statsd

package
Version: v5.0.2 Latest Latest
Warning

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

Go to latest
Published: Nov 29, 2021 License: MIT Imports: 12 Imported by: 1

README

Overview

Package statsd provides a Go dogstatsd client. Dogstatsd extends Statsd, adding tags and histograms.

Documentation

Overview

Package statsd provides a Go dogstatsd client. Dogstatsd extends the popular statsd, adding tags and histograms and pushing upstream to Datadog.

Refer to http://docs.datadoghq.com/guides/dogstatsd/ for information about DogStatsD.

statsd is based on go-statsd-client.

Index

Constants

View Source
const DefaultMaxAgentPayloadSize = 8192

DefaultMaxAgentPayloadSize is the default maximum payload size the agent can receive. This can be adjusted by changing dogstatsd_buffer_size in the agent configuration file datadog.yaml. This is also used as the optimal payload size for UDS datagrams.

View Source
const DefaultUDPBufferPoolSize = 2048

DefaultUDPBufferPoolSize is the default size of the buffer pool for UDP clients.

View Source
const DefaultUDSBufferPoolSize = 512

DefaultUDSBufferPoolSize is the default size of the buffer pool for UDS clients.

View Source
const ErrNoClient = noClientErr("statsd client is nil")

ErrNoClient is returned if statsd reporting methods are invoked on a nil client.

View Source
const MaxUDPPayloadSize = 65467

MaxUDPPayloadSize defines the maximum payload size for a UDP datagram. Its value comes from the calculation: 65535 bytes Max UDP datagram size - 8byte UDP header - 60byte max IP headers any number greater than that will see frames being cut out.

View Source
const OptimalUDPPayloadSize = 1432

OptimalUDPPayloadSize defines the optimal payload size for a UDP datagram, 1432 bytes is optimal for regular networks with an MTU of 1500 so datagrams don't get fragmented. It's generally recommended not to fragment UDP datagrams as losing a single fragment will cause the entire datagram to be lost.

View Source
const UnixAddressPrefix = "unix://"

UnixAddressPrefix holds the prefix to use to enable Unix Domain Socket traffic instead of UDP.

View Source
const WindowsPipeAddressPrefix = `\\.\pipe\`

WindowsPipeAddressPrefix holds the prefix to use to enable Windows Named Pipes traffic instead of UDP.

Variables

This section is empty.

Functions

This section is empty.

Types

type Client

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

A Client is a handle for sending messages to dogstatsd. It is safe to use one Client from multiple goroutines simultaneously.

func CloneWithExtraOptions

func CloneWithExtraOptions(c *Client, options ...Option) (*Client, error)

CloneWithExtraOptions create a new Client with extra options

func New

func New(addr string, options ...Option) (*Client, error)

New returns a pointer to a new Client given an addr in the format "hostname:port" for UDP, "unix:///path/to/socket" for UDS or "\\.\pipe\path\to\pipe" for Windows Named Pipes.

func NewWithWriter

func NewWithWriter(w io.WriteCloser, options ...Option) (*Client, error)

NewWithWriter creates a new Client with given writer. Writer is a io.WriteCloser

func (*Client) Close

func (c *Client) Close() error

Close the client connection.

func (*Client) Count

func (c *Client) Count(name string, value int64, tags []string, rate float64) error

Count tracks how many times something happened per second.

func (*Client) Decr

func (c *Client) Decr(name string, tags []string, rate float64) error

Decr is just Count of -1

func (*Client) Distribution

func (c *Client) Distribution(name string, value float64, tags []string, rate float64) error

Distribution tracks the statistical distribution of a set of values across your infrastructure.

func (*Client) Event

func (c *Client) Event(e *Event) error

Event sends the provided Event.

func (*Client) Flush

func (c *Client) Flush() error

Flush forces a flush of all the queued dogstatsd payloads This method is blocking and will not return until everything is sent through the network. In mutexMode, this will also block sampling new data to the client while the workers and sender are flushed.

func (*Client) Gauge

func (c *Client) Gauge(name string, value float64, tags []string, rate float64) error

Gauge measures the value of a metric at a particular time.

func (*Client) GetTelemetry

func (c *Client) GetTelemetry() Telemetry

GetTelemetry return the telemetry metrics for the client since it started.

func (*Client) Histogram

func (c *Client) Histogram(name string, value float64, tags []string, rate float64) error

Histogram tracks the statistical distribution of a set of values on each host.

func (*Client) Incr

func (c *Client) Incr(name string, tags []string, rate float64) error

Incr is just Count of 1

func (*Client) ServiceCheck

func (c *Client) ServiceCheck(sc *ServiceCheck) error

ServiceCheck sends the provided ServiceCheck.

func (*Client) Set

func (c *Client) Set(name string, value string, tags []string, rate float64) error

Set counts the number of unique elements in a group.

func (*Client) SimpleEvent

func (c *Client) SimpleEvent(title, text string) error

SimpleEvent sends an event with the provided title and text.

func (*Client) SimpleServiceCheck

func (c *Client) SimpleServiceCheck(name string, status ServiceCheckStatus) error

SimpleServiceCheck sends an serviceCheck with the provided name and status.

func (*Client) TimeInMilliseconds

func (c *Client) TimeInMilliseconds(name string, value float64, tags []string, rate float64) error

TimeInMilliseconds sends timing information in milliseconds. It is flushed by statsd with percentiles, mean and other info (https://github.com/etsy/statsd/blob/master/docs/metric_types.md#timing)

func (*Client) Timing

func (c *Client) Timing(name string, value time.Duration, tags []string, rate float64) error

Timing sends timing information, it is an alias for TimeInMilliseconds

type ClientInterface

type ClientInterface interface {
	// Gauge measures the value of a metric at a particular time.
	Gauge(name string, value float64, tags []string, rate float64) error

	// Count tracks how many times something happened per second.
	Count(name string, value int64, tags []string, rate float64) error

	// Histogram tracks the statistical distribution of a set of values on each host.
	Histogram(name string, value float64, tags []string, rate float64) error

	// Distribution tracks the statistical distribution of a set of values across your infrastructure.
	Distribution(name string, value float64, tags []string, rate float64) error

	// Decr is just Count of -1
	Decr(name string, tags []string, rate float64) error

	// Incr is just Count of 1
	Incr(name string, tags []string, rate float64) error

	// Set counts the number of unique elements in a group.
	Set(name string, value string, tags []string, rate float64) error

	// Timing sends timing information, it is an alias for TimeInMilliseconds
	Timing(name string, value time.Duration, tags []string, rate float64) error

	// TimeInMilliseconds sends timing information in milliseconds.
	// It is flushed by statsd with percentiles, mean and other info (https://github.com/etsy/statsd/blob/master/docs/metric_types.md#timing)
	TimeInMilliseconds(name string, value float64, tags []string, rate float64) error

	// Event sends the provided Event.
	Event(e *Event) error

	// SimpleEvent sends an event with the provided title and text.
	SimpleEvent(title, text string) error

	// ServiceCheck sends the provided ServiceCheck.
	ServiceCheck(sc *ServiceCheck) error

	// SimpleServiceCheck sends an serviceCheck with the provided name and status.
	SimpleServiceCheck(name string, status ServiceCheckStatus) error

	// Close the client connection.
	Close() error

	// Flush forces a flush of all the queued dogstatsd payloads.
	Flush() error
}

ClientInterface is an interface that exposes the common client functions for the purpose of being able to provide a no-op client or even mocking. This can aid downstream users' with their testing.

type Event

type Event struct {
	// Title of the event.  Required.
	Title string
	// Text is the description of the event.  Required.
	Text string
	// Timestamp is a timestamp for the event.  If not provided, the dogstatsd
	// server will set this to the current time.
	Timestamp time.Time
	// Hostname for the event.
	Hostname string
	// AggregationKey groups this event with others of the same key.
	AggregationKey string
	// Priority of the event.  Can be statsd.Low or statsd.Normal.
	Priority EventPriority
	// SourceTypeName is a source type for the event.
	SourceTypeName string
	// AlertType can be statsd.Info, statsd.Error, statsd.Warning, or statsd.Success.
	// If absent, the default value applied by the dogstatsd server is Info.
	AlertType EventAlertType
	// Tags for the event.
	Tags []string
}

An Event is an object that can be posted to your DataDog event stream.

func NewEvent

func NewEvent(title, text string) *Event

NewEvent creates a new event with the given title and text. Error checking against these values is done at send-time, or upon running e.Check.

func (*Event) Check

func (e *Event) Check() error

Check verifies that an event is valid.

type EventAlertType

type EventAlertType string

EventAlertType is the alert type for events

const (
	// Info is the "info" AlertType for events
	Info EventAlertType = "info"
	// Error is the "error" AlertType for events
	Error EventAlertType = "error"
	// Warning is the "warning" AlertType for events
	Warning EventAlertType = "warning"
	// Success is the "success" AlertType for events
	Success EventAlertType = "success"
)

type EventPriority

type EventPriority string

EventPriority is the event priority for events

const (
	// Normal is the "normal" Priority for events
	Normal EventPriority = "normal"
	// Low is the "low" Priority for events
	Low EventPriority = "low"
)

type NoOpClient

type NoOpClient struct{}

NoOpClient is a statsd client that does nothing. Can be useful in testing situations for library users.

func (*NoOpClient) Close

func (n *NoOpClient) Close() error

Close does nothing and returns nil

func (*NoOpClient) Count

func (n *NoOpClient) Count(name string, value int64, tags []string, rate float64) error

Count does nothing and returns nil

func (*NoOpClient) Decr

func (n *NoOpClient) Decr(name string, tags []string, rate float64) error

Decr does nothing and returns nil

func (*NoOpClient) Distribution

func (n *NoOpClient) Distribution(name string, value float64, tags []string, rate float64) error

Distribution does nothing and returns nil

func (*NoOpClient) Event

func (n *NoOpClient) Event(e *Event) error

Event does nothing and returns nil

func (*NoOpClient) Flush

func (n *NoOpClient) Flush() error

Flush does nothing and returns nil

func (*NoOpClient) Gauge

func (n *NoOpClient) Gauge(name string, value float64, tags []string, rate float64) error

Gauge does nothing and returns nil

func (*NoOpClient) Histogram

func (n *NoOpClient) Histogram(name string, value float64, tags []string, rate float64) error

Histogram does nothing and returns nil

func (*NoOpClient) Incr

func (n *NoOpClient) Incr(name string, tags []string, rate float64) error

Incr does nothing and returns nil

func (*NoOpClient) ServiceCheck

func (n *NoOpClient) ServiceCheck(sc *ServiceCheck) error

ServiceCheck does nothing and returns nil

func (*NoOpClient) Set

func (n *NoOpClient) Set(name string, value string, tags []string, rate float64) error

Set does nothing and returns nil

func (*NoOpClient) SimpleEvent

func (n *NoOpClient) SimpleEvent(title, text string) error

SimpleEvent does nothing and returns nil

func (*NoOpClient) SimpleServiceCheck

func (n *NoOpClient) SimpleServiceCheck(name string, status ServiceCheckStatus) error

SimpleServiceCheck does nothing and returns nil

func (*NoOpClient) TimeInMilliseconds

func (n *NoOpClient) TimeInMilliseconds(name string, value float64, tags []string, rate float64) error

TimeInMilliseconds does nothing and returns nil

func (*NoOpClient) Timing

func (n *NoOpClient) Timing(name string, value time.Duration, tags []string, rate float64) error

Timing does nothing and returns nil

type Option

type Option func(*Options) error

Option is a client option. Can return an error if validation fails.

func WithAggregationInterval

func WithAggregationInterval(interval time.Duration) Option

WithAggregationInterval sets the interval at which aggregated metrics are flushed. See WithClientSideAggregation and WithExtendedClientSideAggregation for more.

The default interval is 2s. The interval must divide the Agent reporting period (default=10s) evenly to reduce "aliasing" that can cause values to appear irregular/spiky.

For example a 3s aggregation interval will create spikes in the final graph: a application sending a count metric that increments at a constant 1000 time per second will appear noisy with an interval of 3s. This is because client-side aggregation would report every 3 seconds, while the agent is reporting every 10 seconds. This means in each agent bucket, the values are: 9000, 9000, 12000.

func WithBufferFlushInterval

func WithBufferFlushInterval(bufferFlushInterval time.Duration) Option

WithBufferFlushInterval sets the interval after which the current buffer is flushed.

A buffers are used to serialized data, they're flushed either when full (see WithMaxBytesPerPayload) or when it's been open for longer than this interval.

With apps sending a high number of metrics/events/service_checks the interval rarely timeout. But with slow sending apps increasing this value will reduce the number of payload sent on the wire as more data is serialized in the same payload.

Default is 100ms

func WithBufferPoolSize

func WithBufferPoolSize(bufferPoolSize int) Option

WithBufferPoolSize sets the size of the pool of buffers used to serialized metrics, events and service_checks.

The default, 0, will set the option to the optimal size for the transport protocol used: 2048 for UDP and named pipe and 512 for UDS.

func WithChannelMode

func WithChannelMode() Option

WithChannelMode make the client use channels to receive metrics

This determines how the client receive metrics from the app (for example when calling the `Gauge()` method). The client will either drop the metrics if its buffers are full (WithChannelMode option) or block the caller until the metric can be handled (WithMutexMode option). By default the client use mutexes.

WithChannelMode uses a channel (see WithChannelModeBufferSize to configure its size) to receive metrics and drops metrics if the channel is full. Sending metrics in this mode is much slower that WithMutexMode (because of the channel), but will not block the application. This mode is made for application using many goroutines, sending the same metrics, at a very high volume. The goal is to not slow down the application at the cost of dropping metrics and having a lower max throughput.

func WithChannelModeBufferSize

func WithChannelModeBufferSize(bufferSize int) Option

WithChannelModeBufferSize sets the size of the channel holding incoming metrics when WithChannelMode is used.

func WithClientSideAggregation

func WithClientSideAggregation() Option

WithClientSideAggregation enables client side aggregation for Gauges, Counts and Sets.

func WithExtendedClientSideAggregation

func WithExtendedClientSideAggregation() Option

WithExtendedClientSideAggregation enables client side aggregation for all types. This feature is only compatible with Agent's version >=6.25.0 && <7.0.0 or Agent's versions >=7.25.0.

func WithMaxBytesPerPayload

func WithMaxBytesPerPayload(MaxBytesPerPayload int) Option

WithMaxBytesPerPayload sets the maximum number of bytes a single payload can contain.

The deault value 0 which will set the option to the optimal size for the transport protocol used: 1432 for UDP and named pipe and 8192 for UDS.

func WithMaxMessagesPerPayload

func WithMaxMessagesPerPayload(maxMessagesPerPayload int) Option

WithMaxMessagesPerPayload sets the maximum number of metrics, events and/or service checks that a single payload can contain.

The default is 'math.MaxInt32' which will most likely let the WithMaxBytesPerPayload option take precedence. This option can be set to `1` to create an unbuffered client (each metrics/event/service check will be send in its own payload to the agent).

func WithMutexMode

func WithMutexMode() Option

WithMutexMode will use mutex to receive metrics from the app throught the API.

This determines how the client receive metrics from the app (for example when calling the `Gauge()` method). The client will either drop the metrics if its buffers are full (WithChannelMode option) or block the caller until the metric can be handled (WithMutexMode option). By default the client use mutexes.

WithMutexMode uses mutexes to receive metrics which is much faster than channels but can cause some lock contention when used with a high number of goroutines sendint the same metrics. Mutexes are sharded based on the metrics name which limit mutex contention when multiple goroutines send different metrics (see WithWorkersCount). This is the default behavior which will produce the best throughput.

func WithNamespace

func WithNamespace(namespace string) Option

WithNamespace sets a string to be prepend to all metrics, events and service checks name.

A '.' will automatically be added after the namespace if needed. For example a metrics 'test' with a namespace 'prod' will produce a final metric named 'prod.test'.

func WithSenderQueueSize

func WithSenderQueueSize(senderQueueSize int) Option

WithSenderQueueSize sets the size of the sender queue in number of buffers.

After data has been serialized in a buffer they're pushed to a queue that the sender will consume and then each one ot the agent.

The default value 0 will set the option to the optimal size for the transport protocol used: 2048 for UDP and named pipe and 512 for UDS.

func WithTags

func WithTags(tags []string) Option

WithTags sets global tags to be applied to every metrics, events and service checks.

func WithTelemetryAddr

func WithTelemetryAddr(addr string) Option

WithTelemetryAddr sets a different address for telemetry metrics. By default the same address as the client is used for telemetry.

More on this here: https://docs.datadoghq.com/developers/dogstatsd/high_throughput/#client-side-telemetry

func WithWorkersCount

func WithWorkersCount(workersCount int) Option

WithWorkersCount sets the number of workers that will be used to serialized data.

Those workers allow the use of multiple buffers at the same time (see WithBufferPoolSize) to reduce lock contention.

Default is 32.

func WithWriteTimeout

func WithWriteTimeout(writeTimeout time.Duration) Option

WithWriteTimeout sets the timeout for network communication with the Agent, after this interval a payload is dropped. This is only used for UDS and named pipes connection.

func WithoutClientSideAggregation

func WithoutClientSideAggregation() Option

WithoutClientSideAggregation disables client side aggregation.

func WithoutTelemetry

func WithoutTelemetry() Option

WithoutTelemetry disables the client telemetry.

More on this here: https://docs.datadoghq.com/developers/dogstatsd/high_throughput/#client-side-telemetry

type Options

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

Options contains the configuration options for a client.

type ServiceCheck

type ServiceCheck struct {
	// Name of the service check.  Required.
	Name string
	// Status of service check.  Required.
	Status ServiceCheckStatus
	// Timestamp is a timestamp for the serviceCheck.  If not provided, the dogstatsd
	// server will set this to the current time.
	Timestamp time.Time
	// Hostname for the serviceCheck.
	Hostname string
	// A message describing the current state of the serviceCheck.
	Message string
	// Tags for the serviceCheck.
	Tags []string
}

A ServiceCheck is an object that contains status of DataDog service check.

func NewServiceCheck

func NewServiceCheck(name string, status ServiceCheckStatus) *ServiceCheck

NewServiceCheck creates a new serviceCheck with the given name and status. Error checking against these values is done at send-time, or upon running sc.Check.

func (*ServiceCheck) Check

func (sc *ServiceCheck) Check() error

Check verifies that a service check is valid.

type ServiceCheckStatus

type ServiceCheckStatus byte

ServiceCheckStatus support

const (
	// Ok is the "ok" ServiceCheck status
	Ok ServiceCheckStatus = 0
	// Warn is the "warning" ServiceCheck status
	Warn ServiceCheckStatus = 1
	// Critical is the "critical" ServiceCheck status
	Critical ServiceCheckStatus = 2
	// Unknown is the "unknown" ServiceCheck status
	Unknown ServiceCheckStatus = 3
)

type Telemetry

type Telemetry struct {

	// TotalMetrics is the total number of metrics sent by the client before aggregation and sampling.
	TotalMetrics uint64
	// TotalMetricsGauge is the total number of gauges sent by the client before aggregation and sampling.
	TotalMetricsGauge uint64
	// TotalMetricsCount is the total number of counts sent by the client before aggregation and sampling.
	TotalMetricsCount uint64
	// TotalMetricsHistogram is the total number of histograms sent by the client before aggregation and sampling.
	TotalMetricsHistogram uint64
	// TotalMetricsDistribution is the total number of distributions sent by the client before aggregation and
	// sampling.
	TotalMetricsDistribution uint64
	// TotalMetricsSet is the total number of sets sent by the client before aggregation and sampling.
	TotalMetricsSet uint64
	// TotalMetricsTiming is the total number of timings sent by the client before aggregation and sampling.
	TotalMetricsTiming uint64
	// TotalEvents is the total number of events sent by the client before aggregation and sampling.
	TotalEvents uint64
	// TotalServiceChecks is the total number of service_checks sent by the client before aggregation and sampling.
	TotalServiceChecks uint64

	// TotalDroppedOnReceive is the total number metrics/event/service_checks dropped when using ChannelMode (see
	// WithChannelMode option).
	TotalDroppedOnReceive uint64

	// TotalPayloadsSent is the total number of payload (packet on the network) succesfully sent by the client. When
	// using UDP we don't know if packet dropped or not, so all packet are considered as succesfully sent.
	TotalPayloadsSent uint64
	// TotalPayloadsDropped is the total number of payload dropped by the client. This includes all cause of dropped
	// (TotalPayloadsDroppedQueueFull and TotalPayloadsDroppedWriter). When using UDP This won't includes the
	// network dropped.
	TotalPayloadsDropped uint64
	// TotalPayloadsDroppedWriter is the total number of payload dropped by the writer (when using UDS or named
	// pipe) due to network timeout or error.
	TotalPayloadsDroppedWriter uint64
	// TotalPayloadsDroppedQueueFull is the total number of payload dropped internally because the queue of payloads
	// waiting to be sent on the wire is full. This means the client is generating more metrics than can be sent on
	// the wire. If your app sends metrics in batch look at WithSenderQueueSize option to increase the queue size.
	TotalPayloadsDroppedQueueFull uint64

	// TotalBytesSent is the total number of bytes succesfully sent by the client. When using UDP we don't know if
	// packet dropped or not, so all packet are considered as succesfully sent.
	TotalBytesSent uint64
	// TotalBytesDropped is the total number of bytes dropped by the client. This includes all cause of dropped
	// (TotalBytesDroppedQueueFull and TotalBytesDroppedWriter). When using UDP This
	// won't includes the network dropped.
	TotalBytesDropped uint64
	// TotalBytesDroppedWriter is the total number of bytes dropped by the writer (when using UDS or named pipe) due
	// to network timeout or error.
	TotalBytesDroppedWriter uint64
	// TotalBytesDroppedQueueFull is the total number of bytes dropped internally because the queue of payloads
	// waiting to be sent on the wire is full. This means the client is generating more metrics than can be sent on
	// the wire. If your app sends metrics in batch look at WithSenderQueueSize option to increase the queue size.
	TotalBytesDroppedQueueFull uint64

	// AggregationNbContext is the total number of contexts flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContext uint64
	// AggregationNbContextGauge is the total number of contexts for gauges flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContextGauge uint64
	// AggregationNbContextCount is the total number of contexts for counts flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContextCount uint64
	// AggregationNbContextSet is the total number of contexts for sets flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContextSet uint64
	// AggregationNbContextHistogram is the total number of contexts for histograms flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContextHistogram uint64
	// AggregationNbContextDistribution is the total number of contexts for distributions flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContextDistribution uint64
	// AggregationNbContextTiming is the total number of contexts for timings flushed by the aggregator when either
	// WithClientSideAggregation or WithExtendedClientSideAggregation options are enabled.
	AggregationNbContextTiming uint64
}

Telemetry represents internal metrics about the client behavior since it started.

Jump to

Keyboard shortcuts

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