prometheus

package
v1.1.1 Latest Latest
Warning

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

 Go to latest Published: Oct 4, 2019 License: Apache-2.0 Imports: 24 Imported by: 0

README

See go-doc.

Documentation

Overview

Package prometheus is the core instrumentation package. It provides metrics primitives to instrument code for monitoring. It also offers a registry for metrics. Sub-packages allow to expose the registered metrics via HTTP (package promhttp) or push them to a Pushgateway (package push). There is also a sub-package promauto, which provides metrics constructors with automatic registration.

All exported functions and methods are safe to be used concurrently unless specified otherwise.

A Basic Example

As a starting point, a very basic usage example: 

package main

import (
	"log"
	"net/http"

	"github.com/prometheus/client_golang/prometheus"
	"github.com/prometheus/client_golang/prometheus/promhttp"
)

var (
	cpuTemp = prometheus.NewGauge(prometheus.GaugeOpts{
		Name: "cpu_temperature_celsius",
		Help: "Current temperature of the CPU.",
	})
	hdFailures = prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Name: "hd_errors_total",
			Help: "Number of hard-disk errors.",
		},
		[]string{"device"},
	)
)

func init() {
	// Metrics have to be registered to be exposed:
	prometheus.MustRegister(cpuTemp)
	prometheus.MustRegister(hdFailures)
}

func main() {
	cpuTemp.Set(65.3)
	hdFailures.With(prometheus.Labels{"device":"/dev/sda"}).Inc()

	// The Handler function provides a default handler to expose metrics
	// via an HTTP server. "/metrics" is the usual endpoint for that.
	http.Handle("/metrics", promhttp.Handler())
	log.Fatal(http.ListenAndServe(":8080", nil))
}

This is a complete program that exports two metrics, a Gauge and a Counter, the latter with a label attached to turn it into a (one-dimensional) vector.

Metrics

The number of exported identifiers in this package might appear a bit overwhelming. However, in addition to the basic plumbing shown in the example above, you only need to understand the different metric types and their vector versions for basic usage. Furthermore, if you are not concerned with fine-grained control of when and how to register metrics with the registry, have a look at the promauto package, which will effectively allow you to ignore registration altogether in simple cases.

Above, you have already touched the Counter and the Gauge. There are two more advanced metric types: the Summary and Histogram. A more thorough description of those four metric types can be found in the Prometheus docs: https://prometheus.io/docs/concepts/metric_types/

A fifth "type" of metric is Untyped. It behaves like a Gauge, but signals the Prometheus server not to assume anything about its type.

In addition to the fundamental metric types Gauge, Counter, Summary, Histogram, and Untyped, a very important part of the Prometheus data model is the partitioning of samples along dimensions called labels, which results in metric vectors. The fundamental types are GaugeVec, CounterVec, SummaryVec, HistogramVec, and UntypedVec.

While only the fundamental metric types implement the Metric interface, both the metrics and their vector versions implement the Collector interface. A Collector manages the collection of a number of Metrics, but for convenience, a Metric can also “collect itself”. Note that Gauge, Counter, Summary, Histogram, and Untyped are interfaces themselves while GaugeVec, CounterVec, SummaryVec, HistogramVec, and UntypedVec are not.

To create instances of Metrics and their vector versions, you need a suitable …Opts struct, i.e. GaugeOpts, CounterOpts, SummaryOpts, HistogramOpts, or UntypedOpts.

Custom Collectors and constant Metrics

While you could create your own implementations of Metric, most likely you will only ever implement the Collector interface on your own. At a first glance, a custom Collector seems handy to bundle Metrics for common registration (with the prime example of the different metric vectors above, which bundle all the metrics of the same name but with different labels).

There is a more involved use case, too: If you already have metrics available, created outside of the Prometheus context, you don't need the interface of the various Metric types. You essentially want to mirror the existing numbers into Prometheus Metrics during collection. An own implementation of the Collector interface is perfect for that. You can create Metric instances “on the fly” using NewConstMetric, NewConstHistogram, and NewConstSummary (and their respective Must… versions). That will happen in the Collect method. The Describe method has to return separate Desc instances, representative of the “throw-away” metrics to be created later. NewDesc comes in handy to create those Desc instances. Alternatively, you could return no Desc at all, which will mark the Collector “unchecked”. No checks are performed at registration time, but metric consistency will still be ensured at scrape time, i.e. any inconsistencies will lead to scrape errors. Thus, with unchecked Collectors, the responsibility to not collect metrics that lead to inconsistencies in the total scrape result lies with the implementer of the Collector. While this is not a desirable state, it is sometimes necessary. The typical use case is a situation where the exact metrics to be returned by a Collector cannot be predicted at registration time, but the implementer has sufficient knowledge of the whole system to guarantee metric consistency.

The Collector example illustrates the use case. You can also look at the source code of the processCollector (mirroring process metrics), the goCollector (mirroring Go metrics), or the expvarCollector (mirroring expvar metrics) as examples that are used in this package itself.

If you just need to call a function to get a single float value to collect as a metric, GaugeFunc, CounterFunc, or UntypedFunc might be interesting shortcuts.

Advanced Uses of the Registry

While MustRegister is the by far most common way of registering a Collector, sometimes you might want to handle the errors the registration might cause. As suggested by the name, MustRegister panics if an error occurs. With the Register function, the error is returned and can be handled.

An error is returned if the registered Collector is incompatible or inconsistent with already registered metrics. The registry aims for consistency of the collected metrics according to the Prometheus data model. Inconsistencies are ideally detected at registration time, not at collect time. The former will usually be detected at start-up time of a program, while the latter will only happen at scrape time, possibly not even on the first scrape if the inconsistency only becomes relevant later. That is the main reason why a Collector and a Metric have to describe themselves to the registry.

So far, everything we did operated on the so-called default registry, as it can be found in the global DefaultRegisterer variable. With NewRegistry, you can create a custom registry, or you can even implement the Registerer or Gatherer interfaces yourself. The methods Register and Unregister work in the same way on a custom registry as the global functions Register and Unregister on the default registry.

There are a number of uses for custom registries: You can use registries with special properties, see NewPedanticRegistry. You can avoid global state, as it is imposed by the DefaultRegisterer. You can use multiple registries at the same time to expose different metrics in different ways. You can use separate registries for testing purposes.

Also note that the DefaultRegisterer comes registered with a Collector for Go runtime metrics (via NewGoCollector) and a Collector for process metrics (via NewProcessCollector). With a custom registry, you are in control and decide yourself about the Collectors to register.

HTTP Exposition

The Registry implements the Gatherer interface. The caller of the Gather method can then expose the gathered metrics in some way. Usually, the metrics are served via HTTP on the /metrics endpoint. That's happening in the example above. The tools to expose metrics via HTTP are in the promhttp sub-package.

Pushing to the Pushgateway

Function for pushing to the Pushgateway can be found in the push sub-package.

Graphite Bridge

Functions and examples to push metrics from a Gatherer to Graphite can be found in the graphite sub-package.

Other Means of Exposition

More ways of exposing metrics can easily be added by following the approaches of the existing implementations.

Index

Examples

Constants

View Source 
const (
	// DefMaxAge is the default duration for which observations stay
	// relevant.
	DefMaxAge time.Duration = 10 * time.Minute
	// DefAgeBuckets is the default number of buckets used to calculate the
	// age of observations.
	DefAgeBuckets = 5
	// DefBufCap is the standard buffer size for collecting Summary observations.
	DefBufCap = 500
)

Default values for SummaryOpts.

Variables

View Source 
var (
	DefaultRegisterer Registerer = defaultRegistry
	DefaultGatherer   Gatherer   = defaultRegistry
)

DefaultRegisterer and DefaultGatherer are the implementations of the Registerer and Gatherer interface a number of convenience functions in this package act on. Initially, both variables point to the same Registry, which has a process collector (currently on Linux only, see NewProcessCollector) and a Go collector (see NewGoCollector, in particular the note about stop-the-world implication with Go versions older than 1.9) already registered. This approach to keep default instances as global state mirrors the approach of other packages in the Go standard library. Note that there are caveats. Change the variables with caution and only if you understand the consequences. Users who want to avoid global state altogether should not use the convenience functions and act on custom instances instead.

View Source 
var (
	DefBuckets = []float64{.005, .01, .025, .05, .1, .25, .5, 1, 2.5, 5, 10}
)

DefBuckets are the default Histogram buckets. The default buckets are tailored to broadly measure the response time (in seconds) of a network service. Most likely, however, you will be required to define buckets customized to your use case.

Functions

func BuildFQName 

func BuildFQName(namespace, subsystem, name string) string

BuildFQName joins the given three name components by "_". Empty name components are ignored. If the name parameter itself is empty, an empty string is returned, no matter what. Metric implementations included in this library use this function internally to generate the fully-qualified metric name from the name component in their Opts. Users of the library will only need this function if they implement their own Metric or instantiate a Desc (with NewDesc) directly.

func DescribeByCollect added in v0.9.0 

func DescribeByCollect(c Collector, descs chan<- *Desc)

DescribeByCollect is a helper to implement the Describe method of a custom Collector. It collects the metrics from the provided Collector and sends their descriptors to the provided channel.

If a Collector collects the same metrics throughout its lifetime, its Describe method can simply be implemented as: 

func (c customCollector) Describe(ch chan<- *Desc) {
	DescribeByCollect(c, ch)
}

However, this will not work if the metrics collected change dynamically over the lifetime of the Collector in a way that their combined set of descriptors changes as well. The shortcut implementation will then violate the contract of the Describe method. If a Collector sometimes collects no metrics at all (for example vectors like CounterVec, GaugeVec, etc., which only collect metrics after a metric with a fully specified label set has been accessed), it might even get registered as an unchecked Collector (cf. the Register method of the Registerer interface). Hence, only use this shortcut implementation of Describe if you are certain to fulfill the contract.

The Collector example demonstrates a use of DescribeByCollect.

func ExponentialBuckets 

func ExponentialBuckets(start, factor float64, count int) []float64

ExponentialBuckets creates 'count' buckets, where the lowest bucket has an upper bound of 'start' and each following bucket's upper bound is 'factor' times the previous bucket's upper bound. The final +Inf bucket is not counted and not included in the returned slice. The returned slice is meant to be used for the Buckets field of HistogramOpts.

The function panics if 'count' is 0 or negative, if 'start' is 0 or negative, or if 'factor' is less than or equal 1.

func LinearBuckets 

func LinearBuckets(start, width float64, count int) []float64

LinearBuckets creates 'count' buckets, each 'width' wide, where the lowest bucket has an upper bound of 'start'. The final +Inf bucket is not counted and not included in the returned slice. The returned slice is meant to be used for the Buckets field of HistogramOpts.

The function panics if 'count' is zero or negative.

func MustRegister 

func MustRegister(cs ...Collector)

MustRegister registers the provided Collectors with the DefaultRegisterer and panics if any error occurs.

MustRegister is a shortcut for DefaultRegisterer.MustRegister(cs...). See there for more details.

func Register 

func Register(c Collector) error

Register registers the provided Collector with the DefaultRegisterer.

Register is a shortcut for DefaultRegisterer.Register(c). See there for more details.

Example 
package main

import (
	"fmt"
	"net/http"

	"github.com/prometheus/client_golang/prometheus"
	"github.com/prometheus/client_golang/prometheus/promhttp"
)

func main() {
	// Imagine you have a worker pool and want to count the tasks completed.
	taskCounter := prometheus.NewCounter(prometheus.CounterOpts{
		Subsystem: "worker_pool",
		Name:      "completed_tasks_total",
		Help:      "Total number of tasks completed.",
	})
	// This will register fine.
	if err := prometheus.Register(taskCounter); err != nil {
		fmt.Println(err)
	} else {
		fmt.Println("taskCounter registered.")
	}
	// Don't forget to tell the HTTP server about the Prometheus handler.
	// (In a real program, you still need to start the HTTP server...)
	http.Handle("/metrics", promhttp.Handler())

	// Now you can start workers and give every one of them a pointer to
	// taskCounter and let it increment it whenever it completes a task.
	taskCounter.Inc() // This has to happen somewhere in the worker code.

	// But wait, you want to see how individual workers perform. So you need
	// a vector of counters, with one element for each worker.
	taskCounterVec := prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Subsystem: "worker_pool",
			Name:      "completed_tasks_total",
			Help:      "Total number of tasks completed.",
		},
		[]string{"worker_id"},
	)

	// Registering will fail because we already have a metric of that name.
	if err := prometheus.Register(taskCounterVec); err != nil {
		fmt.Println("taskCounterVec not registered:", err)
	} else {
		fmt.Println("taskCounterVec registered.")
	}

	// To fix, first unregister the old taskCounter.
	if prometheus.Unregister(taskCounter) {
		fmt.Println("taskCounter unregistered.")
	}

	// Try registering taskCounterVec again.
	if err := prometheus.Register(taskCounterVec); err != nil {
		fmt.Println("taskCounterVec not registered:", err)
	} else {
		fmt.Println("taskCounterVec registered.")
	}
	// Bummer! Still doesn't work.

	// Prometheus will not allow you to ever export metrics with
	// inconsistent help strings or label names. After unregistering, the
	// unregistered metrics will cease to show up in the /metrics HTTP
	// response, but the registry still remembers that those metrics had
	// been exported before. For this example, we will now choose a
	// different name. (In a real program, you would obviously not export
	// the obsolete metric in the first place.)
	taskCounterVec = prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Subsystem: "worker_pool",
			Name:      "completed_tasks_by_id",
			Help:      "Total number of tasks completed.",
		},
		[]string{"worker_id"},
	)
	if err := prometheus.Register(taskCounterVec); err != nil {
		fmt.Println("taskCounterVec not registered:", err)
	} else {
		fmt.Println("taskCounterVec registered.")
	}
	// Finally it worked!

	// The workers have to tell taskCounterVec their id to increment the
	// right element in the metric vector.
	taskCounterVec.WithLabelValues("42").Inc() // Code from worker 42.

	// Each worker could also keep a reference to their own counter element
	// around. Pick the counter at initialization time of the worker.
	myCounter := taskCounterVec.WithLabelValues("42") // From worker 42 initialization code.
	myCounter.Inc()                                   // Somewhere in the code of that worker.

	// Note that something like WithLabelValues("42", "spurious arg") would
	// panic (because you have provided too many label values). If you want
	// to get an error instead, use GetMetricWithLabelValues(...) instead.
	notMyCounter, err := taskCounterVec.GetMetricWithLabelValues("42", "spurious arg")
	if err != nil {
		fmt.Println("Worker initialization failed:", err)
	}
	if notMyCounter == nil {
		fmt.Println("notMyCounter is nil.")
	}

	// A different (and somewhat tricky) approach is to use
	// ConstLabels. ConstLabels are pairs of label names and label values
	// that never change. Each worker creates and registers an own Counter
	// instance where the only difference is in the value of the
	// ConstLabels. Those Counters can all be registered because the
	// different ConstLabel values guarantee that each worker will increment
	// a different Counter metric.
	counterOpts := prometheus.CounterOpts{
		Subsystem:   "worker_pool",
		Name:        "completed_tasks",
		Help:        "Total number of tasks completed.",
		ConstLabels: prometheus.Labels{"worker_id": "42"},
	}
	taskCounterForWorker42 := prometheus.NewCounter(counterOpts)
	if err := prometheus.Register(taskCounterForWorker42); err != nil {
		fmt.Println("taskCounterVForWorker42 not registered:", err)
	} else {
		fmt.Println("taskCounterForWorker42 registered.")
	}
	// Obviously, in real code, taskCounterForWorker42 would be a member
	// variable of a worker struct, and the "42" would be retrieved with a
	// GetId() method or something. The Counter would be created and
	// registered in the initialization code of the worker.

	// For the creation of the next Counter, we can recycle
	// counterOpts. Just change the ConstLabels.
	counterOpts.ConstLabels = prometheus.Labels{"worker_id": "2001"}
	taskCounterForWorker2001 := prometheus.NewCounter(counterOpts)
	if err := prometheus.Register(taskCounterForWorker2001); err != nil {
		fmt.Println("taskCounterVForWorker2001 not registered:", err)
	} else {
		fmt.Println("taskCounterForWorker2001 registered.")
	}

	taskCounterForWorker2001.Inc()
	taskCounterForWorker42.Inc()
	taskCounterForWorker2001.Inc()

	// Yet another approach would be to turn the workers themselves into
	// Collectors and register them. See the Collector example for details.

}

Output:

taskCounter registered.
taskCounterVec not registered: a previously registered descriptor with the same fully-qualified name as Desc{fqName: "worker_pool_completed_tasks_total", help: "Total number of tasks completed.", constLabels: {}, variableLabels: [worker_id]} has different label names or a different help string
taskCounter unregistered.
taskCounterVec not registered: a previously registered descriptor with the same fully-qualified name as Desc{fqName: "worker_pool_completed_tasks_total", help: "Total number of tasks completed.", constLabels: {}, variableLabels: [worker_id]} has different label names or a different help string
taskCounterVec registered.
Worker initialization failed: inconsistent label cardinality: expected 1 label values but got 2 in []string{"42", "spurious arg"}
notMyCounter is nil.
taskCounterForWorker42 registered.
taskCounterForWorker2001 registered.

func Unregister 

func Unregister(c Collector) bool

Unregister removes the registration of the provided Collector from the DefaultRegisterer.

Unregister is a shortcut for DefaultRegisterer.Unregister(c). See there for more details.

func WriteToTextfile added in v0.9.1 

func WriteToTextfile(filename string, g Gatherer) error

WriteToTextfile calls Gather on the provided Gatherer, encodes the result in the Prometheus text format, and writes it to a temporary file. Upon success, the temporary file is renamed to the provided filename.

This is intended for use with the textfile collector of the node exporter. Note that the node exporter expects the filename to be suffixed with ".prom".

Types

type AlreadyRegisteredError 

type AlreadyRegisteredError struct {
	ExistingCollector, NewCollector Collector
}

AlreadyRegisteredError is returned by the Register method if the Collector to be registered has already been registered before, or a different Collector that collects the same metrics has been registered before. Registration fails in that case, but you can detect from the kind of error what has happened. The error contains fields for the existing Collector and the (rejected) new Collector that equals the existing one. This can be used to find out if an equal Collector has been registered before and switch over to using the old one, as demonstrated in the example.

Example 
package main

import (
	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	reqCounter := prometheus.NewCounter(prometheus.CounterOpts{
		Name: "requests_total",
		Help: "The total number of requests served.",
	})
	if err := prometheus.Register(reqCounter); err != nil {
		if are, ok := err.(prometheus.AlreadyRegisteredError); ok {
			// A counter for that metric has been registered before.
			// Use the old counter from now on.
			reqCounter = are.ExistingCollector.(prometheus.Counter)
		} else {
			// Something else went wrong!
			panic(err)
		}
	}
	reqCounter.Inc()
}

Output:

func (AlreadyRegisteredError) Error 

func (err AlreadyRegisteredError) Error() string

type Collector 

type Collector interface {
	// Describe sends the super-set of all possible descriptors of metrics
	// collected by this Collector to the provided channel and returns once
	// the last descriptor has been sent. The sent descriptors fulfill the
	// consistency and uniqueness requirements described in the Desc
	// documentation.
	//
	// It is valid if one and the same Collector sends duplicate
	// descriptors. Those duplicates are simply ignored. However, two
	// different Collectors must not send duplicate descriptors.
	//
	// Sending no descriptor at all marks the Collector as “unchecked”,
	// i.e. no checks will be performed at registration time, and the
	// Collector may yield any Metric it sees fit in its Collect method.
	//
	// This method idempotently sends the same descriptors throughout the
	// lifetime of the Collector. It may be called concurrently and
	// therefore must be implemented in a concurrency safe way.
	//
	// If a Collector encounters an error while executing this method, it
	// must send an invalid descriptor (created with NewInvalidDesc) to
	// signal the error to the registry.
	Describe(chan<- *Desc)
	// Collect is called by the Prometheus registry when collecting
	// metrics. The implementation sends each collected metric via the
	// provided channel and returns once the last metric has been sent. The
	// descriptor of each sent metric is one of those returned by Describe
	// (unless the Collector is unchecked, see above). Returned metrics that
	// share the same descriptor must differ in their variable label
	// values.
	//
	// This method may be called concurrently and must therefore be
	// implemented in a concurrency safe way. Blocking occurs at the expense
	// of total performance of rendering all registered metrics. Ideally,
	// Collector implementations support concurrent readers.
	Collect(chan<- Metric)
}

Collector is the interface implemented by anything that can be used by Prometheus to collect metrics. A Collector has to be registered for collection. See Registerer.Register.

The stock metrics provided by this package (Gauge, Counter, Summary, Histogram, Untyped) are also Collectors (which only ever collect one metric, namely itself). An implementer of Collector may, however, collect multiple metrics in a coordinated fashion and/or create metrics on the fly. Examples for collectors already implemented in this library are the metric vectors (i.e. collection of multiple instances of the same Metric but with different label values) like GaugeVec or SummaryVec, and the ExpvarCollector.

Example 
package main

import (
	"log"
	"net/http"

	"github.com/prometheus/client_golang/prometheus"
	"github.com/prometheus/client_golang/prometheus/promhttp"
)

// ClusterManager is an example for a system that might have been built without
// Prometheus in mind. It models a central manager of jobs running in a
// cluster. Thus, we implement a custom Collector called
// ClusterManagerCollector, which collects information from a ClusterManager
// using its provided methods and turns them into Prometheus Metrics for
// collection.
//
// An additional challenge is that multiple instances of the ClusterManager are
// run within the same binary, each in charge of a different zone. We need to
// make use of wrapping Registerers to be able to register each
// ClusterManagerCollector instance with Prometheus.
type ClusterManager struct {
	Zone string
	// Contains many more fields not listed in this example.
}

// ReallyExpensiveAssessmentOfTheSystemState is a mock for the data gathering a
// real cluster manager would have to do. Since it may actually be really
// expensive, it must only be called once per collection. This implementation,
// obviously, only returns some made-up data.
func (c *ClusterManager) ReallyExpensiveAssessmentOfTheSystemState() (
	oomCountByHost map[string]int, ramUsageByHost map[string]float64,
) {
	// Just example fake data.
	oomCountByHost = map[string]int{
		"foo.example.org": 42,
		"bar.example.org": 2001,
	}
	ramUsageByHost = map[string]float64{
		"foo.example.org": 6.023e23,
		"bar.example.org": 3.14,
	}
	return
}

// ClusterManagerCollector implements the Collector interface.
type ClusterManagerCollector struct {
	ClusterManager *ClusterManager
}

// Descriptors used by the ClusterManagerCollector below.
var (
	oomCountDesc = prometheus.NewDesc(
		"clustermanager_oom_crashes_total",
		"Number of OOM crashes.",
		[]string{"host"}, nil,
	)
	ramUsageDesc = prometheus.NewDesc(
		"clustermanager_ram_usage_bytes",
		"RAM usage as reported to the cluster manager.",
		[]string{"host"}, nil,
	)
)

// Describe is implemented with DescribeByCollect. That's possible because the
// Collect method will always return the same two metrics with the same two
// descriptors.
func (cc ClusterManagerCollector) Describe(ch chan<- *prometheus.Desc) {
	prometheus.DescribeByCollect(cc, ch)
}

// Collect first triggers the ReallyExpensiveAssessmentOfTheSystemState. Then it
// creates constant metrics for each host on the fly based on the returned data.
//
// Note that Collect could be called concurrently, so we depend on
// ReallyExpensiveAssessmentOfTheSystemState to be concurrency-safe.
func (cc ClusterManagerCollector) Collect(ch chan<- prometheus.Metric) {
	oomCountByHost, ramUsageByHost := cc.ClusterManager.ReallyExpensiveAssessmentOfTheSystemState()
	for host, oomCount := range oomCountByHost {
		ch <- prometheus.MustNewConstMetric(
			oomCountDesc,
			prometheus.CounterValue,
			float64(oomCount),
			host,
		)
	}
	for host, ramUsage := range ramUsageByHost {
		ch <- prometheus.MustNewConstMetric(
			ramUsageDesc,
			prometheus.GaugeValue,
			ramUsage,
			host,
		)
	}
}

// NewClusterManager first creates a Prometheus-ignorant ClusterManager
// instance. Then, it creates a ClusterManagerCollector for the just created
// ClusterManager. Finally, it registers the ClusterManagerCollector with a
// wrapping Registerer that adds the zone as a label. In this way, the metrics
// collected by different ClusterManagerCollectors do not collide.
func NewClusterManager(zone string, reg prometheus.Registerer) *ClusterManager {
	c := &ClusterManager{
		Zone: zone,
	}
	cc := ClusterManagerCollector{ClusterManager: c}
	prometheus.WrapRegistererWith(prometheus.Labels{"zone": zone}, reg).MustRegister(cc)
	return c
}

func main() {
	// Since we are dealing with custom Collector implementations, it might
	// be a good idea to try it out with a pedantic registry.
	reg := prometheus.NewPedanticRegistry()

	// Construct cluster managers. In real code, we would assign them to
	// variables to then do something with them.
	NewClusterManager("db", reg)
	NewClusterManager("ca", reg)

	// Add the standard process and Go metrics to the custom registry.
	reg.MustRegister(
		prometheus.NewProcessCollector(prometheus.ProcessCollectorOpts{}),
		prometheus.NewGoCollector(),
	)

	http.Handle("/metrics", promhttp.HandlerFor(reg, promhttp.HandlerOpts{}))
	log.Fatal(http.ListenAndServe(":8080", nil))
}

Output:

func NewBuildInfoCollector added in v1.1.1 

func NewBuildInfoCollector() Collector

NewBuildInfoCollector returns a collector collecting a single metric "go_build_info" with the constant value 1 and three labels "path", "version", and "checksum". Their label values contain the main module path, version, and checksum, respectively. The labels will only have meaningful values if the binary is built with Go module support and from source code retrieved from the source repository (rather than the local file system). This is usually accomplished by building from outside of GOPATH, specifying the full address of the main package, e.g. "GO111MODULE=on go run github.com/prometheus/client_golang/examples/random". If built without Go module support, all label values will be "unknown". If built with Go module support but using the source code from the local file system, the "path" will be set appropriately, but "checksum" will be empty and "version" will be "(devel)".

This collector uses only the build information for the main module. See https://github.com/povilasv/prommod for an example of a collector for the module dependencies.

func NewExpvarCollector 

func NewExpvarCollector(exports map[string]*Desc) Collector

NewExpvarCollector returns a newly allocated expvar Collector that still has to be registered with a Prometheus registry.

An expvar Collector collects metrics from the expvar interface. It provides a quick way to expose numeric values that are already exported via expvar as Prometheus metrics. Note that the data models of expvar and Prometheus are fundamentally different, and that the expvar Collector is inherently slower than native Prometheus metrics. Thus, the expvar Collector is probably great for experiments and prototying, but you should seriously consider a more direct implementation of Prometheus metrics for monitoring production systems.

The exports map has the following meaning:

The keys in the map correspond to expvar keys, i.e. for every expvar key you want to export as Prometheus metric, you need an entry in the exports map. The descriptor mapped to each key describes how to export the expvar value. It defines the name and the help string of the Prometheus metric proxying the expvar value. The type will always be Untyped.

For descriptors without variable labels, the expvar value must be a number or a bool. The number is then directly exported as the Prometheus sample value. (For a bool, 'false' translates to 0 and 'true' to 1). Expvar values that are not numbers or bools are silently ignored.

If the descriptor has one variable label, the expvar value must be an expvar map. The keys in the expvar map become the various values of the one Prometheus label. The values in the expvar map must be numbers or bools again as above.

For descriptors with more than one variable label, the expvar must be a nested expvar map, i.e. where the values of the topmost map are maps again etc. until a depth is reached that corresponds to the number of labels. The leaves of that structure must be numbers or bools as above to serve as the sample values.

Anything that does not fit into the scheme above is silently ignored.

Example 
package main

import (
	"expvar"
	"fmt"
	"sort"
	"strings"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	expvarCollector := prometheus.NewExpvarCollector(map[string]*prometheus.Desc{
		"memstats": prometheus.NewDesc(
			"expvar_memstats",
			"All numeric memstats as one metric family. Not a good role-model, actually... ;-)",
			[]string{"type"}, nil,
		),
		"lone-int": prometheus.NewDesc(
			"expvar_lone_int",
			"Just an expvar int as an example.",
			nil, nil,
		),
		"http-request-map": prometheus.NewDesc(
			"expvar_http_request_total",
			"How many http requests processed, partitioned by status code and http method.",
			[]string{"code", "method"}, nil,
		),
	})
	prometheus.MustRegister(expvarCollector)

	// The Prometheus part is done here. But to show that this example is
	// doing anything, we have to manually export something via expvar.  In
	// real-life use-cases, some library would already have exported via
	// expvar what we want to re-export as Prometheus metrics.
	expvar.NewInt("lone-int").Set(42)
	expvarMap := expvar.NewMap("http-request-map")
	var (
		expvarMap1, expvarMap2                             expvar.Map
		expvarInt11, expvarInt12, expvarInt21, expvarInt22 expvar.Int
	)
	expvarMap1.Init()
	expvarMap2.Init()
	expvarInt11.Set(3)
	expvarInt12.Set(13)
	expvarInt21.Set(11)
	expvarInt22.Set(212)
	expvarMap1.Set("POST", &expvarInt11)
	expvarMap1.Set("GET", &expvarInt12)
	expvarMap2.Set("POST", &expvarInt21)
	expvarMap2.Set("GET", &expvarInt22)
	expvarMap.Set("404", &expvarMap1)
	expvarMap.Set("200", &expvarMap2)
	// Results in the following expvar map:
	// "http-request-count": {"200": {"POST": 11, "GET": 212}, "404": {"POST": 3, "GET": 13}}

	// Let's see what the scrape would yield, but exclude the memstats metrics.
	metricStrings := []string{}
	metric := dto.Metric{}
	metricChan := make(chan prometheus.Metric)
	go func() {
		expvarCollector.Collect(metricChan)
		close(metricChan)
	}()
	for m := range metricChan {
		if !strings.Contains(m.Desc().String(), "expvar_memstats") {
			metric.Reset()
			m.Write(&metric)
			metricStrings = append(metricStrings, metric.String())
		}
	}
	sort.Strings(metricStrings)
	for _, s := range metricStrings {
		fmt.Println(strings.TrimRight(s, " "))
	}
}

Output:

label:<name:"code" value:"200" > label:<name:"method" value:"GET" > untyped:<value:212 >
label:<name:"code" value:"200" > label:<name:"method" value:"POST" > untyped:<value:11 >
label:<name:"code" value:"404" > label:<name:"method" value:"GET" > untyped:<value:13 >
label:<name:"code" value:"404" > label:<name:"method" value:"POST" > untyped:<value:3 >
untyped:<value:42 >

func NewGoCollector 

func NewGoCollector() Collector

NewGoCollector returns a collector that exports metrics about the current Go process. This includes memory stats. To collect those, runtime.ReadMemStats is called. This requires to “stop the world”, which usually only happens for garbage collection (GC). Take the following implications into account when deciding whether to use the Go collector:

1. The performance impact of stopping the world is the more relevant the more frequently metrics are collected. However, with Go1.9 or later the stop-the-world time per metrics collection is very short (~25µs) so that the performance impact will only matter in rare cases. However, with older Go versions, the stop-the-world duration depends on the heap size and can be quite significant (~1.7 ms/GiB as per https://go-review.googlesource.com/c/go/+/34937).

2. During an ongoing GC, nothing else can stop the world. Therefore, if the metrics collection happens to coincide with GC, it will only complete after GC has finished. Usually, GC is fast enough to not cause problems. However, with a very large heap, GC might take multiple seconds, which is enough to cause scrape timeouts in common setups. To avoid this problem, the Go collector will use the memstats from a previous collection if runtime.ReadMemStats takes more than 1s. However, if there are no previously collected memstats, or their collection is more than 5m ago, the collection will block until runtime.ReadMemStats succeeds. (The problem might be solved in Go1.13, see https://github.com/golang/go/issues/19812 for the related Go issue.)

func NewProcessCollector 

func NewProcessCollector(opts ProcessCollectorOpts) Collector

NewProcessCollector returns a collector which exports the current state of process metrics including CPU, memory and file descriptor usage as well as the process start time. The detailed behavior is defined by the provided ProcessCollectorOpts. The zero value of ProcessCollectorOpts creates a collector for the current process with an empty namespace string and no error reporting.

The collector only works on operating systems with a Linux-style proc filesystem and on Microsoft Windows. On other operating systems, it will not collect any metrics.

type Counter 

type Counter interface {
	Metric
	Collector

	// Inc increments the counter by 1. Use Add to increment it by arbitrary
	// non-negative values.
	Inc()
	// Add adds the given value to the counter. It panics if the value is <
	// 0.
	Add(float64)
}

Counter is a Metric that represents a single numerical value that only ever goes up. That implies that it cannot be used to count items whose number can also go down, e.g. the number of currently running goroutines. Those "counters" are represented by Gauges.

A Counter is typically used to count requests served, tasks completed, errors occurred, etc.

To create Counter instances, use NewCounter.

func NewCounter 

func NewCounter(opts CounterOpts) Counter

NewCounter creates a new Counter based on the provided CounterOpts.

The returned implementation tracks the counter value in two separate variables, a float64 and a uint64. The latter is used to track calls of the Inc method and calls of the Add method with a value that can be represented as a uint64. This allows atomic increments of the counter with optimal performance. (It is common to have an Inc call in very hot execution paths.) Both internal tracking values are added up in the Write method. This has to be taken into account when it comes to precision and overflow behavior.

type CounterFunc 

type CounterFunc interface {
	Metric
	Collector
}

CounterFunc is a Counter whose value is determined at collect time by calling a provided function.

To create CounterFunc instances, use NewCounterFunc.

func NewCounterFunc 

func NewCounterFunc(opts CounterOpts, function func() float64) CounterFunc

NewCounterFunc creates a new CounterFunc based on the provided CounterOpts. The value reported is determined by calling the given function from within the Write method. Take into account that metric collection may happen concurrently. If that results in concurrent calls to Write, like in the case where a CounterFunc is directly registered with Prometheus, the provided function must be concurrency-safe. The function should also honor the contract for a Counter (values only go up, not down), but compliance will not be checked.

type CounterOpts 

type CounterOpts Opts

CounterOpts is an alias for Opts. See there for doc comments.

type CounterVec 

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

CounterVec is a Collector that bundles a set of Counters that all share the same Desc, but have different values for their variable labels. This is used if you want to count the same thing partitioned by various dimensions (e.g. number of HTTP requests, partitioned by response code and method). Create instances with NewCounterVec.

Example 
package main

import (
	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	httpReqs := prometheus.NewCounterVec(
		prometheus.CounterOpts{
			Name: "http_requests_total",
			Help: "How many HTTP requests processed, partitioned by status code and HTTP method.",
		},
		[]string{"code", "method"},
	)
	prometheus.MustRegister(httpReqs)

	httpReqs.WithLabelValues("404", "POST").Add(42)

	// If you have to access the same set of labels very frequently, it
	// might be good to retrieve the metric only once and keep a handle to
	// it. But beware of deletion of that metric, see below!
	m := httpReqs.WithLabelValues("200", "GET")
	for i := 0; i < 1000000; i++ {
		m.Inc()
	}
	// Delete a metric from the vector. If you have previously kept a handle
	// to that metric (as above), future updates via that handle will go
	// unseen (even if you re-create a metric with the same label set
	// later).
	httpReqs.DeleteLabelValues("200", "GET")
	// Same thing with the more verbose Labels syntax.
	httpReqs.Delete(prometheus.Labels{"method": "GET", "code": "200"})
}

Output:

func NewCounterVec 

func NewCounterVec(opts CounterOpts, labelNames []string) *CounterVec

NewCounterVec creates a new CounterVec based on the provided CounterOpts and partitioned by the given label names.

func (*CounterVec) CurryWith added in v0.9.0 

func (v *CounterVec) CurryWith(labels Labels) (*CounterVec, error)

CurryWith returns a vector curried with the provided labels, i.e. the returned vector has those labels pre-set for all labeled operations performed on it. The cardinality of the curried vector is reduced accordingly. The order of the remaining labels stays the same (just with the curried labels taken out of the sequence – which is relevant for the (GetMetric)WithLabelValues methods). It is possible to curry a curried vector, but only with labels not yet used for currying before.

The metrics contained in the CounterVec are shared between the curried and uncurried vectors. They are just accessed differently. Curried and uncurried vectors behave identically in terms of collection. Only one must be registered with a given registry (usually the uncurried version). The Reset method deletes all metrics, even if called on a curried vector.

func (CounterVec) Delete added in v0.9.0 

func (m CounterVec) Delete(labels Labels) bool

Delete deletes the metric where the variable labels are the same as those passed in as labels. It returns true if a metric was deleted.

It is not an error if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc. However, such inconsistent Labels can never match an actual metric, so the method will always return false in that case.

This method is used for the same purpose as DeleteLabelValues(...string). See there for pros and cons of the two methods.

func (CounterVec) DeleteLabelValues added in v0.9.0 

func (m CounterVec) DeleteLabelValues(lvs ...string) bool

DeleteLabelValues removes the metric where the variable labels are the same as those passed in as labels (same order as the VariableLabels in Desc). It returns true if a metric was deleted.

It is not an error if the number of label values is not the same as the number of VariableLabels in Desc. However, such inconsistent label count can never match an actual metric, so the method will always return false in that case.

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider Delete(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the CounterVec example.

func (*CounterVec) GetMetricWith 

func (v *CounterVec) GetMetricWith(labels Labels) (Counter, error)

GetMetricWith returns the Counter for the given Labels map (the label names must match those of the VariableLabels in Desc). If that label map is accessed for the first time, a new Counter is created. Implications of creating a Counter without using it and keeping the Counter for later use are the same as for GetMetricWithLabelValues.

An error is returned if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc (minus any curried labels).

This method is used for the same purpose as GetMetricWithLabelValues(...string). See there for pros and cons of the two methods.

func (*CounterVec) GetMetricWithLabelValues 

func (v *CounterVec) GetMetricWithLabelValues(lvs ...string) (Counter, error)

GetMetricWithLabelValues returns the Counter for the given slice of label values (same order as the VariableLabels in Desc). If that combination of label values is accessed for the first time, a new Counter is created.

It is possible to call this method without using the returned Counter to only create the new Counter but leave it at its starting value 0. See also the SummaryVec example.

Keeping the Counter for later use is possible (and should be considered if performance is critical), but keep in mind that Reset, DeleteLabelValues and Delete can be used to delete the Counter from the CounterVec. In that case, the Counter will still exist, but it will not be exported anymore, even if a Counter with the same label values is created later.

An error is returned if the number of label values is not the same as the number of VariableLabels in Desc (minus any curried labels).

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider GetMetricWith(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the GaugeVec example.

func (*CounterVec) MustCurryWith added in v0.9.0 

func (v *CounterVec) MustCurryWith(labels Labels) *CounterVec

MustCurryWith works as CurryWith but panics where CurryWith would have returned an error.

func (*CounterVec) With 

func (v *CounterVec) With(labels Labels) Counter

With works as GetMetricWith, but panics where GetMetricWithLabels would have returned an error. Not returning an error allows shortcuts like 

myVec.With(prometheus.Labels{"code": "404", "method": "GET"}).Add(42)

func (*CounterVec) WithLabelValues 

func (v *CounterVec) WithLabelValues(lvs ...string) Counter

WithLabelValues works as GetMetricWithLabelValues, but panics where GetMetricWithLabelValues would have returned an error. Not returning an error allows shortcuts like 

myVec.WithLabelValues("404", "GET").Add(42)

type Desc 

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

Desc is the descriptor used by every Prometheus Metric. It is essentially the immutable meta-data of a Metric. The normal Metric implementations included in this package manage their Desc under the hood. Users only have to deal with Desc if they use advanced features like the ExpvarCollector or custom Collectors and Metrics.

Descriptors registered with the same registry have to fulfill certain consistency and uniqueness criteria if they share the same fully-qualified name: They must have the same help string and the same label names (aka label dimensions) in each, constLabels and variableLabels, but they must differ in the values of the constLabels.

Descriptors that share the same fully-qualified names and the same label values of their constLabels are considered equal.

Use NewDesc to create new Desc instances.

func NewDesc 

func NewDesc(fqName, help string, variableLabels []string, constLabels Labels) *Desc

NewDesc allocates and initializes a new Desc. Errors are recorded in the Desc and will be reported on registration time. variableLabels and constLabels can be nil if no such labels should be set. fqName must not be empty.

variableLabels only contain the label names. Their label values are variable and therefore not part of the Desc. (They are managed within the Metric.)

For constLabels, the label values are constant. Therefore, they are fully specified in the Desc. See the Collector example for a usage pattern.

func NewInvalidDesc 

func NewInvalidDesc(err error) *Desc

NewInvalidDesc returns an invalid descriptor, i.e. a descriptor with the provided error set. If a collector returning such a descriptor is registered, registration will fail with the provided error. NewInvalidDesc can be used by a Collector to signal inability to describe itself.

func (*Desc) String 

func (d *Desc) String() string

type Gatherer 

type Gatherer interface {
	// Gather calls the Collect method of the registered Collectors and then
	// gathers the collected metrics into a lexicographically sorted slice
	// of uniquely named MetricFamily protobufs. Gather ensures that the
	// returned slice is valid and self-consistent so that it can be used
	// for valid exposition. As an exception to the strict consistency
	// requirements described for metric.Desc, Gather will tolerate
	// different sets of label names for metrics of the same metric family.
	//
	// Even if an error occurs, Gather attempts to gather as many metrics as
	// possible. Hence, if a non-nil error is returned, the returned
	// MetricFamily slice could be nil (in case of a fatal error that
	// prevented any meaningful metric collection) or contain a number of
	// MetricFamily protobufs, some of which might be incomplete, and some
	// might be missing altogether. The returned error (which might be a
	// MultiError) explains the details. Note that this is mostly useful for
	// debugging purposes. If the gathered protobufs are to be used for
	// exposition in actual monitoring, it is almost always better to not
	// expose an incomplete result and instead disregard the returned
	// MetricFamily protobufs in case the returned error is non-nil.
	Gather() ([]*dto.MetricFamily, error)
}

Gatherer is the interface for the part of a registry in charge of gathering the collected metrics into a number of MetricFamilies. The Gatherer interface comes with the same general implication as described for the Registerer interface.

type GathererFunc 

type GathererFunc func() ([]*dto.MetricFamily, error)

GathererFunc turns a function into a Gatherer.

func (GathererFunc) Gather 

func (gf GathererFunc) Gather() ([]*dto.MetricFamily, error)

Gather implements Gatherer.

type Gatherers 

type Gatherers []Gatherer

Gatherers is a slice of Gatherer instances that implements the Gatherer interface itself. Its Gather method calls Gather on all Gatherers in the slice in order and returns the merged results. Errors returned from the Gather calls are all returned in a flattened MultiError. Duplicate and inconsistent Metrics are skipped (first occurrence in slice order wins) and reported in the returned error.

Gatherers can be used to merge the Gather results from multiple Registries. It also provides a way to directly inject existing MetricFamily protobufs into the gathering by creating a custom Gatherer with a Gather method that simply returns the existing MetricFamily protobufs. Note that no registration is involved (in contrast to Collector registration), so obviously registration-time checks cannot happen. Any inconsistencies between the gathered MetricFamilies are reported as errors by the Gather method, and inconsistent Metrics are dropped. Invalid parts of the MetricFamilies (e.g. syntactically invalid metric or label names) will go undetected.

Example 
package main

import (
	"bytes"
	"fmt"
	"strings"

	"github.com/prometheus/common/expfmt"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	reg := prometheus.NewRegistry()
	temp := prometheus.NewGaugeVec(
		prometheus.GaugeOpts{
			Name: "temperature_kelvin",
			Help: "Temperature in Kelvin.",
		},
		[]string{"location"},
	)
	reg.MustRegister(temp)
	temp.WithLabelValues("outside").Set(273.14)
	temp.WithLabelValues("inside").Set(298.44)

	var parser expfmt.TextParser

	text := `
# TYPE humidity_percent gauge
# HELP humidity_percent Humidity in %.
humidity_percent{location="outside"} 45.4
humidity_percent{location="inside"} 33.2
# TYPE temperature_kelvin gauge
# HELP temperature_kelvin Temperature in Kelvin.
temperature_kelvin{location="somewhere else"} 4.5
`

	parseText := func() ([]*dto.MetricFamily, error) {
		parsed, err := parser.TextToMetricFamilies(strings.NewReader(text))
		if err != nil {
			return nil, err
		}
		var result []*dto.MetricFamily
		for _, mf := range parsed {
			result = append(result, mf)
		}
		return result, nil
	}

	gatherers := prometheus.Gatherers{
		reg,
		prometheus.GathererFunc(parseText),
	}

	gathering, err := gatherers.Gather()
	if err != nil {
		fmt.Println(err)
	}

	out := &bytes.Buffer{}
	for _, mf := range gathering {
		if _, err := expfmt.MetricFamilyToText(out, mf); err != nil {
			panic(err)
		}
	}
	fmt.Print(out.String())
	fmt.Println("----------")

	// Note how the temperature_kelvin metric family has been merged from
	// different sources. Now try
	text = `
# TYPE humidity_percent gauge
# HELP humidity_percent Humidity in %.
humidity_percent{location="outside"} 45.4
humidity_percent{location="inside"} 33.2
# TYPE temperature_kelvin gauge
# HELP temperature_kelvin Temperature in Kelvin.
# Duplicate metric:
temperature_kelvin{location="outside"} 265.3
 # Missing location label (note that this is undesirable but valid):
temperature_kelvin 4.5
`

	gathering, err = gatherers.Gather()
	if err != nil {
		fmt.Println(err)
	}
	// Note that still as many metrics as possible are returned:
	out.Reset()
	for _, mf := range gathering {
		if _, err := expfmt.MetricFamilyToText(out, mf); err != nil {
			panic(err)
		}
	}
	fmt.Print(out.String())

}

Output:

# HELP humidity_percent Humidity in %.
# TYPE humidity_percent gauge
humidity_percent{location="inside"} 33.2
humidity_percent{location="outside"} 45.4
# HELP temperature_kelvin Temperature in Kelvin.
# TYPE temperature_kelvin gauge
temperature_kelvin{location="inside"} 298.44
temperature_kelvin{location="outside"} 273.14
temperature_kelvin{location="somewhere else"} 4.5
----------
collected metric "temperature_kelvin" { label:<name:"location" value:"outside" > gauge:<value:265.3 > } was collected before with the same name and label values
# HELP humidity_percent Humidity in %.
# TYPE humidity_percent gauge
humidity_percent{location="inside"} 33.2
humidity_percent{location="outside"} 45.4
# HELP temperature_kelvin Temperature in Kelvin.
# TYPE temperature_kelvin gauge
temperature_kelvin 4.5
temperature_kelvin{location="inside"} 298.44
temperature_kelvin{location="outside"} 273.14

func (Gatherers) Gather 

func (gs Gatherers) Gather() ([]*dto.MetricFamily, error)

Gather implements Gatherer.

type Gauge 

type Gauge interface {
	Metric
	Collector

	// Set sets the Gauge to an arbitrary value.
	Set(float64)
	// Inc increments the Gauge by 1. Use Add to increment it by arbitrary
	// values.
	Inc()
	// Dec decrements the Gauge by 1. Use Sub to decrement it by arbitrary
	// values.
	Dec()
	// Add adds the given value to the Gauge. (The value can be negative,
	// resulting in a decrease of the Gauge.)
	Add(float64)
	// Sub subtracts the given value from the Gauge. (The value can be
	// negative, resulting in an increase of the Gauge.)
	Sub(float64)

	// SetToCurrentTime sets the Gauge to the current Unix time in seconds.
	SetToCurrentTime()
}

Gauge is a Metric that represents a single numerical value that can arbitrarily go up and down.

A Gauge is typically used for measured values like temperatures or current memory usage, but also "counts" that can go up and down, like the number of running goroutines.

To create Gauge instances, use NewGauge.

Example 
package main

import (
	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	opsQueued := prometheus.NewGauge(prometheus.GaugeOpts{
		Namespace: "our_company",
		Subsystem: "blob_storage",
		Name:      "ops_queued",
		Help:      "Number of blob storage operations waiting to be processed.",
	})
	prometheus.MustRegister(opsQueued)

	// 10 operations queued by the goroutine managing incoming requests.
	opsQueued.Add(10)
	// A worker goroutine has picked up a waiting operation.
	opsQueued.Dec()
	// And once more...
	opsQueued.Dec()
}

Output:

func NewGauge 

func NewGauge(opts GaugeOpts) Gauge

NewGauge creates a new Gauge based on the provided GaugeOpts.

The returned implementation is optimized for a fast Set method. If you have a choice for managing the value of a Gauge via Set vs. Inc/Dec/Add/Sub, pick the former. For example, the Inc method of the returned Gauge is slower than the Inc method of a Counter returned by NewCounter. This matches the typical scenarios for Gauges and Counters, where the former tends to be Set-heavy and the latter Inc-heavy.

type GaugeFunc 

type GaugeFunc interface {
	Metric
	Collector
}

GaugeFunc is a Gauge whose value is determined at collect time by calling a provided function.

To create GaugeFunc instances, use NewGaugeFunc.

Example 
package main

import (
	"fmt"
	"runtime"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	if err := prometheus.Register(prometheus.NewGaugeFunc(
		prometheus.GaugeOpts{
			Subsystem: "runtime",
			Name:      "goroutines_count",
			Help:      "Number of goroutines that currently exist.",
		},
		func() float64 { return float64(runtime.NumGoroutine()) },
	)); err == nil {
		fmt.Println("GaugeFunc 'goroutines_count' registered.")
	}
	// Note that the count of goroutines is a gauge (and not a counter) as
	// it can go up and down.

}

Output:

GaugeFunc 'goroutines_count' registered.

func NewGaugeFunc 

func NewGaugeFunc(opts GaugeOpts, function func() float64) GaugeFunc

NewGaugeFunc creates a new GaugeFunc based on the provided GaugeOpts. The value reported is determined by calling the given function from within the Write method. Take into account that metric collection may happen concurrently. If that results in concurrent calls to Write, like in the case where a GaugeFunc is directly registered with Prometheus, the provided function must be concurrency-safe.

type GaugeOpts 

type GaugeOpts Opts

GaugeOpts is an alias for Opts. See there for doc comments.

type GaugeVec 

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

GaugeVec is a Collector that bundles a set of Gauges that all share the same Desc, but have different values for their variable labels. This is used if you want to count the same thing partitioned by various dimensions (e.g. number of operations queued, partitioned by user and operation type). Create instances with NewGaugeVec.

Example 
package main

import (
	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	opsQueued := prometheus.NewGaugeVec(
		prometheus.GaugeOpts{
			Namespace: "our_company",
			Subsystem: "blob_storage",
			Name:      "ops_queued",
			Help:      "Number of blob storage operations waiting to be processed, partitioned by user and type.",
		},
		[]string{
			// Which user has requested the operation?
			"user",
			// Of what type is the operation?
			"type",
		},
	)
	prometheus.MustRegister(opsQueued)

	// Increase a value using compact (but order-sensitive!) WithLabelValues().
	opsQueued.WithLabelValues("bob", "put").Add(4)
	// Increase a value with a map using WithLabels. More verbose, but order
	// doesn't matter anymore.
	opsQueued.With(prometheus.Labels{"type": "delete", "user": "alice"}).Inc()
}

Output:

func NewGaugeVec 

func NewGaugeVec(opts GaugeOpts, labelNames []string) *GaugeVec

NewGaugeVec creates a new GaugeVec based on the provided GaugeOpts and partitioned by the given label names.

func (*GaugeVec) CurryWith added in v0.9.0 

func (v *GaugeVec) CurryWith(labels Labels) (*GaugeVec, error)

CurryWith returns a vector curried with the provided labels, i.e. the returned vector has those labels pre-set for all labeled operations performed on it. The cardinality of the curried vector is reduced accordingly. The order of the remaining labels stays the same (just with the curried labels taken out of the sequence – which is relevant for the (GetMetric)WithLabelValues methods). It is possible to curry a curried vector, but only with labels not yet used for currying before.

The metrics contained in the GaugeVec are shared between the curried and uncurried vectors. They are just accessed differently. Curried and uncurried vectors behave identically in terms of collection. Only one must be registered with a given registry (usually the uncurried version). The Reset method deletes all metrics, even if called on a curried vector.

func (GaugeVec) Delete added in v0.9.0 

func (m GaugeVec) Delete(labels Labels) bool

Delete deletes the metric where the variable labels are the same as those passed in as labels. It returns true if a metric was deleted.

It is not an error if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc. However, such inconsistent Labels can never match an actual metric, so the method will always return false in that case.

This method is used for the same purpose as DeleteLabelValues(...string). See there for pros and cons of the two methods.

func (GaugeVec) DeleteLabelValues added in v0.9.0 

func (m GaugeVec) DeleteLabelValues(lvs ...string) bool

DeleteLabelValues removes the metric where the variable labels are the same as those passed in as labels (same order as the VariableLabels in Desc). It returns true if a metric was deleted.

It is not an error if the number of label values is not the same as the number of VariableLabels in Desc. However, such inconsistent label count can never match an actual metric, so the method will always return false in that case.

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider Delete(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the CounterVec example.

func (*GaugeVec) GetMetricWith 

func (v *GaugeVec) GetMetricWith(labels Labels) (Gauge, error)

GetMetricWith returns the Gauge for the given Labels map (the label names must match those of the VariableLabels in Desc). If that label map is accessed for the first time, a new Gauge is created. Implications of creating a Gauge without using it and keeping the Gauge for later use are the same as for GetMetricWithLabelValues.

An error is returned if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc (minus any curried labels).

This method is used for the same purpose as GetMetricWithLabelValues(...string). See there for pros and cons of the two methods.

func (*GaugeVec) GetMetricWithLabelValues 

func (v *GaugeVec) GetMetricWithLabelValues(lvs ...string) (Gauge, error)

GetMetricWithLabelValues returns the Gauge for the given slice of label values (same order as the VariableLabels in Desc). If that combination of label values is accessed for the first time, a new Gauge is created.

It is possible to call this method without using the returned Gauge to only create the new Gauge but leave it at its starting value 0. See also the SummaryVec example.

Keeping the Gauge for later use is possible (and should be considered if performance is critical), but keep in mind that Reset, DeleteLabelValues and Delete can be used to delete the Gauge from the GaugeVec. In that case, the Gauge will still exist, but it will not be exported anymore, even if a Gauge with the same label values is created later. See also the CounterVec example.

An error is returned if the number of label values is not the same as the number of VariableLabels in Desc (minus any curried labels).

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider GetMetricWith(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map).

func (*GaugeVec) MustCurryWith added in v0.9.0 

func (v *GaugeVec) MustCurryWith(labels Labels) *GaugeVec

MustCurryWith works as CurryWith but panics where CurryWith would have returned an error.

func (*GaugeVec) With 

func (v *GaugeVec) With(labels Labels) Gauge

With works as GetMetricWith, but panics where GetMetricWithLabels would have returned an error. Not returning an error allows shortcuts like 

myVec.With(prometheus.Labels{"code": "404", "method": "GET"}).Add(42)

func (*GaugeVec) WithLabelValues 

func (v *GaugeVec) WithLabelValues(lvs ...string) Gauge

WithLabelValues works as GetMetricWithLabelValues, but panics where GetMetricWithLabelValues would have returned an error. Not returning an error allows shortcuts like 

myVec.WithLabelValues("404", "GET").Add(42)

type Histogram 

type Histogram interface {
	Metric
	Collector

	// Observe adds a single observation to the histogram.
	Observe(float64)
}

A Histogram counts individual observations from an event or sample stream in configurable buckets. Similar to a summary, it also provides a sum of observations and an observation count.

On the Prometheus server, quantiles can be calculated from a Histogram using the histogram_quantile function in the query language.

Note that Histograms, in contrast to Summaries, can be aggregated with the Prometheus query language (see the documentation for detailed procedures). However, Histograms require the user to pre-define suitable buckets, and they are in general less accurate. The Observe method of a Histogram has a very low performance overhead in comparison with the Observe method of a Summary.

To create Histogram instances, use NewHistogram.

Example 
package main

import (
	"fmt"
	"math"

	"github.com/golang/protobuf/proto"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	temps := prometheus.NewHistogram(prometheus.HistogramOpts{
		Name:    "pond_temperature_celsius",
		Help:    "The temperature of the frog pond.", // Sorry, we can't measure how badly it smells.
		Buckets: prometheus.LinearBuckets(20, 5, 5),  // 5 buckets, each 5 centigrade wide.
	})

	// Simulate some observations.
	for i := 0; i < 1000; i++ {
		temps.Observe(30 + math.Floor(120*math.Sin(float64(i)*0.1))/10)
	}

	// Just for demonstration, let's check the state of the histogram by
	// (ab)using its Write method (which is usually only used by Prometheus
	// internally).
	metric := &dto.Metric{}
	temps.Write(metric)
	fmt.Println(proto.MarshalTextString(metric))

}

Output:

histogram: <
  sample_count: 1000
  sample_sum: 29969.50000000001
  bucket: <
    cumulative_count: 192
    upper_bound: 20
  >
  bucket: <
    cumulative_count: 366
    upper_bound: 25
  >
  bucket: <
    cumulative_count: 501
    upper_bound: 30
  >
  bucket: <
    cumulative_count: 638
    upper_bound: 35
  >
  bucket: <
    cumulative_count: 816
    upper_bound: 40
  >
>

func NewHistogram 

func NewHistogram(opts HistogramOpts) Histogram

NewHistogram creates a new Histogram based on the provided HistogramOpts. It panics if the buckets in HistogramOpts are not in strictly increasing order.

type HistogramOpts 

type HistogramOpts struct {
	// Namespace, Subsystem, and Name are components of the fully-qualified
	// name of the Histogram (created by joining these components with
	// "_"). Only Name is mandatory, the others merely help structuring the
	// name. Note that the fully-qualified name of the Histogram must be a
	// valid Prometheus metric name.
	Namespace string
	Subsystem string
	Name      string

	// Help provides information about this Histogram.
	//
	// Metrics with the same fully-qualified name must have the same Help
	// string.
	Help string

	// ConstLabels are used to attach fixed labels to this metric. Metrics
	// with the same fully-qualified name must have the same label names in
	// their ConstLabels.
	//
	// ConstLabels are only used rarely. In particular, do not use them to
	// attach the same labels to all your metrics. Those use cases are
	// better covered by target labels set by the scraping Prometheus
	// server, or by one specific metric (e.g. a build_info or a
	// machine_role metric). See also
	// https://prometheus.io/docs/instrumenting/writing_exporters/#target-labels,-not-static-scraped-labels
	ConstLabels Labels

	// Buckets defines the buckets into which observations are counted. Each
	// element in the slice is the upper inclusive bound of a bucket. The
	// values must be sorted in strictly increasing order. There is no need
	// to add a highest bucket with +Inf bound, it will be added
	// implicitly. The default value is DefBuckets.
	Buckets []float64
}

HistogramOpts bundles the options for creating a Histogram metric. It is mandatory to set Name to a non-empty string. All other fields are optional and can safely be left at their zero value, although it is strongly encouraged to set a Help string.

type HistogramVec 

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

HistogramVec is a Collector that bundles a set of Histograms that all share the same Desc, but have different values for their variable labels. This is used if you want to count the same thing partitioned by various dimensions (e.g. HTTP request latencies, partitioned by status code and method). Create instances with NewHistogramVec.

func NewHistogramVec 

func NewHistogramVec(opts HistogramOpts, labelNames []string) *HistogramVec

NewHistogramVec creates a new HistogramVec based on the provided HistogramOpts and partitioned by the given label names.

func (*HistogramVec) CurryWith added in v0.9.0 

func (v *HistogramVec) CurryWith(labels Labels) (ObserverVec, error)

CurryWith returns a vector curried with the provided labels, i.e. the returned vector has those labels pre-set for all labeled operations performed on it. The cardinality of the curried vector is reduced accordingly. The order of the remaining labels stays the same (just with the curried labels taken out of the sequence – which is relevant for the (GetMetric)WithLabelValues methods). It is possible to curry a curried vector, but only with labels not yet used for currying before.

The metrics contained in the HistogramVec are shared between the curried and uncurried vectors. They are just accessed differently. Curried and uncurried vectors behave identically in terms of collection. Only one must be registered with a given registry (usually the uncurried version). The Reset method deletes all metrics, even if called on a curried vector.

func (HistogramVec) Delete added in v0.9.0 

func (m HistogramVec) Delete(labels Labels) bool

Delete deletes the metric where the variable labels are the same as those passed in as labels. It returns true if a metric was deleted.

It is not an error if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc. However, such inconsistent Labels can never match an actual metric, so the method will always return false in that case.

This method is used for the same purpose as DeleteLabelValues(...string). See there for pros and cons of the two methods.

func (HistogramVec) DeleteLabelValues added in v0.9.0 

func (m HistogramVec) DeleteLabelValues(lvs ...string) bool

DeleteLabelValues removes the metric where the variable labels are the same as those passed in as labels (same order as the VariableLabels in Desc). It returns true if a metric was deleted.

It is not an error if the number of label values is not the same as the number of VariableLabels in Desc. However, such inconsistent label count can never match an actual metric, so the method will always return false in that case.

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider Delete(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the CounterVec example.

func (*HistogramVec) GetMetricWith 

func (v *HistogramVec) GetMetricWith(labels Labels) (Observer, error)

GetMetricWith returns the Histogram for the given Labels map (the label names must match those of the VariableLabels in Desc). If that label map is accessed for the first time, a new Histogram is created. Implications of creating a Histogram without using it and keeping the Histogram for later use are the same as for GetMetricWithLabelValues.

An error is returned if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc (minus any curried labels).

This method is used for the same purpose as GetMetricWithLabelValues(...string). See there for pros and cons of the two methods.

func (*HistogramVec) GetMetricWithLabelValues 

func (v *HistogramVec) GetMetricWithLabelValues(lvs ...string) (Observer, error)

GetMetricWithLabelValues returns the Histogram for the given slice of label values (same order as the VariableLabels in Desc). If that combination of label values is accessed for the first time, a new Histogram is created.

It is possible to call this method without using the returned Histogram to only create the new Histogram but leave it at its starting value, a Histogram without any observations.

Keeping the Histogram for later use is possible (and should be considered if performance is critical), but keep in mind that Reset, DeleteLabelValues and Delete can be used to delete the Histogram from the HistogramVec. In that case, the Histogram will still exist, but it will not be exported anymore, even if a Histogram with the same label values is created later. See also the CounterVec example.

An error is returned if the number of label values is not the same as the number of VariableLabels in Desc (minus any curried labels).

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider GetMetricWith(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the GaugeVec example.

func (*HistogramVec) MustCurryWith added in v0.9.0 

func (v *HistogramVec) MustCurryWith(labels Labels) ObserverVec

MustCurryWith works as CurryWith but panics where CurryWith would have returned an error.

func (*HistogramVec) With 

func (v *HistogramVec) With(labels Labels) Observer

With works as GetMetricWith but panics where GetMetricWithLabels would have returned an error. Not returning an error allows shortcuts like 

myVec.With(prometheus.Labels{"code": "404", "method": "GET"}).Observe(42.21)

func (*HistogramVec) WithLabelValues 

func (v *HistogramVec) WithLabelValues(lvs ...string) Observer

WithLabelValues works as GetMetricWithLabelValues, but panics where GetMetricWithLabelValues would have returned an error. Not returning an error allows shortcuts like 

myVec.WithLabelValues("404", "GET").Observe(42.21)

type Labels 

type Labels map[string]string

Labels represents a collection of label name -> value mappings. This type is commonly used with the With(Labels) and GetMetricWith(Labels) methods of metric vector Collectors, e.g.: 

myVec.With(Labels{"code": "404", "method": "GET"}).Add(42)

The other use-case is the specification of constant label pairs in Opts or to create a Desc.

type Metric 

type Metric interface {
	// Desc returns the descriptor for the Metric. This method idempotently
	// returns the same descriptor throughout the lifetime of the
	// Metric. The returned descriptor is immutable by contract. A Metric
	// unable to describe itself must return an invalid descriptor (created
	// with NewInvalidDesc).
	Desc() *Desc
	// Write encodes the Metric into a "Metric" Protocol Buffer data
	// transmission object.
	//
	// Metric implementations must observe concurrency safety as reads of
	// this metric may occur at any time, and any blocking occurs at the
	// expense of total performance of rendering all registered
	// metrics. Ideally, Metric implementations should support concurrent
	// readers.
	//
	// While populating dto.Metric, it is the responsibility of the
	// implementation to ensure validity of the Metric protobuf (like valid
	// UTF-8 strings or syntactically valid metric and label names). It is
	// recommended to sort labels lexicographically. Callers of Write should
	// still make sure of sorting if they depend on it.
	Write(*dto.Metric) error
}

A Metric models a single sample value with its meta data being exported to Prometheus. Implementations of Metric in this package are Gauge, Counter, Histogram, Summary, and Untyped.

func MustNewConstHistogram 

func MustNewConstHistogram(
	desc *Desc,
	count uint64,
	sum float64,
	buckets map[float64]uint64,
	labelValues ...string,
) Metric

MustNewConstHistogram is a version of NewConstHistogram that panics where NewConstMetric would have returned an error.

func MustNewConstMetric 

func MustNewConstMetric(desc *Desc, valueType ValueType, value float64, labelValues ...string) Metric

MustNewConstMetric is a version of NewConstMetric that panics where NewConstMetric would have returned an error.

func MustNewConstSummary 

func MustNewConstSummary(
	desc *Desc,
	count uint64,
	sum float64,
	quantiles map[float64]float64,
	labelValues ...string,
) Metric

MustNewConstSummary is a version of NewConstSummary that panics where NewConstMetric would have returned an error.

func NewConstHistogram 

func NewConstHistogram(
	desc *Desc,
	count uint64,
	sum float64,
	buckets map[float64]uint64,
	labelValues ...string,
) (Metric, error)

NewConstHistogram returns a metric representing a Prometheus histogram with fixed values for the count, sum, and bucket counts. As those parameters cannot be changed, the returned value does not implement the Histogram interface (but only the Metric interface). Users of this package will not have much use for it in regular operations. However, when implementing custom Collectors, it is useful as a throw-away metric that is generated on the fly to send it to Prometheus in the Collect method.

buckets is a map of upper bounds to cumulative counts, excluding the +Inf bucket.

NewConstHistogram returns an error if the length of labelValues is not consistent with the variable labels in Desc or if Desc is invalid.

Example 
package main

import (
	"fmt"

	"github.com/golang/protobuf/proto"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	desc := prometheus.NewDesc(
		"http_request_duration_seconds",
		"A histogram of the HTTP request durations.",
		[]string{"code", "method"},
		prometheus.Labels{"owner": "example"},
	)

	// Create a constant histogram from values we got from a 3rd party telemetry system.
	h := prometheus.MustNewConstHistogram(
		desc,
		4711, 403.34,
		map[float64]uint64{25: 121, 50: 2403, 100: 3221, 200: 4233},
		"200", "get",
	)

	// Just for demonstration, let's check the state of the histogram by
	// (ab)using its Write method (which is usually only used by Prometheus
	// internally).
	metric := &dto.Metric{}
	h.Write(metric)
	fmt.Println(proto.MarshalTextString(metric))

}

Output:

label: <
  name: "code"
  value: "200"
>
label: <
  name: "method"
  value: "get"
>
label: <
  name: "owner"
  value: "example"
>
histogram: <
  sample_count: 4711
  sample_sum: 403.34
  bucket: <
    cumulative_count: 121
    upper_bound: 25
  >
  bucket: <
    cumulative_count: 2403
    upper_bound: 50
  >
  bucket: <
    cumulative_count: 3221
    upper_bound: 100
  >
  bucket: <
    cumulative_count: 4233
    upper_bound: 200
  >
>

func NewConstMetric 

func NewConstMetric(desc *Desc, valueType ValueType, value float64, labelValues ...string) (Metric, error)

NewConstMetric returns a metric with one fixed value that cannot be changed. Users of this package will not have much use for it in regular operations. However, when implementing custom Collectors, it is useful as a throw-away metric that is generated on the fly to send it to Prometheus in the Collect method. NewConstMetric returns an error if the length of labelValues is not consistent with the variable labels in Desc or if Desc is invalid.

func NewConstSummary 

func NewConstSummary(
	desc *Desc,
	count uint64,
	sum float64,
	quantiles map[float64]float64,
	labelValues ...string,
) (Metric, error)

NewConstSummary returns a metric representing a Prometheus summary with fixed values for the count, sum, and quantiles. As those parameters cannot be changed, the returned value does not implement the Summary interface (but only the Metric interface). Users of this package will not have much use for it in regular operations. However, when implementing custom Collectors, it is useful as a throw-away metric that is generated on the fly to send it to Prometheus in the Collect method.

quantiles maps ranks to quantile values. For example, a median latency of 0.23s and a 99th percentile latency of 0.56s would be expressed as: 

map[float64]float64{0.5: 0.23, 0.99: 0.56}

NewConstSummary returns an error if the length of labelValues is not consistent with the variable labels in Desc or if Desc is invalid.

Example 
package main

import (
	"fmt"

	"github.com/golang/protobuf/proto"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	desc := prometheus.NewDesc(
		"http_request_duration_seconds",
		"A summary of the HTTP request durations.",
		[]string{"code", "method"},
		prometheus.Labels{"owner": "example"},
	)

	// Create a constant summary from values we got from a 3rd party telemetry system.
	s := prometheus.MustNewConstSummary(
		desc,
		4711, 403.34,
		map[float64]float64{0.5: 42.3, 0.9: 323.3},
		"200", "get",
	)

	// Just for demonstration, let's check the state of the summary by
	// (ab)using its Write method (which is usually only used by Prometheus
	// internally).
	metric := &dto.Metric{}
	s.Write(metric)
	fmt.Println(proto.MarshalTextString(metric))

}

Output:

label: <
  name: "code"
  value: "200"
>
label: <
  name: "method"
  value: "get"
>
label: <
  name: "owner"
  value: "example"
>
summary: <
  sample_count: 4711
  sample_sum: 403.34
  quantile: <
    quantile: 0.5
    value: 42.3
  >
  quantile: <
    quantile: 0.9
    value: 323.3
  >
>

func NewInvalidMetric 

func NewInvalidMetric(desc *Desc, err error) Metric

NewInvalidMetric returns a metric whose Write method always returns the provided error. It is useful if a Collector finds itself unable to collect a metric and wishes to report an error to the registry.

func NewMetricWithTimestamp added in v0.9.0 

func NewMetricWithTimestamp(t time.Time, m Metric) Metric

NewMetricWithTimestamp returns a new Metric wrapping the provided Metric in a way that it has an explicit timestamp set to the provided Time. This is only useful in rare cases as the timestamp of a Prometheus metric should usually be set by the Prometheus server during scraping. Exceptions include mirroring metrics with given timestamps from other metric sources.

NewMetricWithTimestamp works best with MustNewConstMetric, MustNewConstHistogram, and MustNewConstSummary, see example.

Currently, the exposition formats used by Prometheus are limited to millisecond resolution. Thus, the provided time will be rounded down to the next full millisecond value.

Example 
package main

import (
	"fmt"
	"time"

	"github.com/golang/protobuf/proto"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	desc := prometheus.NewDesc(
		"temperature_kelvin",
		"Current temperature in Kelvin.",
		nil, nil,
	)

	// Create a constant gauge from values we got from an external
	// temperature reporting system. Those values are reported with a slight
	// delay, so we want to add the timestamp of the actual measurement.
	temperatureReportedByExternalSystem := 298.15
	timeReportedByExternalSystem := time.Date(2009, time.November, 10, 23, 0, 0, 12345678, time.UTC)
	s := prometheus.NewMetricWithTimestamp(
		timeReportedByExternalSystem,
		prometheus.MustNewConstMetric(
			desc, prometheus.GaugeValue, temperatureReportedByExternalSystem,
		),
	)

	// Just for demonstration, let's check the state of the gauge by
	// (ab)using its Write method (which is usually only used by Prometheus
	// internally).
	metric := &dto.Metric{}
	s.Write(metric)
	fmt.Println(proto.MarshalTextString(metric))

}

Output:

gauge: <
  value: 298.15
>
timestamp_ms: 1257894000012

type MultiError 

type MultiError []error

MultiError is a slice of errors implementing the error interface. It is used by a Gatherer to report multiple errors during MetricFamily gathering.

func (*MultiError) Append added in v0.9.0 

func (errs *MultiError) Append(err error)

Append appends the provided error if it is not nil.

func (MultiError) Error 

func (errs MultiError) Error() string

func (MultiError) MaybeUnwrap 

func (errs MultiError) MaybeUnwrap() error

MaybeUnwrap returns nil if len(errs) is 0. It returns the first and only contained error as error if len(errs is 1). In all other cases, it returns the MultiError directly. This is helpful for returning a MultiError in a way that only uses the MultiError if needed.

type Observer added in v0.9.0 

type Observer interface {
	Observe(float64)
}

Observer is the interface that wraps the Observe method, which is used by Histogram and Summary to add observations.

type ObserverFunc added in v0.9.0 

type ObserverFunc func(float64)

The ObserverFunc type is an adapter to allow the use of ordinary functions as Observers. If f is a function with the appropriate signature, ObserverFunc(f) is an Observer that calls f.

This adapter is usually used in connection with the Timer type, and there are two general use cases:

The most common one is to use a Gauge as the Observer for a Timer. See the "Gauge" Timer example.

The more advanced use case is to create a function that dynamically decides which Observer to use for observing the duration. See the "Complex" Timer example.

func (ObserverFunc) Observe added in v0.9.0 

func (f ObserverFunc) Observe(value float64)

Observe calls f(value). It implements Observer.

type ObserverVec added in v0.9.0 

type ObserverVec interface {
	GetMetricWith(Labels) (Observer, error)
	GetMetricWithLabelValues(lvs ...string) (Observer, error)
	With(Labels) Observer
	WithLabelValues(...string) Observer
	CurryWith(Labels) (ObserverVec, error)
	MustCurryWith(Labels) ObserverVec

	Collector
}

ObserverVec is an interface implemented by `HistogramVec` and `SummaryVec`.

type Opts 

type Opts struct {
	// Namespace, Subsystem, and Name are components of the fully-qualified
	// name of the Metric (created by joining these components with
	// "_"). Only Name is mandatory, the others merely help structuring the
	// name. Note that the fully-qualified name of the metric must be a
	// valid Prometheus metric name.
	Namespace string
	Subsystem string
	Name      string

	// Help provides information about this metric.
	//
	// Metrics with the same fully-qualified name must have the same Help
	// string.
	Help string

	// ConstLabels are used to attach fixed labels to this metric. Metrics
	// with the same fully-qualified name must have the same label names in
	// their ConstLabels.
	//
	// ConstLabels are only used rarely. In particular, do not use them to
	// attach the same labels to all your metrics. Those use cases are
	// better covered by target labels set by the scraping Prometheus
	// server, or by one specific metric (e.g. a build_info or a
	// machine_role metric). See also
	// https://prometheus.io/docs/instrumenting/writing_exporters/#target-labels,-not-static-scraped-labels
	ConstLabels Labels
}

Opts bundles the options for creating most Metric types. Each metric implementation XXX has its own XXXOpts type, but in most cases, it is just be an alias of this type (which might change when the requirement arises.)

It is mandatory to set Name to a non-empty string. All other fields are optional and can safely be left at their zero value, although it is strongly encouraged to set a Help string.

type ProcessCollectorOpts added in v0.9.0 

type ProcessCollectorOpts struct {
	// PidFn returns the PID of the process the collector collects metrics
	// for. It is called upon each collection. By default, the PID of the
	// current process is used, as determined on construction time by
	// calling os.Getpid().
	PidFn func() (int, error)
	// If non-empty, each of the collected metrics is prefixed by the
	// provided string and an underscore ("_").
	Namespace string
	// If true, any error encountered during collection is reported as an
	// invalid metric (see NewInvalidMetric). Otherwise, errors are ignored
	// and the collected metrics will be incomplete. (Possibly, no metrics
	// will be collected at all.) While that's usually not desired, it is
	// appropriate for the common "mix-in" of process metrics, where process
	// metrics are nice to have, but failing to collect them should not
	// disrupt the collection of the remaining metrics.
	ReportErrors bool
}

ProcessCollectorOpts defines the behavior of a process metrics collector created with NewProcessCollector.

type Registerer 

type Registerer interface {
	// Register registers a new Collector to be included in metrics
	// collection. It returns an error if the descriptors provided by the
	// Collector are invalid or if they — in combination with descriptors of
	// already registered Collectors — do not fulfill the consistency and
	// uniqueness criteria described in the documentation of metric.Desc.
	//
	// If the provided Collector is equal to a Collector already registered
	// (which includes the case of re-registering the same Collector), the
	// returned error is an instance of AlreadyRegisteredError, which
	// contains the previously registered Collector.
	//
	// A Collector whose Describe method does not yield any Desc is treated
	// as unchecked. Registration will always succeed. No check for
	// re-registering (see previous paragraph) is performed. Thus, the
	// caller is responsible for not double-registering the same unchecked
	// Collector, and for providing a Collector that will not cause
	// inconsistent metrics on collection. (This would lead to scrape
	// errors.)
	Register(Collector) error
	// MustRegister works like Register but registers any number of
	// Collectors and panics upon the first registration that causes an
	// error.
	MustRegister(...Collector)
	// Unregister unregisters the Collector that equals the Collector passed
	// in as an argument.  (Two Collectors are considered equal if their
	// Describe method yields the same set of descriptors.) The function
	// returns whether a Collector was unregistered. Note that an unchecked
	// Collector cannot be unregistered (as its Describe method does not
	// yield any descriptor).
	//
	// Note that even after unregistering, it will not be possible to
	// register a new Collector that is inconsistent with the unregistered
	// Collector, e.g. a Collector collecting metrics with the same name but
	// a different help string. The rationale here is that the same registry
	// instance must only collect consistent metrics throughout its
	// lifetime.
	Unregister(Collector) bool
}

Registerer is the interface for the part of a registry in charge of registering and unregistering. Users of custom registries should use Registerer as type for registration purposes (rather than the Registry type directly). In that way, they are free to use custom Registerer implementation (e.g. for testing purposes).

func WrapRegistererWith added in v0.9.0 

func WrapRegistererWith(labels Labels, reg Registerer) Registerer

WrapRegistererWith returns a Registerer wrapping the provided Registerer. Collectors registered with the returned Registerer will be registered with the wrapped Registerer in a modified way. The modified Collector adds the provided Labels to all Metrics it collects (as ConstLabels). The Metrics collected by the unmodified Collector must not duplicate any of those labels.

WrapRegistererWith provides a way to add fixed labels to a subset of Collectors. It should not be used to add fixed labels to all metrics exposed.

Conflicts between Collectors registered through the original Registerer with Collectors registered through the wrapping Registerer will still be detected. Any AlreadyRegisteredError returned by the Register method of either Registerer will contain the ExistingCollector in the form it was provided to the respective registry.

The Collector example demonstrates a use of WrapRegistererWith.

func WrapRegistererWithPrefix added in v0.9.0 

func WrapRegistererWithPrefix(prefix string, reg Registerer) Registerer

WrapRegistererWithPrefix returns a Registerer wrapping the provided Registerer. Collectors registered with the returned Registerer will be registered with the wrapped Registerer in a modified way. The modified Collector adds the provided prefix to the name of all Metrics it collects.

WrapRegistererWithPrefix is useful to have one place to prefix all metrics of a sub-system. To make this work, register metrics of the sub-system with the wrapping Registerer returned by WrapRegistererWithPrefix. It is rarely useful to use the same prefix for all metrics exposed. In particular, do not prefix metric names that are standardized across applications, as that would break horizontal monitoring, for example the metrics provided by the Go collector (see NewGoCollector) and the process collector (see NewProcessCollector). (In fact, those metrics are already prefixed with “go_” or “process_”, respectively.)

Conflicts between Collectors registered through the original Registerer with Collectors registered through the wrapping Registerer will still be detected. Any AlreadyRegisteredError returned by the Register method of either Registerer will contain the ExistingCollector in the form it was provided to the respective registry.

type Registry 

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

Registry registers Prometheus collectors, collects their metrics, and gathers them into MetricFamilies for exposition. It implements both Registerer and Gatherer. The zero value is not usable. Create instances with NewRegistry or NewPedanticRegistry.

func NewPedanticRegistry 

func NewPedanticRegistry() *Registry

NewPedanticRegistry returns a registry that checks during collection if each collected Metric is consistent with its reported Desc, and if the Desc has actually been registered with the registry. Unchecked Collectors (those whose Describe method does not yield any descriptors) are excluded from the check.

Usually, a Registry will be happy as long as the union of all collected Metrics is consistent and valid even if some metrics are not consistent with their own Desc or a Desc provided by their registered Collector. Well-behaved Collectors and Metrics will only provide consistent Descs. This Registry is useful to test the implementation of Collectors and Metrics.

func NewRegistry 

func NewRegistry() *Registry

NewRegistry creates a new vanilla Registry without any Collectors pre-registered.

func (*Registry) Gather 

func (r *Registry) Gather() ([]*dto.MetricFamily, error)

Gather implements Gatherer.

func (*Registry) MustRegister 

func (r *Registry) MustRegister(cs ...Collector)

MustRegister implements Registerer.

func (*Registry) Register 

func (r *Registry) Register(c Collector) error

Register implements Registerer.

func (*Registry) Unregister 

func (r *Registry) Unregister(c Collector) bool

Unregister implements Registerer.

type Summary 

type Summary interface {
	Metric
	Collector

	// Observe adds a single observation to the summary.
	Observe(float64)
}

A Summary captures individual observations from an event or sample stream and summarizes them in a manner similar to traditional summary statistics: 1. sum of observations, 2. observation count, 3. rank estimations.

A typical use-case is the observation of request latencies. By default, a Summary provides the median, the 90th and the 99th percentile of the latency as rank estimations. However, the default behavior will change in the upcoming v1.0.0 of the library. There will be no rank estimations at all by default. For a sane transition, it is recommended to set the desired rank estimations explicitly.

Note that the rank estimations cannot be aggregated in a meaningful way with the Prometheus query language (i.e. you cannot average or add them). If you need aggregatable quantiles (e.g. you want the 99th percentile latency of all queries served across all instances of a service), consider the Histogram metric type. See the Prometheus documentation for more details.

To create Summary instances, use NewSummary.

Example 
package main

import (
	"fmt"
	"math"

	"github.com/golang/protobuf/proto"

	dto "github.com/prometheus/client_model/go"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	temps := prometheus.NewSummary(prometheus.SummaryOpts{
		Name:       "pond_temperature_celsius",
		Help:       "The temperature of the frog pond.",
		Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
	})

	// Simulate some observations.
	for i := 0; i < 1000; i++ {
		temps.Observe(30 + math.Floor(120*math.Sin(float64(i)*0.1))/10)
	}

	// Just for demonstration, let's check the state of the summary by
	// (ab)using its Write method (which is usually only used by Prometheus
	// internally).
	metric := &dto.Metric{}
	temps.Write(metric)
	fmt.Println(proto.MarshalTextString(metric))

}

Output:

summary: <
  sample_count: 1000
  sample_sum: 29969.50000000001
  quantile: <
    quantile: 0.5
    value: 31.1
  >
  quantile: <
    quantile: 0.9
    value: 41.3
  >
  quantile: <
    quantile: 0.99
    value: 41.9
  >
>

func NewSummary 

func NewSummary(opts SummaryOpts) Summary

NewSummary creates a new Summary based on the provided SummaryOpts.

type SummaryOpts 

type SummaryOpts struct {
	// Namespace, Subsystem, and Name are components of the fully-qualified
	// name of the Summary (created by joining these components with
	// "_"). Only Name is mandatory, the others merely help structuring the
	// name. Note that the fully-qualified name of the Summary must be a
	// valid Prometheus metric name.
	Namespace string
	Subsystem string
	Name      string

	// Help provides information about this Summary.
	//
	// Metrics with the same fully-qualified name must have the same Help
	// string.
	Help string

	// ConstLabels are used to attach fixed labels to this metric. Metrics
	// with the same fully-qualified name must have the same label names in
	// their ConstLabels.
	//
	// Due to the way a Summary is represented in the Prometheus text format
	// and how it is handled by the Prometheus server internally, “quantile”
	// is an illegal label name. Construction of a Summary or SummaryVec
	// will panic if this label name is used in ConstLabels.
	//
	// ConstLabels are only used rarely. In particular, do not use them to
	// attach the same labels to all your metrics. Those use cases are
	// better covered by target labels set by the scraping Prometheus
	// server, or by one specific metric (e.g. a build_info or a
	// machine_role metric). See also
	// https://prometheus.io/docs/instrumenting/writing_exporters/#target-labels,-not-static-scraped-labels
	ConstLabels Labels

	// Objectives defines the quantile rank estimates with their respective
	// absolute error. If Objectives[q] = e, then the value reported for q
	// will be the φ-quantile value for some φ between q-e and q+e.  The
	// default value is an empty map, resulting in a summary without
	// quantiles.
	Objectives map[float64]float64

	// MaxAge defines the duration for which an observation stays relevant
	// for the summary. Must be positive. The default value is DefMaxAge.
	MaxAge time.Duration

	// AgeBuckets is the number of buckets used to exclude observations that
	// are older than MaxAge from the summary. A higher number has a
	// resource penalty, so only increase it if the higher resolution is
	// really required. For very high observation rates, you might want to
	// reduce the number of age buckets. With only one age bucket, you will
	// effectively see a complete reset of the summary each time MaxAge has
	// passed. The default value is DefAgeBuckets.
	AgeBuckets uint32

	// BufCap defines the default sample stream buffer size.  The default
	// value of DefBufCap should suffice for most uses. If there is a need
	// to increase the value, a multiple of 500 is recommended (because that
	// is the internal buffer size of the underlying package
	// "github.com/bmizerany/perks/quantile").
	BufCap uint32
}

SummaryOpts bundles the options for creating a Summary metric. It is mandatory to set Name to a non-empty string. While all other fields are optional and can safely be left at their zero value, it is recommended to set a help string and to explicitly set the Objectives field to the desired value as the default value will change in the upcoming v1.0.0 of the library.

type SummaryVec 

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

SummaryVec is a Collector that bundles a set of Summaries that all share the same Desc, but have different values for their variable labels. This is used if you want to count the same thing partitioned by various dimensions (e.g. HTTP request latencies, partitioned by status code and method). Create instances with NewSummaryVec.

Example 
package main

import (
	"fmt"
	"math"

	"github.com/golang/protobuf/proto"

	"github.com/prometheus/client_golang/prometheus"
)

func main() {
	temps := prometheus.NewSummaryVec(
		prometheus.SummaryOpts{
			Name:       "pond_temperature_celsius",
			Help:       "The temperature of the frog pond.",
			Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
		},
		[]string{"species"},
	)

	// Simulate some observations.
	for i := 0; i < 1000; i++ {
		temps.WithLabelValues("litoria-caerulea").Observe(30 + math.Floor(120*math.Sin(float64(i)*0.1))/10)
		temps.WithLabelValues("lithobates-catesbeianus").Observe(32 + math.Floor(100*math.Cos(float64(i)*0.11))/10)
	}

	// Create a Summary without any observations.
	temps.WithLabelValues("leiopelma-hochstetteri")

	// Just for demonstration, let's check the state of the summary vector
	// by registering it with a custom registry and then let it collect the
	// metrics.
	reg := prometheus.NewRegistry()
	reg.MustRegister(temps)

	metricFamilies, err := reg.Gather()
	if err != nil || len(metricFamilies) != 1 {
		panic("unexpected behavior of custom test registry")
	}
	fmt.Println(proto.MarshalTextString(metricFamilies[0]))

}

Output:

name: "pond_temperature_celsius"
help: "The temperature of the frog pond."
type: SUMMARY
metric: <
  label: <
    name: "species"
    value: "leiopelma-hochstetteri"
  >
  summary: <
    sample_count: 0
    sample_sum: 0
    quantile: <
      quantile: 0.5
      value: nan
    >
    quantile: <
      quantile: 0.9
      value: nan
    >
    quantile: <
      quantile: 0.99
      value: nan
    >
  >
>
metric: <
  label: <
    name: "species"
    value: "lithobates-catesbeianus"
  >
  summary: <
    sample_count: 1000
    sample_sum: 31956.100000000017
    quantile: <
      quantile: 0.5
      value: 32.4
    >
    quantile: <
      quantile: 0.9
      value: 41.4
    >
    quantile: <
      quantile: 0.99
      value: 41.9
    >
  >
>
metric: <
  label: <
    name: "species"
    value: "litoria-caerulea"
  >
  summary: <
    sample_count: 1000
    sample_sum: 29969.50000000001
    quantile: <
      quantile: 0.5
      value: 31.1
    >
    quantile: <
      quantile: 0.9
      value: 41.3
    >
    quantile: <
      quantile: 0.99
      value: 41.9
    >
  >
>

func NewSummaryVec 

func NewSummaryVec(opts SummaryOpts, labelNames []string) *SummaryVec

NewSummaryVec creates a new SummaryVec based on the provided SummaryOpts and partitioned by the given label names.

Due to the way a Summary is represented in the Prometheus text format and how it is handled by the Prometheus server internally, “quantile” is an illegal label name. NewSummaryVec will panic if this label name is used.

func (*SummaryVec) CurryWith added in v0.9.0 

func (v *SummaryVec) CurryWith(labels Labels) (ObserverVec, error)

CurryWith returns a vector curried with the provided labels, i.e. the returned vector has those labels pre-set for all labeled operations performed on it. The cardinality of the curried vector is reduced accordingly. The order of the remaining labels stays the same (just with the curried labels taken out of the sequence – which is relevant for the (GetMetric)WithLabelValues methods). It is possible to curry a curried vector, but only with labels not yet used for currying before.

The metrics contained in the SummaryVec are shared between the curried and uncurried vectors. They are just accessed differently. Curried and uncurried vectors behave identically in terms of collection. Only one must be registered with a given registry (usually the uncurried version). The Reset method deletes all metrics, even if called on a curried vector.

func (SummaryVec) Delete added in v0.9.0 

func (m SummaryVec) Delete(labels Labels) bool

Delete deletes the metric where the variable labels are the same as those passed in as labels. It returns true if a metric was deleted.

It is not an error if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc. However, such inconsistent Labels can never match an actual metric, so the method will always return false in that case.

This method is used for the same purpose as DeleteLabelValues(...string). See there for pros and cons of the two methods.

func (SummaryVec) DeleteLabelValues added in v0.9.0 

func (m SummaryVec) DeleteLabelValues(lvs ...string) bool

DeleteLabelValues removes the metric where the variable labels are the same as those passed in as labels (same order as the VariableLabels in Desc). It returns true if a metric was deleted.

It is not an error if the number of label values is not the same as the number of VariableLabels in Desc. However, such inconsistent label count can never match an actual metric, so the method will always return false in that case.

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider Delete(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the CounterVec example.

func (*SummaryVec) GetMetricWith 

func (v *SummaryVec) GetMetricWith(labels Labels) (Observer, error)

GetMetricWith returns the Summary for the given Labels map (the label names must match those of the VariableLabels in Desc). If that label map is accessed for the first time, a new Summary is created. Implications of creating a Summary without using it and keeping the Summary for later use are the same as for GetMetricWithLabelValues.

An error is returned if the number and names of the Labels are inconsistent with those of the VariableLabels in Desc (minus any curried labels).

This method is used for the same purpose as GetMetricWithLabelValues(...string). See there for pros and cons of the two methods.

func (*SummaryVec) GetMetricWithLabelValues 

func (v *SummaryVec) GetMetricWithLabelValues(lvs ...string) (Observer, error)

GetMetricWithLabelValues returns the Summary for the given slice of label values (same order as the VariableLabels in Desc). If that combination of label values is accessed for the first time, a new Summary is created.

It is possible to call this method without using the returned Summary to only create the new Summary but leave it at its starting value, a Summary without any observations.

Keeping the Summary for later use is possible (and should be considered if performance is critical), but keep in mind that Reset, DeleteLabelValues and Delete can be used to delete the Summary from the SummaryVec. In that case, the Summary will still exist, but it will not be exported anymore, even if a Summary with the same label values is created later. See also the CounterVec example.

An error is returned if the number of label values is not the same as the number of VariableLabels in Desc (minus any curried labels).

Note that for more than one label value, this method is prone to mistakes caused by an incorrect order of arguments. Consider GetMetricWith(Labels) as an alternative to avoid that type of mistake. For higher label numbers, the latter has a much more readable (albeit more verbose) syntax, but it comes with a performance overhead (for creating and processing the Labels map). See also the GaugeVec example.

func (*SummaryVec) MustCurryWith added in v0.9.0 

func (v *SummaryVec) MustCurryWith(labels Labels) ObserverVec

MustCurryWith works as CurryWith but panics where CurryWith would have returned an error.

func (*SummaryVec) With 

func (v *SummaryVec) With(labels Labels) Observer

With works as GetMetricWith, but panics where GetMetricWithLabels would have returned an error. Not returning an error allows shortcuts like 

myVec.With(prometheus.Labels{"code": "404", "method": "GET"}).Observe(42.21)

func (*SummaryVec) WithLabelValues 

func (v *SummaryVec) WithLabelValues(lvs ...string) Observer

WithLabelValues works as GetMetricWithLabelValues, but panics where GetMetricWithLabelValues would have returned an error. Not returning an error allows shortcuts like 

myVec.WithLabelValues("404", "GET").Observe(42.21)

type Timer added in v0.9.0 

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

Timer is a helper type to time functions. Use NewTimer to create new instances.

Example 
package main

import (
	"math/rand"
	"time"

	"github.com/prometheus/client_golang/prometheus"
)

var (
	requestDuration = prometheus.NewHistogram(prometheus.HistogramOpts{
		Name:    "example_request_duration_seconds",
		Help:    "Histogram for the runtime of a simple example function.",
		Buckets: prometheus.LinearBuckets(0.01, 0.01, 10),
	})
)

func main() {
	// timer times this example function. It uses a Histogram, but a Summary
	// would also work, as both implement Observer. Check out
	// https://prometheus.io/docs/practices/histograms/ for differences.
	timer := prometheus.NewTimer(requestDuration)
	defer timer.ObserveDuration()

	// Do something here that takes time.
	time.Sleep(time.Duration(rand.NormFloat64()*10000+50000) * time.Microsecond)
}

Output:
Example (Complex) 
package main

import (
	"net/http"

	"github.com/prometheus/client_golang/prometheus"
)

var (
	// apiRequestDuration tracks the duration separate for each HTTP status
	// class (1xx, 2xx, ...). This creates a fair amount of time series on
	// the Prometheus server. Usually, you would track the duration of
	// serving HTTP request without partitioning by outcome. Do something
	// like this only if needed. Also note how only status classes are
	// tracked, not every single status code. The latter would create an
	// even larger amount of time series. Request counters partitioned by
	// status code are usually OK as each counter only creates one time
	// series. Histograms are way more expensive, so partition with care and
	// only where you really need separate latency tracking. Partitioning by
	// status class is only an example. In concrete cases, other partitions
	// might make more sense.
	apiRequestDuration = prometheus.NewHistogramVec(
		prometheus.HistogramOpts{
			Name:    "api_request_duration_seconds",
			Help:    "Histogram for the request duration of the public API, partitioned by status class.",
			Buckets: prometheus.ExponentialBuckets(0.1, 1.5, 5),
		},
		[]string{"status_class"},
	)
)

func handler(w http.ResponseWriter, r *http.Request) {
	status := http.StatusOK
	// The ObserverFunc gets called by the deferred ObserveDuration and
	// decides which Histogram's Observe method is called.
	timer := prometheus.NewTimer(prometheus.ObserverFunc(func(v float64) {
		switch {
		case status >= 500: // Server error.
			apiRequestDuration.WithLabelValues("5xx").Observe(v)
		case status >= 400: // Client error.
			apiRequestDuration.WithLabelValues("4xx").Observe(v)
		case status >= 300: // Redirection.
			apiRequestDuration.WithLabelValues("3xx").Observe(v)
		case status >= 200: // Success.
			apiRequestDuration.WithLabelValues("2xx").Observe(v)
		default: // Informational.
			apiRequestDuration.WithLabelValues("1xx").Observe(v)
		}
	}))
	defer timer.ObserveDuration()

	// Handle the request. Set status accordingly.
	// ...
}

func main() {
	http.HandleFunc("/api", handler)
}

Output:
Example (Gauge) 
package main

import (
	"os"

	"github.com/prometheus/client_golang/prometheus"
)

var (
	// If a function is called rarely (i.e. not more often than scrapes
	// happen) or ideally only once (like in a batch job), it can make sense
	// to use a Gauge for timing the function call. For timing a batch job
	// and pushing the result to a Pushgateway, see also the comprehensive
	// example in the push package.
	funcDuration = prometheus.NewGauge(prometheus.GaugeOpts{
		Name: "example_function_duration_seconds",
		Help: "Duration of the last call of an example function.",
	})
)

func run() error {
	// The Set method of the Gauge is used to observe the duration.
	timer := prometheus.NewTimer(prometheus.ObserverFunc(funcDuration.Set))
	defer timer.ObserveDuration()

	// Do something. Return errors as encountered. The use of 'defer' above
	// makes sure the function is still timed properly.
	return nil
}

func main() {
	if err := run(); err != nil {
		os.Exit(1)
	}
}

Output:

func NewTimer added in v0.9.0 

func NewTimer(o Observer) *Timer

NewTimer creates a new Timer. The provided Observer is used to observe a duration in seconds. Timer is usually used to time a function call in the following way: 

func TimeMe() {
    timer := NewTimer(myHistogram)
    defer timer.ObserveDuration()
    // Do actual work.
}

func (*Timer) ObserveDuration added in v0.9.0 

func (t *Timer) ObserveDuration() time.Duration

ObserveDuration records the duration passed since the Timer was created with NewTimer. It calls the Observe method of the Observer provided during construction with the duration in seconds as an argument. The observed duration is also returned. ObserveDuration is usually called with a defer statement.

Note that this method is only guaranteed to never observe negative durations if used with Go1.9+.

type UntypedFunc 

type UntypedFunc interface {
	Metric
	Collector
}

UntypedFunc works like GaugeFunc but the collected metric is of type "Untyped". UntypedFunc is useful to mirror an external metric of unknown type.

To create UntypedFunc instances, use NewUntypedFunc.

func NewUntypedFunc 

func NewUntypedFunc(opts UntypedOpts, function func() float64) UntypedFunc

NewUntypedFunc creates a new UntypedFunc based on the provided UntypedOpts. The value reported is determined by calling the given function from within the Write method. Take into account that metric collection may happen concurrently. If that results in concurrent calls to Write, like in the case where an UntypedFunc is directly registered with Prometheus, the provided function must be concurrency-safe.

type UntypedOpts 

type UntypedOpts Opts

UntypedOpts is an alias for Opts. See there for doc comments.

type ValueType 

type ValueType int

ValueType is an enumeration of metric types that represent a simple value. 

const (
	CounterValue ValueType
	GaugeValue
	UntypedValue
)

Possible values for the ValueType enum.

Source Files

View all Source files

Directories

Path Synopsis
Package graphite provides a bridge to push Prometheus metrics to a Graphite server.
 Package graphite provides a bridge to push Prometheus metrics to a Graphite server.
Package promauto provides constructors for the usual Prometheus metrics that return them already registered with the global registry (prometheus.DefaultRegisterer).
 Package promauto provides constructors for the usual Prometheus metrics that return them already registered with the global registry (prometheus.DefaultRegisterer).
Package promhttp provides tooling around HTTP servers and clients.
 Package promhttp provides tooling around HTTP servers and clients.
Package push provides functions to push metrics to a Pushgateway.
 Package push provides functions to push metrics to a Pushgateway.
Package testutil provides helpers to test code using the prometheus package of client_golang.
 Package testutil provides helpers to test code using the prometheus package of client_golang.

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.