- prometheus.Metric)
  • func (j *JobMetricCollector) Describe(descs chan<- *prometheus.Desc)
• type JobStatus
  • func (j JobStatus) String() string
  • func (j JobStatus) TimePoint() (start TimePoint, end TimePoint)
• type JobTimePoint
  • func (t JobTimePoint) Index() int
  • func (t JobTimePoint) Status() Status
• type ListJobFunc
• type ListQueueFunc
• type QueueMetricCollector
  • func NewQueueMetricsCollector(queueFunc ListQueueFunc) *QueueMetricCollector
  • func (q *QueueMetricCollector) Collect(metrics chan<- prometheus.Metric)
  • func (q *QueueMetricCollector) Describe(descs chan<- *prometheus.Desc)
• type Status
• type TimePoint
• type TimePointManager
  • func NewJobMetricTimePointManager() TimePointManager
• type Timestamps

Constants

const (
	MetricJobTime    = "pf_metric_job_time"
	MetricQueueInfo  = "pf_metric_queue_info"
	MetricJobGPUInfo = "pf_metric_job_gpu_info"
)

const (
	JobIDLabel          = "jobID"
	GpuIdxLabel         = "id"
	StatusLabel         = "status"
	QueueIDLabel        = "queueID"
	FinishedStatusLabel = "finishedStatus"
	QueueNameLabel      = "queueName"
	UserNameLabel       = "userName"
	ResourceLabel       = "resource"
	TypeLabel           = "type"
	BaiduGpuIndexLabel  = "baidu_com_gpu_idx"
)

const (
	MinTimePoint = T1
	MinStatus    = StatusDBInserting
	MaxTimePoint = T8
	MaxStatus    = StatusRunning
)

const (
	MaxNum       = 10000
	Timeout      = time.Hour
	ZeroDuration = time.Duration(0)
)

const (
	QueueTypeMaxResource    = "maxResource"
	QueueTypeMinResource    = "minResource"
	QueueTypeScalarResource = "scalarResource"
)

const (
	PromQLQueryPodAnnotations = "kube_pod_annotations{pod~=\"%s\"}"
	PromQLQueryPodLabels      = "kube_pod_labels{pod~=\"%s\"}"
)

const (
	DefaultMetricPort = 8231
)

const (
	QueryTimeout = time.Second * 1
)

Variables

var (
	ZeroTime = time.Time{}
)

Functions

func GetAnnotationsFromRuntimeInfo

func GetAnnotationsFromRuntimeInfo(info interface{}) map[string]string

GetAnnotationsFromRuntimeInfo get annotations from info map

func GetQueryLabelsFromPrometheus

func GetQueryLabelsFromPrometheus(query string) map[string]string

GetQueryLabelsFromPrometheus return query labels from prometheus Deprecated

func InitMetrics

func InitMetrics()

func StartMetricsService

func StartMetricsService(port int, queueFunc ListQueueFunc, jobFunc ListJobFunc) string

Types

type Info

type Info map[string]string

type JobMetricCollector

type JobMetricCollector struct {
	// contains filtered or unexported fields
}

func NewJobMetricsCollector

func NewJobMetricsCollector(manager TimePointManager, listJob ListJobFunc) *JobMetricCollector

func (*JobMetricCollector) Collect

func (j *JobMetricCollector) Collect(metrics chan<- prometheus.Metric)

func (*JobMetricCollector) Describe

func (j *JobMetricCollector) Describe(descs chan<- *prometheus.Desc)

type JobStatus

type JobStatus int

const (
	StatusUnknown JobStatus = iota
	StatusDBInserting
	StatusEnqueue
	StatusDequeue
	StatusSubmitting
	StatusPending
	StatusCreating
	StatusRunning
)

func (JobStatus) String

func (j JobStatus) String() string

func (JobStatus) TimePoint

func (j JobStatus) TimePoint() (start TimePoint, end TimePoint)

type JobTimePoint

type JobTimePoint int

const (
	// T1 api query time
	T1 JobTimePoint = iota
	// T2 db insert time
	T2
	// T3 enqueue time
	T3
	// T4 dequeue time
	T4
	// T5 submit time
	T5
	// T6 scheduled time
	// TODO: T6 is not supported yet
	T6
	// T7 run time
	T7
	// T8 finish(success/fail) time
	T8
)

func (JobTimePoint) Index

func (t JobTimePoint) Index() int

func (JobTimePoint) Status

func (t JobTimePoint) Status() Status

type ListJobFunc

type ListJobFunc func() []model.Job

type ListQueueFunc

type ListQueueFunc func() []model.Queue

type QueueMetricCollector

type QueueMetricCollector struct {
	// contains filtered or unexported fields
}

func NewQueueMetricsCollector

func NewQueueMetricsCollector(queueFunc ListQueueFunc) *QueueMetricCollector

func (*QueueMetricCollector) Collect

func (q *QueueMetricCollector) Collect(metrics chan<- prometheus.Metric)

func (*QueueMetricCollector) Describe

func (q *QueueMetricCollector) Describe(descs chan<- *prometheus.Desc)

type Status

type Status interface {
	String() string
	TimePoint() (start TimePoint, end TimePoint)
}

type TimePoint

type TimePoint interface {
	Status() Status
	Index() int
}

type TimePointManager

type TimePointManager interface {
	AddTimestamp(key string, timePoint TimePoint, timestamp time.Time, extraInfos ...Info)
	GetStatusTime(key string, status Status) (time.Duration, bool)
	GetTimestamp(key string, timePoint TimePoint) (time.Time, bool)
	GetTimestampsCache() map[string]Timestamps
	GetStatusCount(status Status) int64
	GetInfo(key string) (Info, bool)
}

var (
	Job           TimePointManager
	PromAPIClient prom_v1.API
)

func NewJobMetricTimePointManager

func NewJobMetricTimePointManager() TimePointManager

NewJobMetricTimePointManager Implementation of default job metric manager

type Timestamps

type Timestamps interface {
	GetStatusTime(status Status) (time.Duration, bool)
	GetTimestamp(timePoint TimePoint) (time.Time, bool)
	AddTimestamp(timePoint TimePoint, timestamp time.Time)
}