Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
img
LICENSE
README.md
meta.yaml

README.md

Go library for SignalFx

Metadata associated with the Go library for SignalFx can be found here. The relevant code for the library can be found here.

    import "github.com/signalfx/golib/sfxclient"

Package signalfx creates convenient Go functions and wrappers to send metrics to SignalFx.

The core of the library is HTTPDatapointSink which allows users to send metrics to SignalFx ad-hoc. A Scheduler is built on top of this to facility easy management of metrics for multiple SignalFx reporters at once in more complex libraries.

HTTPDatapointSink

The simplest way to send metrics to SignalFx is with HTTPDatapointSink. The only struct parameter that needs to be configured is AuthToken. To make it easier to create common Datapoint objects, wrappers exist for Gauge and Cumulative. An example of sending a hello world metric would look like this:

    func SendHelloWorld() {
        client := NewHTTPDatapointSink()
        client.AuthToken = "ABCDXYZ"
        ctx := context.Background()
        client.AddDatapoints(ctx, []*datapoint.Datapoint{
            GaugeF("hello.world", nil, 1.0),
        })
    }

Scheduler

To facilitate periodic sending of datapoints to SignalFx, a Scheduler abstraction exists. You can use this to report custom metrics to SignalFx at some periodic interval.

    type CustomApplication struct {
        queue chan int64
    }
    func (c *CustomApplication) Datapoints() []*datapoint.Datapoint {
        return []*datapoint.Datapoint {
          sfxclient.Gauge("queue.size", nil, len(queue)),
        }
    }
    func main() {
        scheduler := sfxclient.NewScheduler()
        scheduler.Sink.(*HTTPDatapointSink).AuthToken = "ABCD-XYZ"
        app := &CustomApplication{}
        scheduler.AddCallback(app)
        go scheduler.Schedule(context.Background())
    }

RollingBucket and CumulativeBucket

Because counting things and calculating percentiles like p99 or median are common operations, RollingBucket and CumulativeBucket exist to make this easier. They implement the Collector interface which allows users to add them to an existing Scheduler.

Index

Constants

const ClientVersion = "1.0"

ClientVersion is the version of this library and is embedded into the user agent

const DefaultReportingDelay = time.Second * 20

DefaultReportingDelay is the default interval Scheduler users to report metrics to SignalFx

const DefaultTimeout = time.Second * 5

DefaultTimeout is the default time to fail signalfx datapoint requests if they don't succeed

const IngestEndpointV2 = "https://ingest.signalfx.com/v2/datapoint"

IngestEndpointV2 is the v2 version of the signalfx ingest endpoint

const TokenHeaderName = "X-Sf-Token"

TokenHeaderName is the header key for the auth token in the HTTP request

Variables

var DefaultBucketWidth = time.Second * 20

DefaultBucketWidth is the default width that a RollingBucket should flush histogram values

var DefaultErrorHandler = func(err error) error {
	log.DefaultLogger.Log(log.Err, err, "Unable to handle error")
	return nil
}

DefaultErrorHandler is the default way to handle errors by a scheduler. It simply prints them to stdout

var DefaultHistogramSize = 80

DefaultHistogramSize is the default number of windows RollingBucket uses for created histograms

var DefaultMaxBufferSize = 100

DefaultMaxBufferSize is the default number of past bucket Quantile values RollingBucket saves until a Datapoints() call

var DefaultQuantiles = []float64{.25, .5, .9, .99}

DefaultQuantiles are the default set of percentiles RollingBucket should collect

var DefaultUserAgent = fmt.Sprintf("golib-sfxclient/%s (gover %s)", ClientVersion, runtime.Version())

DefaultUserAgent is the UserAgent string sent to signalfx

func Cumulative

func Cumulative(metricName string, dimensions map[string]string, val int64) *datapoint.Datapoint

Cumulative creates a SignalFx cumulative counter for integer values.

func CumulativeF

func CumulativeF(metricName string, dimensions map[string]string, val float64) *datapoint.Datapoint

CumulativeF creates a SignalFx cumulative counter for float values.

func CumulativeP

func CumulativeP(metricName string, dimensions map[string]string, val *int64) *datapoint.Datapoint

CumulativeP creates a SignalFx cumulative counter for integer values from a pointer that is loaded atomically.

func Gauge

func Gauge(metricName string, dimensions map[string]string, val int64) *datapoint.Datapoint

Gauge creates a SignalFx gauge for integer values.

func GaugeF

func GaugeF(metricName string, dimensions map[string]string, val float64) *datapoint.Datapoint

GaugeF creates a SignalFx gauge for floating point values.

type Collector

type Collector interface {
	Datapoints() []*datapoint.Datapoint
}

Collector is anything Scheduler can track that emits points

var GoMetricsSource Collector = &goMetrics{}

GoMetricsSource is a singleton Collector that collects basic go system stats. It currently collects from runtime.ReadMemStats and adds a few extra metrics like uptime of the process and other runtime package functions.

func NewMultiCollector

func NewMultiCollector(collectors ...Collector) Collector

NewMultiCollector returns a collector that is the aggregate of every given collector. It can be used to turn multiple collectors into a single collector.

type CumulativeBucket

type CumulativeBucket struct {
	MetricName string
	Dimensions map[string]string
}

A CumulativeBucket tracks groups of values, reporting the count/sum/sum of squares as a cumulative counter.

func (*CumulativeBucket) Add

func (b *CumulativeBucket) Add(val int64)

Add an item to the bucket, later reporting the result in the next report cycle.

func (*CumulativeBucket) Datapoints

func (b *CumulativeBucket) Datapoints() []*datapoint.Datapoint

Datapoints returns the count/sum/sumsquare datapoints, or nil if there is no set metric name

func (*CumulativeBucket) MultiAdd

func (b *CumulativeBucket) MultiAdd(res *Result)

MultiAdd many items into the bucket at once using a Result. This can be more efficient as it involves only a constant number of atomic operations.

type HTTPDatapointSink

type HTTPDatapointSink struct {
	AuthToken string
	UserAgent string
	Endpoint  string
	Client    http.Client
}

HTTPDatapointSink will accept signalfx datapoints and forward them to SignalFx via HTTP.

func NewHTTPDatapointSink

func NewHTTPDatapointSink() *HTTPDatapointSink

NewHTTPDatapointSink creates a default NewHTTPDatapointSink using package level constants as defaults, including an empty auth token. If sending directly to SiganlFx, you will be required to explicitly set the AuthToken

func (*HTTPDatapointSink) AddDatapoints

func (h *HTTPDatapointSink) AddDatapoints(ctx context.Context, points []*datapoint.Datapoint) (err error)

AddDatapoints forwards the datapoints to SignalFx.

type HashableCollector

type HashableCollector struct {
	Callback func() []*datapoint.Datapoint
}

HashableCollector is a Collector function that can be inserted into a hashmap. You can use it to wrap a functional callback and insert it into a Scheduler.

func CollectorFunc

func CollectorFunc(callback func() []*datapoint.Datapoint) *HashableCollector

CollectorFunc wraps a function to make it a Collector.

func (*HashableCollector) Datapoints

func (h *HashableCollector) Datapoints() []*datapoint.Datapoint

Datapoints calls the wrapped function.

type MultiCollector

type MultiCollector []Collector

MultiCollector acts like a datapoint collector over multiple collectors.

func (MultiCollector) Datapoints

func (m MultiCollector) Datapoints() []*datapoint.Datapoint

Datapoints returns the datapoints from every collector.

type Result

type Result struct {
	Count        int64
	Sum          int64
	SumOfSquares float64
}

Result is a cumulated result of items that can be added to a CumulativeBucket at once

func (\Result) Add

func (r *Result) Add(val int64)

Add a single number to the bucket. This does not use atomic operations and is not thread safe, but adding a finished Result into a CumulativeBucket is thread safe.

type RollingBucket

type RollingBucket struct {
	// MetricName is the metric name used when the RollingBucket is reported to SignalFx
	MetricName string
	// Dimensions are the dimensions used when the RollingBucket is reported to SignalFx
	Dimensions map[string]string
	// Quantiles are an array of values [0 - 1.0] that are the histogram quantiles reported to
	// SignalFx during a Datapoints() call.  For example, [.5] would only report the median.
	Quantiles []float64
	// BucketWidth is how long in time a bucket accumulates values before a flush is forced
	BucketWidth time.Duration
	// Hist is an efficient tracker of numeric values for a histogram
	Hist *gohistogram.NumericHistogram
	// MaxFlushBufferSize is the maximum size of a window to keep for the RollingBucket before
	// quantiles are dropped.  It is ideally close to len(quantiles) * 3 + 15
	MaxFlushBufferSize int
	// Timer is used to track time.Now() during default value add calls
	Timer timekeeper.TimeKeeper
}

RollingBucket keeps histogram style metrics over a BucketWidth window of time. It allows users to collect and report percentile metrics like like median or p99, as well as min/max/sum/count and sum of square from a set of points.

func NewRollingBucket

func NewRollingBucket(metricName string, dimensions map[string]string) *RollingBucket

NewRollingBucket creates a new RollingBucket using default values for Quantiles, BucketWidth, and the histogram tracker.

func (*RollingBucket) Add

func (r *RollingBucket) Add(v float64)

Add a value to the rolling bucket histogram. If the current time is already calculated, it may be more efficient to call AddAt in order to save another time.Time() call.

func (*RollingBucket) AddAt

func (r *RollingBucket) AddAt(v float64, t time.Time)

AddAt is like Add but also takes a time to pretend the value comes at.

func (*RollingBucket) Datapoints

func (r *RollingBucket) Datapoints() []*datapoint.Datapoint

Datapoints returns basic bucket stats every time and will only the first time called for each window return that window's points. For efficiency sake, Datapoints() will only return histogram window values once. Because of this, it is suggested to always forward datapoints returned by this call to SignalFx.

type Scheduler

type Scheduler struct {
	Sink             Sink
	Timer            timekeeper.TimeKeeper
	SendZeroTime     bool
	ErrorHandler     func(error) error
	ReportingDelayNs int64
}

A Scheduler reports metrics to SignalFx at some timely manner.

func NewScheduler

func NewScheduler() *Scheduler

NewScheduler creates a default SignalFx scheduler that can report metrics to SignalFx at some interval.

func (*Scheduler) AddCallback

func (s *Scheduler) AddCallback(db Collector)

AddCallback adds a collector to the default group.

func (*Scheduler) AddGroupedCallback

func (s *Scheduler) AddGroupedCallback(group string, db Collector)

AddGroupedCallback adds a collector to a specific group.

func (*Scheduler) DefaultDimensions

func (s *Scheduler) DefaultDimensions(dims map[string]string)

DefaultDimensions adds a dimension map that are appended to all metrics in the default group.

func (*Scheduler) GroupedDefaultDimensions

func (s *Scheduler) GroupedDefaultDimensions(group string, dims map[string]string)

GroupedDefaultDimensions adds default dimensions to a specific group.

func (*Scheduler) RemoveCallback

func (s *Scheduler) RemoveCallback(db Collector)

RemoveCallback removes a collector from the default group.

func (*Scheduler) RemoveGroupedCallback

func (s *Scheduler) RemoveGroupedCallback(group string, db Collector)

RemoveGroupedCallback removes a collector from a specific group.

func (*Scheduler) ReportOnce

func (s *Scheduler) ReportOnce(ctx context.Context) error

ReportOnce will report any metrics saved in this reporter to SignalFx

func (*Scheduler) ReportingDelay

func (s *Scheduler) ReportingDelay(delay time.Duration)

ReportingDelay sets the interval metrics are reported to SignalFx.

func (*Scheduler) Schedule

func (s *Scheduler) Schedule(ctx context.Context) error

Schedule will run until either the ErrorHandler returns an error or the context is canceled. This is intended to be run inside a goroutine.

func (*Scheduler) Var

func (s *Scheduler) Var() expvar.Var

Var returns an expvar variable that prints the values of the previously reported datapoints.

type Sink

type Sink interface {
	AddDatapoints(ctx context.Context, points []*datapoint.Datapoint) (err error)
}

Sink is anything that can receive points collected by a Scheduler. This can be useful for stubbing out your collector to test the points that will be sent to SignalFx.

type WithDimensions

type WithDimensions struct {
	Dimensions map[string]string
	Collector  Collector
}

WithDimensions adds dimensions on top of the datapoints of a collector. This can be used to take an existing Collector and include extra dimensions.

func (*WithDimensions) Datapoints

func (w *WithDimensions) Datapoints() []*datapoint.Datapoint

Datapoints calls datapoints and adds on Dimensions