metrics

package
v0.0.0-...-7983b3b Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Apr 8, 2024 License: Apache-2.0 Imports: 29 Imported by: 0

README

Metrics

The agent offers different type of metric. Each metrics offers 2 methods addSample and flush.

  • addSample: add a new sample to the metrics.
  • flush: aggregate all samples received since the last flush and return a Series to be forwarded to the Datadog backend.
gauge

Gauge tracks the value of a metric. The last value received is the one returned by the flush method.

counter

Counter tracks how many times something happened per second. Counters are only used by DogStatsD and are very similar to Count: the main diffence is that they are sent as Rate.

count

Count is used to count the number of events that occur between 2 flushes. Each sample's value is added to the value that's flushed.

histogram

Histogram tracks the distribution of samples added over one flush period.

historate

Historate tracks the distribution of samples added over one flush period for "rate" like metrics. Warning this doesn't use the harmonic mean, beware of what it means when using it.

monotonic_count

MonotonicCount tracks a raw counter, based on increasing counter values. Samples that have a lower value than the previous sample are ignored (since it usually means that the underlying raw counter has been reset).

Example:

Submitting samples 2, 3, 6, 7 returns 5 (i.e. 7-2) on flush, then submitting samples 10, 11 on the same MonotonicCount returns 4 (i.e. 11-7) on the second flush.

percentile

Percentile tracks the distribution of samples added over one flush period. Designed to be globally accurate for percentiles.

Percentile is not usable yet; it is still undergoing development and testing.

rate

Rate tracks the rate of a metric over 2 successive flushes (ie: no metrics will be returned on the first flush).

set

Set tracks the number of unique elements in a set. This is only used by DogStatsD; you cannot create sets from an Agent check.

service_check

Service checks track the status of any service: OK, WARNING, CRITICAL, or UNKNOWN. The Agent does not aggregate service checks, it sends every check straight to Datadog's backend.

event

Events represent discrete moments in time, e.g. thrown exceptions, code deploys, etc. The Agent does not aggregate events, it sends every event straight to your Datadog event stream.

Documentation

Index

Constants

This section is empty.

Variables

View Source
var (
	DistributionMetricTypes = map[MetricType]struct{}{
		DistributionType: {},
	}
)

DistributionMetricTypes contains the MetricTypes that are used for percentiles

Functions

This section is empty.

Types

type APIMetricType

type APIMetricType int

APIMetricType represents an API metric type

const (
	APIGaugeType APIMetricType = iota
	APIRateType
	APICountType
)

Enumeration of the existing API metric types

func (APIMetricType) MarshalText

func (a APIMetricType) MarshalText() ([]byte, error)

MarshalText implements the encoding.TextMarshal interface to marshal an APIMetricType to a serialized byte slice

func (APIMetricType) String

func (a APIMetricType) String() string

String returns a string representation of APIMetricType

func (*APIMetricType) UnmarshalText

func (a *APIMetricType) UnmarshalText(buf []byte) error

UnmarshalText is a custom unmarshaller for APIMetricType (used for testing)

type AddSampleTelemetry

type AddSampleTelemetry struct {
	Total     telemetry.SimpleCounter
	Stateful  telemetry.SimpleCounter
	Stateless telemetry.SimpleCounter
}

AddSampleTelemetry counts number of new metrics added.

func (*AddSampleTelemetry) Inc

func (a *AddSampleTelemetry) Inc(isStateful bool)

Inc should be called once for each new metric added to the map.

isStateful should be the value returned by isStateful method for the new metric.

type CheckMetrics

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

CheckMetrics stores metrics for the check sampler.

This is similar to ContextMetrics, but provides additional facility to remove metrics.

Metrics can be requested to be removed by calling Expire(). Metrics that keep state between flushes (see Metric.isStateful) are kept for additional statefulTimeout seconds after expiration, as a precaution against checks that send metrics intermittently. Older stateful metrics need to be cleaned up by calling RemoveExpired().

func NewCheckMetrics

func NewCheckMetrics(expireMetrics bool, statefulTimeout time.Duration) CheckMetrics

NewCheckMetrics returns new CheckMetrics instance.

func (*CheckMetrics) AddSample

func (cm *CheckMetrics) AddSample(contextKey ckey.ContextKey, sample *MetricSample, timestamp float64, interval int64) error

AddSample adds a new sample to the metric with contextKey, initializing a new metric as necessary.

If contextKey is scheduled for removal (see Expire), it will be unscheduled.

See also ContextMetrics.AddSample().

func (*CheckMetrics) Expire

func (cm *CheckMetrics) Expire(contextKeys []ckey.ContextKey, timestamp float64)

Expire enables metric data for given context keys to be removed.

Metrics that do not keep state between flushes, will be removed immediately.

Metrics that do keep state, will be kept around for additional `cm.statefulTimeout` time after timestamp, before ultimately removed (See RemoveExpired). Call to AddSample will cancel delayed removal.

func (*CheckMetrics) Flush

func (cm *CheckMetrics) Flush(timestamp float64) ([]*Serie, map[ckey.ContextKey]error)

Flush flushes every metrics in the CheckMetrics (see ContextMetrics.Flush)

func (*CheckMetrics) RemoveExpired

func (cm *CheckMetrics) RemoveExpired(timestamp float64)

RemoveExpired removes stateful metrics that have expired before the given timestamp.

type CheckMetricsTelemetryAccumulator

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

CheckMetricsTelemetryAccumulator aggregates telemetry collected from multiple CheckMetrics instances.

func (*CheckMetricsTelemetryAccumulator) Flush

Flush updates telemetry counters based on aggregated statistics.

func (*CheckMetricsTelemetryAccumulator) VisitCheckMetrics

func (c *CheckMetricsTelemetryAccumulator) VisitCheckMetrics(cm *CheckMetrics)

VisitCheckMetrics adds metrics from CheckMetrics instance to the accumulator.

type ContextMetrics

type ContextMetrics map[ckey.ContextKey]Metric

ContextMetrics stores all the metrics by context key

func MakeContextMetrics

func MakeContextMetrics() ContextMetrics

MakeContextMetrics returns a new ContextMetrics

func (ContextMetrics) AddSample

func (m ContextMetrics) AddSample(contextKey ckey.ContextKey, sample *MetricSample, timestamp float64, interval int64, t *AddSampleTelemetry) error

AddSample add a sample to the current ContextMetrics and initialize a new metrics if needed.

func (ContextMetrics) Flush

func (m ContextMetrics) Flush(timestamp float64) ([]*Serie, map[ckey.ContextKey]error)

Flush flushes every metrics in the ContextMetrics. Returns the slice of Series and a map of errors by context key.

type Count

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

Count is used to count the number of events that occur between 2 flushes. Each sample's value is added to the value that's flushed

type Counter

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

Counter tracks how many times something happened per second. Counters are only used by DogStatsD and are very similar to Count: the main diffence is that they are sent as Rate.

func NewCounter

func NewCounter(interval int64) *Counter

NewCounter return a new initialized Counter

type Event

type Event struct {
	Title          string         `json:"msg_title"`
	Text           string         `json:"msg_text"`
	Ts             int64          `json:"timestamp"`
	Priority       EventPriority  `json:"priority,omitempty"`
	Host           string         `json:"host"`
	Tags           []string       `json:"tags,omitempty"`
	AlertType      EventAlertType `json:"alert_type,omitempty"`
	AggregationKey string         `json:"aggregation_key,omitempty"`
	SourceTypeName string         `json:"source_type_name,omitempty"`
	EventType      string         `json:"event_type,omitempty"`
	OriginID       string         `json:"-"`
	K8sOriginID    string         `json:"-"`
	Cardinality    string         `json:"-"`
	EventContext   *EventContext  `json:"context,omitempty"`
}

Event holds an event (w/ serialization to DD agent 5 intake format)

func (*Event) String

func (e *Event) String() string

Return a JSON string

type EventAlertType

type EventAlertType string

EventAlertType represents the alert type of an event

const (
	EventAlertTypeError   EventAlertType = "error"
	EventAlertTypeWarning EventAlertType = "warning"
	EventAlertTypeInfo    EventAlertType = "info"
	EventAlertTypeSuccess EventAlertType = "success"
)

Enumeration of the existing event alert types, and their values

func GetAlertTypeFromString

func GetAlertTypeFromString(val string) (EventAlertType, error)

GetAlertTypeFromString returns the EventAlertType from its string representation

type EventContext

type EventContext struct {
	SourceIdentifier   string                 `json:"source_identifier,omitempty" msg:"source_identifier,omitempty"`
	ElementIdentifiers []string               `json:"element_identifiers" msg:"element_identifiers"`
	Source             string                 `json:"source" msg:"source"`
	Category           string                 `json:"category" msg:"category"`
	Data               map[string]interface{} `json:"data,omitempty" msg:"data,omitempty"`
	SourceLinks        []SourceLink           `json:"source_links" msg:"source_links"`

} // [sts]

EventContext enriches the event with some more context and allows correlation to topology in StackState

type EventPriority

type EventPriority string

EventPriority represents the priority of an event

const (
	EventPriorityNormal EventPriority = "normal"
	EventPriorityLow    EventPriority = "low"
)

Enumeration of the existing event priorities, and their values

func GetEventPriorityFromString

func GetEventPriorityFromString(val string) (EventPriority, error)

GetEventPriorityFromString returns the EventPriority from its string representation

type Events

type Events []*Event

Events represents a list of events ready to be serialize

func (Events) CreateMarshalersBySourceType

func (events Events) CreateMarshalersBySourceType() []marshaler.StreamJSONMarshaler

CreateMarshalersBySourceType creates a collection of marshaler.StreamJSONMarshaler. Each StreamJSONMarshaler is composed of all events for a specific source type name.

func (Events) CreateSingleMarshaler

func (events Events) CreateSingleMarshaler() marshaler.StreamJSONMarshaler

CreateSingleMarshaler creates marshaler.StreamJSONMarshaler where each item is composed of all events for a specific source type name.

func (Events) Marshal

func (events Events) Marshal() ([]byte, error)

Marshal serialize events using agent-payload definition

func (Events) MarshalJSON

func (events Events) MarshalJSON() ([]byte, error)

MarshalJSON serializes events to JSON so it can be sent to the Agent 5 intake (we don't use the v1 event endpoint because it only supports 1 event per payload) FIXME(olivier): to be removed when v2 endpoints are available

func (Events) MarshalSplitCompress

func (events Events) MarshalSplitCompress(bufferContext *marshaler.BufferContext) ([]*[]byte, error)

MarshalSplitCompress not implemented

func (Events) SplitPayload

func (events Events) SplitPayload(times int) ([]marshaler.AbstractMarshaler, error)

SplitPayload breaks the payload into times number of pieces

type Gauge

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

Gauge tracks the value of a metric

type Histogram

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

Histogram tracks the distribution of samples added over one flush period

func NewHistogram

func NewHistogram(interval int64) *Histogram

NewHistogram returns a newly initialized histogram

type HistogramBucket

type HistogramBucket struct {
	Name            string
	Value           int64
	LowerBound      float64
	UpperBound      float64
	Monotonic       bool
	Tags            []string
	Host            string
	Timestamp       float64
	FlushFirstValue bool
}

HistogramBucket represents a prometheus/openmetrics histogram bucket

func (*HistogramBucket) GetHost

func (m *HistogramBucket) GetHost() string

GetHost returns the bucket host

func (*HistogramBucket) GetName

func (m *HistogramBucket) GetName() string

GetName returns the bucket name

func (*HistogramBucket) GetTags

GetTags returns the bucket tags.

type Historate

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

Historate tracks the distribution of samples added over one flush period for "rate" like metrics. Warning this doesn't use the harmonic mean, beware of what it means when using it.

func NewHistorate

func NewHistorate(interval int64) *Historate

NewHistorate returns a newly-initialized historate

type IntakeEvents

type IntakeEvents struct {
	Events []Event
}

IntakeEvents are used in the transactional batcher to keep state of events in a payload [sts]

func (IntakeEvents) IntakeFormat

func (ie IntakeEvents) IntakeFormat() map[string][]Event

IntakeFormat returns a map of events grouped by source type name

type Metric

type Metric interface {
	// contains filtered or unexported methods
}

Metric is the interface of all metric types

type MetricSample

type MetricSample struct {
	Name            string
	Value           float64
	RawValue        string
	Mtype           MetricType
	Tags            []string
	Host            string
	SampleRate      float64
	Timestamp       float64
	FlushFirstValue bool
	OriginID        string
	K8sOriginID     string
	Cardinality     string
}

MetricSample represents a raw metric sample

func (*MetricSample) Copy

func (m *MetricSample) Copy() *MetricSample

Copy returns a deep copy of the m MetricSample

func (*MetricSample) GetHost

func (m *MetricSample) GetHost() string

GetHost returns the metric sample host

func (*MetricSample) GetName

func (m *MetricSample) GetName() string

GetName returns the metric sample name

func (*MetricSample) GetTags

func (m *MetricSample) GetTags(tb *tagset.HashingTagsAccumulator)

GetTags returns the metric sample tags

type MetricSampleContext

type MetricSampleContext interface {
	GetName() string
	GetHost() string
	GetTags(*tagset.HashingTagsAccumulator)
}

MetricSampleContext allows to access a sample context data

type MetricSamplePool

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

MetricSamplePool is a pool of metrics sample

func NewMetricSamplePool

func NewMetricSamplePool(batchSize int) *MetricSamplePool

NewMetricSamplePool creates a new MetricSamplePool

func (*MetricSamplePool) GetBatch

func (m *MetricSamplePool) GetBatch() []MetricSample

GetBatch gets a batch of metric samples from the pool

func (*MetricSamplePool) PutBatch

func (m *MetricSamplePool) PutBatch(batch []MetricSample)

PutBatch puts a batch back into the pool

type MetricType

type MetricType int

MetricType is the representation of an aggregator metric type

const (
	GaugeType MetricType = iota
	RateType
	CountType
	MonotonicCountType
	CounterType
	HistogramType
	HistorateType
	SetType
	DistributionType
)

metric type constants enumeration

func (MetricType) String

func (m MetricType) String() string

String returns a string representation of MetricType

type MonotonicCount

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

MonotonicCount tracks a raw counter, based on increasing counter values. Samples that have a lower value than the previous sample are ignored (since it usually means that the underlying raw counter has been reset). Example:

submitting samples 2, 3, 6, 7 returns 5 (i.e. 7-2) on flush ;
then submitting samples 10, 11 on the same MonotonicCount returns 4 (i.e. 11-7) on flush

type NoSerieError

type NoSerieError struct{}

NoSerieError is the error returned by a metric when not enough samples have been submitted to generate a serie

func (NoSerieError) Error

func (e NoSerieError) Error() string

type Point

type Point struct {
	Ts    float64
	Value float64
}

Point represents a metric value at a specific time

func (*Point) MarshalJSON

func (p *Point) MarshalJSON() ([]byte, error)

MarshalJSON return a Point as an array of value (to be compatible with v1 API) FIXME(maxime): to be removed when v2 endpoints are available Note: it is not used with jsoniter, encodePoints takes over

func (*Point) UnmarshalJSON

func (p *Point) UnmarshalJSON(buf []byte) error

UnmarshalJSON is a custom unmarshaller for Point (used for testing)

type Rate

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

Rate tracks the rate of a metric over 2 successive flushes

type Serie

type Serie struct {
	Name           string          `json:"metric"`
	Points         []Point         `json:"points"`
	Tags           []string        `json:"tags"`
	Host           string          `json:"host"`
	Device         string          `json:"device,omitempty"` // FIXME(olivier): remove as soon as the v1 API can handle `device` as a regular tag
	MType          APIMetricType   `json:"type"`
	Interval       int64           `json:"interval"`
	SourceTypeName string          `json:"source_type_name,omitempty"`
	ContextKey     ckey.ContextKey `json:"-"`
	NameSuffix     string          `json:"-"`
}

Serie holds a timeseries (w/ json serialization to DD API format)

func (Serie) String

func (e Serie) String() string

String could be used for debug logging

type Series

type Series []*Serie

Series represents a list of Serie ready to be serialize

func (Series) DescribeItem

func (series Series) DescribeItem(i int) string

DescribeItem returns a text description for logs

func (Series) Len

func (series Series) Len() int

Len returns the number of items to marshal

func (Series) MarshalJSON

func (series Series) MarshalJSON() ([]byte, error)

MarshalJSON serializes timeseries to JSON so it can be sent to V1 endpoints FIXME(maxime): to be removed when v2 endpoints are available

func (Series) MarshalSplitCompress

func (series Series) MarshalSplitCompress(bufferContext *marshaler.BufferContext) ([]*[]byte, error)

MarshalSplitCompress uses the stream compressor to marshal and compress series payloads. If a compressed payload is larger than the max, a new payload will be generated. This method returns a slice of compressed protobuf marshaled MetricPayload objects.

func (Series) MarshalStrings

func (series Series) MarshalStrings() ([]string, [][]string)

MarshalStrings converts the timeseries to a sorted slice of string slices

func (Series) SplitPayload

func (series Series) SplitPayload(times int) ([]marshaler.AbstractMarshaler, error)

SplitPayload breaks the payload into, at least, "times" number of pieces

func (Series) WriteFooter

func (series Series) WriteFooter(stream *jsoniter.Stream) error

WriteFooter prints the payload footer for this type

func (Series) WriteHeader

func (series Series) WriteHeader(stream *jsoniter.Stream) error

WriteHeader writes the payload header for this type

func (Series) WriteItem

func (series Series) WriteItem(stream *jsoniter.Stream, i int) error

WriteItem prints the json representation of an item

type ServiceCheck

type ServiceCheck struct {
	CheckName   string             `json:"check"`
	Host        string             `json:"host_name"`
	Ts          int64              `json:"timestamp"`
	Status      ServiceCheckStatus `json:"status"`
	Message     string             `json:"message"`
	Tags        []string           `json:"tags"`
	OriginID    string             `json:"-"`
	K8sOriginID string             `json:"-"`
	Cardinality string             `json:"-"`
}

ServiceCheck holds a service check (w/ serialization to DD api format)

func (ServiceCheck) String

func (sc ServiceCheck) String() string

type ServiceCheckStatus

type ServiceCheckStatus int

ServiceCheckStatus represents the status associated with a service check

const (
	ServiceCheckOK       ServiceCheckStatus = iota
	ServiceCheckWarning  ServiceCheckStatus = 1
	ServiceCheckCritical ServiceCheckStatus = 2
	ServiceCheckUnknown  ServiceCheckStatus = 3
)

Enumeration of the existing service check statuses, and their values

func GetServiceCheckStatus

func GetServiceCheckStatus(val int) (ServiceCheckStatus, error)

GetServiceCheckStatus returns the ServiceCheckStatus from and integer value

func (ServiceCheckStatus) String

func (s ServiceCheckStatus) String() string

String returns a string representation of ServiceCheckStatus

type ServiceChecks

type ServiceChecks []*ServiceCheck

ServiceChecks represents a list of service checks ready to be serialize

func (ServiceChecks) DescribeItem

func (sc ServiceChecks) DescribeItem(i int) string

DescribeItem returns a text description for logs

func (ServiceChecks) Len

func (sc ServiceChecks) Len() int

Len returns the number of items to marshal

func (ServiceChecks) MarshalJSON

func (sc ServiceChecks) MarshalJSON() ([]byte, error)

MarshalJSON serializes service checks to JSON so it can be sent to V1 endpoints FIXME(olivier): to be removed when v2 endpoints are available

func (ServiceChecks) MarshalSplitCompress

func (sc ServiceChecks) MarshalSplitCompress(bufferContext *marshaler.BufferContext) ([]*[]byte, error)

MarshalSplitCompress not implemented

func (ServiceChecks) MarshalStrings

func (sc ServiceChecks) MarshalStrings() ([]string, [][]string)

MarshalStrings converts the service checks to a sorted slice of string slices

func (ServiceChecks) SplitPayload

func (sc ServiceChecks) SplitPayload(times int) ([]marshaler.AbstractMarshaler, error)

SplitPayload breaks the payload into times number of pieces

func (ServiceChecks) WriteFooter

func (sc ServiceChecks) WriteFooter(stream *jsoniter.Stream) error

WriteFooter prints the payload footer for this type

func (ServiceChecks) WriteHeader

func (sc ServiceChecks) WriteHeader(stream *jsoniter.Stream) error

WriteHeader writes the payload header for this type

func (ServiceChecks) WriteItem

func (sc ServiceChecks) WriteItem(stream *jsoniter.Stream, i int) error

WriteItem prints the json representation of an item

type Set

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

Set tracks the number of unique elements in a set. They are only use by DogStatsD

func NewSet

func NewSet() *Set

NewSet return a new initialized Set

type SketchPoint

type SketchPoint struct {
	Sketch *quantile.Sketch `json:"sketch"`
	Ts     int64            `json:"ts"`
}

A SketchPoint represents a quantile sketch at a specific time

type SketchSeries

type SketchSeries struct {
	Name       string          `json:"metric"`
	Tags       []string        `json:"tags"`
	Host       string          `json:"host"`
	Interval   int64           `json:"interval"`
	Points     []SketchPoint   `json:"points"`
	ContextKey ckey.ContextKey `json:"-"`
}

A SketchSeries is a timeseries of quantile sketches.

type SketchSeriesList

type SketchSeriesList []SketchSeries

A SketchSeriesList implements marshaler.Marshaler

func (SketchSeriesList) Marshal

func (sl SketchSeriesList) Marshal() ([]byte, error)

Marshal encodes this series list.

func (SketchSeriesList) MarshalJSON

func (sl SketchSeriesList) MarshalJSON() ([]byte, error)

MarshalJSON serializes sketch series to JSON. Quite slow, but hopefully this method is called only in the `agent check` command

func (SketchSeriesList) MarshalSplitCompress

func (sl SketchSeriesList) MarshalSplitCompress(bufferContext *marshaler.BufferContext) ([]*[]byte, error)

MarshalSplitCompress uses the stream compressor to marshal and compress sketch series payloads. If a compressed payload is larger than the max, a new payload will be generated. This method returns a slice of compressed protobuf marshaled gogen.SketchPayload objects. gogen.SketchPayload is not directly marshaled - instead it's contents are marshaled individually, packed with the appropriate protobuf metadata, and compressed in stream. The resulting payloads (when decompressed) are binary equal to the result of marshaling the whole object at once.

func (SketchSeriesList) SplitPayload

func (sl SketchSeriesList) SplitPayload(times int) ([]marshaler.AbstractMarshaler, error)

SplitPayload breaks the payload into times number of pieces

type SourceLink struct {
	Title string `json:"title" msg:"title"`
	URL   string `json:"url" msg:"url"`

} // [sts]

SourceLink points to links that may contain more information about this event

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL