api.md

Metrics API

Status: Stable, except where otherwise specified

Table of Contents

• Overview
• MeterProvider
  • MeterProvider operations
    • Get a Meter
• Meter
  • Meter operations
• Instrument
  • General characteristics
    • Instrument name syntax
    • Instrument unit
    • Instrument description
    • Instrument advisory parameters
      • Instrument advisory parameter: ExplicitBucketBoundaries
      • Instrument advisory parameter: Attributes
    • Synchronous and Asynchronous instruments
      • Synchronous Instrument API
      • Asynchronous Instrument API
  • Counter
    • Counter creation
    • Counter operations
      • Add
  • Asynchronous Counter
    • Asynchronous Counter creation
    • Asynchronous Counter operations
  • Histogram
    • Histogram creation
    • Histogram operations
      • Record
  • Gauge
    • Gauge creation
    • Gauge operations
      • Record
  • Asynchronous Gauge
    • Asynchronous Gauge creation
    • Asynchronous Gauge operations
  • UpDownCounter
    • UpDownCounter creation
    • UpDownCounter operations
      • Add
  • Asynchronous UpDownCounter
    • Asynchronous UpDownCounter creation
    • Asynchronous UpDownCounter operations
• Measurement
  • Multiple-instrument callbacks
• Compatibility requirements
• Concurrency requirements

Overview

The Metrics API consists of these main components:

• MeterProvider is the entry point of the API. It provides access to Meters.
• Meter is the class responsible for creating Instruments.
• Instrument is responsible for reporting Measurements.

Here is an example of the object hierarchy inside a process instrumented with the metrics API:

+-- MeterProvider(default)
    |
    +-- Meter(name='io.opentelemetry.runtime', version='1.0.0')
    |   |
    |   +-- Instrument<Asynchronous Gauge, int>(name='cpython.gc', attributes=['generation'], unit='kB')
    |   |
    |   +-- instruments...
    |
    +-- Meter(name='io.opentelemetry.contrib.mongodb.client', version='2.3.0')
        |
        +-- Instrument<Counter, int>(name='client.exception', attributes=['type'], unit='1')
        |
        +-- Instrument<Histogram, double>(name='client.duration', attributes=['server.address', 'server.port'], unit='ms')
        |
        +-- instruments...

+-- MeterProvider(custom)
    |
    +-- Meter(name='bank.payment', version='23.3.5')
        |
        +-- instruments...

MeterProvider

Meters can be accessed with a MeterProvider.

In implementations of the API, the MeterProvider is expected to be the stateful object that holds any configuration.

Normally, the MeterProvider is expected to be accessed from a central place. Thus, the API SHOULD provide a way to set/register and access a global default MeterProvider.

MeterProvider operations

The MeterProvider MUST provide the following functions:

• Get a Meter

Get a Meter

This API MUST accept the following parameters:

• name: This name uniquely identifies the instrumentation scope, such as the instrumentation library (e.g. io.opentelemetry.contrib.mongodb), package, module or class name. If an application or library has built-in OpenTelemetry instrumentation, both Instrumented library and Instrumentation library can refer to the same library. In that scenario, the name denotes a module name or component name within that library or application.

• version: Specifies the version of the instrumentation scope if the scope has a version (e.g. a library version). Example value: 1.0.0.

  Users can provide a version, but it is up to their discretion. Therefore, this API needs to be structured to accept a version, but MUST NOT obligate a user to provide one.

• [since 1.4.0] schema_url: Specifies the Schema URL that should be recorded in the emitted telemetry.

  Users can provide a schema_url, but it is up to their discretion. Therefore, this API needs to be structured to accept a schema_url, but MUST NOT obligate a user to provide one.

• [since 1.13.0] attributes: Specifies the instrumentation scope attributes to associate with emitted telemetry.

  Users can provide attributes to associate with the instrumentation scope, but it is up to their discretion. Therefore, this API MUST be structured to accept a variable number of attributes, including none.

Meters are identified by name, version, and schema_url fields. When more than one Meter of the same name, version, and schema_url is created, it is unspecified whether or under which conditions the same or different Meter instances are returned. It is a user error to create Meters with different attributes but the same identity.

The term identical applied to Meters describes instances where all identifying fields are equal. The term distinct applied to Meters describes instances where at least one identifying field has a different value.

Meter

The meter is responsible for creating Instruments.

Note: Meter SHOULD NOT be responsible for the configuration. This should be the responsibility of the MeterProvider instead.

Meter operations

The Meter MUST provide functions to create new Instruments:

• Create a new Counter
• Create a new Asynchronous Counter
• Create a new Histogram
• Create a new Asynchronous Gauge
• Create a new UpDownCounter
• Create a new Asynchronous UpDownCounter

Also see the respective sections below for more information on instrument creation.

Instrument

Instruments are used to report Measurements. Each Instrument will have the following parameters:

• The name of the Instrument
• The kind of the Instrument - whether it is a Counter or one of the other kinds, whether it is synchronous or asynchronous
• An optional unit of measure
• An optional description
• Optional advisory parameters (experimental)

Instruments are associated with the Meter during creation. Instruments are identified by the name, kind, unit, and description.

Language-level features such as the distinction between integer and floating point numbers SHOULD be considered as identifying.

The term identical applied to an Instrument describes instances where all identifying fields are equal.

General characteristics

Instrument name syntax

The instrument name syntax is defined below using the Augmented Backus-Naur Form:

instrument-name = ALPHA 0*254 ("_" / "." / "-" / "/" / ALPHA / DIGIT)

ALPHA = %x41-5A / %x61-7A; A-Z / a-z
DIGIT = %x30-39 ; 0-9

• They are not null or empty strings.
• They are case-insensitive, ASCII strings.
• The first character must be an alphabetic character.
• Subsequent characters must belong to the alphanumeric characters, '_', '.', '-', and '/'.
• They can have a maximum length of 255 characters.

Instrument unit

The unit is an optional string provided by the author of the Instrument. The API SHOULD treat it as an opaque string.

• It MUST be case-sensitive (e.g. kb and kB are different units), ASCII string.
• It can have a maximum length of 63 characters. The number 63 is chosen to allow the unit strings (including the \0 terminator on certain language runtimes) to be stored and compared as fixed size array/struct when performance is critical.

Instrument description

The description is an optional free-form text provided by the author of the instrument. The API MUST treat it as an opaque string.

• It MUST support BMP (Unicode Plane 0), which is basically only the first three bytes of UTF-8 (or utf8mb3). OpenTelemetry API authors MAY decide if they want to support more Unicode Planes.
• It MUST support at least 1023 characters. OpenTelemetry API authors MAY decide if they want to support more.

Instrument advisory parameters

Status: Mixed

advisory parameters are an optional set of recommendations provided by the author of the Instrument, aimed at assisting implementations in providing useful output with minimal configuration. They differ from other parameters in that Implementations MAY ignore advisory parameters.

OpenTelemetry SDKs MUST handle advisory parameters as described here.

advisory parameters may be general, or may be accepted only for specific instrument kinds.

Instrument advisory parameter: ExplicitBucketBoundaries

Status: Stable

Applies to Histogram instrument type.

ExplicitBucketBoundaries (double[]) is the recommended set of bucket boundaries to use if aggregating to explicit bucket Histogram metric data point.

Instrument advisory parameter: Attributes

Status: Experimental

Applies to all instrument types.

Attributes (a list of attribute keys) is the recommended set of attribute keys to be used for the resulting metrics.

Synchronous and Asynchronous instruments

Instruments are categorized on whether they are synchronous or asynchronous:

• Synchronous instruments (e.g. Counter) are meant to be invoked inline with application/business processing logic. For example, an HTTP client could use a Counter to record the number of bytes it has received. Measurements recorded by synchronous instruments can be associated with the Context.

• Asynchronous instruments (e.g. Asynchronous Gauge) give the user a way to register callback function, and the callback function will be invoked only on demand (see SDK collection for reference). For example, a piece of embedded software could use an asynchronous gauge to collect the temperature from a sensor every 15 seconds, which means the callback function will only be invoked every 15 seconds. Measurements recorded by asynchronous instruments cannot be associated with the Context.

Please note that the term synchronous and asynchronous have nothing to do with the asynchronous pattern.

Synchronous Instrument API

The API to construct synchronous instruments MUST accept the following parameters:

• A name of the Instrument.

  The name needs to be provided by a user. If possible, the API SHOULD be structured so a user is obligated to provide this parameter. If it is not possible to structurally enforce this obligation, the API MUST be documented in a way to communicate to users that this parameter is needed.

  The API SHOULD be documented in a way to communicate to users that the name parameter needs to conform to the instrument name syntax. The API SHOULD NOT validate the name; that is left to implementations of the API.

• A unit of measure.

  Users can provide a unit, but it is up to their discretion. Therefore, this API needs to be structured to accept a unit, but MUST NOT obligate a user to provide one.

  The unit parameter needs to support the instrument unit rule. Meaning, the API MUST accept a case-sensitive string that supports ASCII character encoding and can hold at least 63 characters. The API SHOULD NOT validate the unit.

• A description describing the Instrument in human-readable terms.

  Users can provide a description, but it is up to their discretion. Therefore, this API needs to be structured to accept a description, but MUST NOT obligate a user to provide one.

  The description needs to support the instrument description rule. Meaning, the API MUST accept a string that supports at least BMP (Unicode Plane 0) encoded characters and hold at least 1023 characters.

• advisory parameters associated with the instrument kind.

  Users can provide advisory parameters, but its up to their discretion. Therefore, this API needs to be structured to accept advisory parameters, but MUST NOT obligate the user to provide it.

  advisory parameters need to be structured as described in instrument advisory parameters, with parameters that are general and specific to a particular instrument kind. The API SHOULD NOT validate advisory parameters.

Asynchronous Instrument API

Asynchronous instruments have associated callback functions which are responsible for reporting Measurements. Callback functions will be called only when the Meter is being observed. The order of callback execution is not specified.

The API to construct asynchronous instruments MUST accept the following parameters:

• A name of the Instrument.

  The name needs to be provided by a user. If possible, the API SHOULD be structured so a user is obligated to provide this parameter. If it is not possible to structurally enforce this obligation, the API MUST be documented in a way to communicate to users that this parameter is needed.

  The API SHOULD be documented in a way to communicate to users that the name parameter needs to conform to the instrument name syntax. The API SHOULD NOT validate the name, that is left to implementations of the API.

• A unit of measure.

  Users can provide a unit, but it is up to their discretion. Therefore, this API needs to be structured to accept a unit, but MUST NOT obligate a user to provide one.

  The unit parameter needs to support the instrument unit rule. Meaning, the API MUST accept a case-sensitive string that supports ASCII character encoding and can hold at least 63 characters. The API SHOULD NOT validate the unit.

• A description describing the Instrument in human-readable terms.

  Users can provide a description, but it is up to their discretion. Therefore, this API needs to be structured to accept a description, but MUST NOT obligate a user to provide one.

  The description needs to support the instrument description rule. Meaning, the API MUST accept a string that supports at least BMP (Unicode Plane 0) encoded characters and hold at least 1023 characters.

• advisory parameters associated with the instrument kind.

  Users can provide advisory parameters, but its up to their discretion. Therefore, this API needs to be structured to accept advisory parameters, but MUST NOT obligate the user to provide it.

  advisory parameters need to be structured as described in instrument advisory parameters, with parameters that are general and specific to a particular instrument kind. The API SHOULD NOT validate advisory parameters.

• callback functions that report Measurements of the created instrument.

  Users can provide callback functions, but it is up to their discretion. Therefore, this API MUST be structured to accept a variable number of callback functions, including none.

The API MUST support creation of asynchronous instruments by passing zero or more callback functions to be permanently registered to the newly created instrument.

A Callback is the conceptual entity created each time a callback function is registered through an OpenTelemetry API.

The API SHOULD support registration of callback functions associated with asynchronous instruments after they are created.

Where the API supports registration of callback functions after asynchronous instrumentation creation, the user MUST be able to undo registration of the specific callback after its registration by some means.

Every currently registered Callback associated with a set of instruments MUST be evaluated exactly once during collection prior to reading data for that instrument set.

Callback functions MUST be documented as follows for the end user:

• Callback functions SHOULD be reentrant safe. The SDK expects to evaluate callbacks for each MetricReader independently.
• Callback functions SHOULD NOT take an indefinite amount of time.
• Callback functions SHOULD NOT make duplicate observations (more than one Measurement with the same attributes) across all registered callbacks.

The resulting behavior when a callback violates any of these RECOMMENDATIONS is explicitly not specified at the API level.

OpenTelemetry API authors MAY decide what is the idiomatic approach for capturing measurements from callback functions. Here are some examples:

• Return a list (or tuple, generator, enumerator, etc.) of individual Measurement values.
• Pass an Observable Result as a formal parameter of the callback, where result.Observe() captures individual Measurement values.

Callbacks registered at the time of instrument creation MUST apply to the single instruments which is under construction.

Callbacks registered after the time of instrument creation MAY be associated with multiple instruments.

Idiomatic APIs for multiple-instrument Callbacks MUST distinguish the instrument associated with each observed Measurement value.

Multiple-instrument Callbacks MUST be associated at the time of registration with a declared set of asynchronous instruments from the same Meter instance. This requirement that Instruments be declaratively associated with Callbacks allows an SDK to execute only those Callbacks that are necessary to evaluate instruments that are in use by a configured View.

The API MUST treat observations from a single Callback as logically taking place at a single instant, such that when recorded, observations from a single callback MUST be reported with identical timestamps.

The API SHOULD provide some way to pass state to the callback. OpenTelemetry API authors MAY decide what is the idiomatic approach (e.g. it could be an additional parameter to the callback function, or captured by the lambda closure, or something else).

Counter

Counter is a synchronous Instrument which supports non-negative increments.

Example uses for Counter:

• count the number of bytes received
• count the number of requests completed
• count the number of accounts created
• count the number of checkpoints run
• count the number of HTTP 5xx errors

Counter creation

There MUST NOT be any API for creating a Counter other than with a Meter. This MAY be called CreateCounter. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateUInt64Counter, CreateDoubleCounter, CreateCounter<UInt64>, CreateCounter<double>.

See the general requirements for synchronous instruments.

Here are some examples that OpenTelemetry API authors might consider:

# Python

exception_counter = meter.create_counter(name="exceptions", description="number of exceptions caught", value_type=int)

// C#

var counterExceptions = meter.CreateCounter<UInt64>("exceptions", description="number of exceptions caught");

readonly struct PowerConsumption
{
    [HighCardinality]
    string customer;
};

var counterPowerUsed = meter.CreateCounter<double, PowerConsumption>("power_consumption", unit="kWh");

Counter operations

Add

Increment the Counter by a fixed amount.

This API SHOULD NOT return a value (it MAY return a dummy value if required by certain programming languages or systems, for example null, undefined).

This API MUST accept the following parameter:

• A numeric increment value.

  The increment value needs to be provided by a user. If possible, this API SHOULD be structured so a user is obligated to provide this parameter. If it is not possible to structurally enforce this obligation, this API MUST be documented in a way to communicate to users that this parameter is needed.

  The increment value is expected to be non-negative. This API SHOULD be documented in a way to communicate to users that this value is expected to be non-negative. This API SHOULD NOT validate this value, that is left to implementations of the API.

• Attributes to associate with the increment value.

  Users can provide attributes to associate with the increment value, but it is up to their discretion. Therefore, this API MUST be structured to accept a variable number of attributes, including none.

The OpenTelemetry API authors MAY decide to allow flexible attributes to be passed in as arguments. If the attribute names and types are provided during the counter creation, the OpenTelemetry API authors MAY allow attribute values to be passed in using a more efficient way (e.g. strong typed struct allocated on the callstack, tuple). The API MUST allow callers to provide flexible attributes at invocation time rather than having to register all the possible attribute names during the instrument creation. Here are some examples that OpenTelemetry API authors might consider:

# Python

exception_counter.add(1, {"exception_type": "IOError", "handled_by_user": True})
exception_counter.add(1, exception_type="IOError", handled_by_user=True)

// C#

counterExceptions.Add(1, ("exception_type", "FileLoadException"), ("handled_by_user", true));

counterPowerUsed.Add(13.5, new PowerConsumption { customer = "Tom" });
counterPowerUsed.Add(200, new PowerConsumption { customer = "Jerry" }, ("is_green_energy", true));

Asynchronous Counter

Asynchronous Counter is an asynchronous Instrument which reports monotonically increasing value(s) when the instrument is being observed.

Example uses for Asynchronous Counter:

• CPU time, which could be reported for each thread, each process or the entire system. For example "the CPU time for process A running in user mode, measured in seconds".
• The number of page faults for each process.

Asynchronous Counter creation

There MUST NOT be any API for creating an Asynchronous Counter other than with a Meter. This MAY be called CreateObservableCounter. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateUInt64ObservableCounter, CreateDoubleObservableCounter, CreateObservableCounter<UInt64>, CreateObservableCounter<double>.

It is highly recommended that implementations use the name ObservableCounter (or any language idiomatic variation, e.g. observable_counter) unless there is a strong reason not to do so. Please note that the name has nothing to do with asynchronous pattern and observer pattern.

See the general requirements for asynchronous instruments.

Note: Unlike Counter.Add() which takes the increment/delta value, the callback function reports the absolute value of the counter. To determine the reported rate the counter is changing, the difference between successive measurements is used.

OpenTelemetry API authors MAY decide what is the idiomatic approach. Here are some examples:

• Return a list (or tuple, generator, enumerator, etc.) of Measurements.
• Use an observable result argument to allow individual Measurements to be reported.

User code is recommended not to provide more than one Measurement with the same attributes in a single callback. If it happens, OpenTelemetry SDK authors MAY decide how to handle it in the SDK. For example, during the callback invocation if two measurements value=1, attributes={pid:4, bitness:64} and value=2, attributes={pid:4, bitness:64} are reported, OpenTelemetry SDK authors MAY decide to simply let them pass through (so the downstream consumer can handle duplication), drop the entire data, pick the last one, or something else. The API MUST treat observations from a single callback as logically taking place at a single instant, such that when recorded, observations from a single callback MUST be reported with identical timestamps.

The API SHOULD provide some way to pass state to the callback. OpenTelemetry API authors MAY decide what is the idiomatic approach (e.g. it could be an additional parameter to the callback function, or captured by the lambda closure, or something else).

Here are some examples that OpenTelemetry API authors might consider:

# Python

def pf_callback():
    # Note: in the real world these would be retrieved from the operating system
    return (
        (8,        ("pid", 0),   ("bitness", 64)),
        (37741921, ("pid", 4),   ("bitness", 64)),
        (10465,    ("pid", 880), ("bitness", 32)),
    )

meter.create_observable_counter(name="PF", description="process page faults", pf_callback)

# Python

def pf_callback(result):
    # Note: in the real world these would be retrieved from the operating system
    result.Observe(8,        ("pid", 0),   ("bitness", 64))
    result.Observe(37741921, ("pid", 4),   ("bitness", 64))
    result.Observe(10465,    ("pid", 880), ("bitness", 32))

meter.create_observable_counter(name="PF", description="process page faults", pf_callback)

// C#

// A simple scenario where only one value is reported

interface IAtomicClock
{
    UInt64 GetCaesiumOscillates();
}

IAtomicClock clock = AtomicClock.Connect();

meter.CreateObservableCounter<UInt64>("caesium_oscillates", () => clock.GetCaesiumOscillates());

Asynchronous Counter operations

Asynchronous Counter uses an idiomatic interface for reporting measurements through a callback, which is registered during Asynchronous Counter creation.

For callback functions registered after an asynchronous instrument is created, the API is required to support a mechanism for unregistration. For example, the object returned from register_callback can support an unregister() method directly.

# Python
class Device:
    """A device with one counter"""

    def __init__(self, meter, x):
        self.x = x
        counter = meter.create_observable_counter(name="usage", description="count of items used")
        self.cb = counter.register_callback(self.counter_callback)

    def counter_callback(self, result):
        result.Observe(self.read_counter(), {'x', self.x})

    def read_counter(self):
        return 100  # ...

    def stop(self):
        self.cb.unregister()

Histogram

Histogram is a synchronous Instrument which can be used to report arbitrary values that are likely to be statistically meaningful. It is intended for statistics such as histograms, summaries, and percentile.

Example uses for Histogram:

• the request duration
• the size of the response payload

Histogram creation

There MUST NOT be any API for creating a Histogram other than with a Meter. This MAY be called CreateHistogram. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateUInt64Histogram, CreateDoubleHistogram, CreateHistogram<UInt64>, CreateHistogram<double>.

See the general requirements for synchronous instruments.

Here are some examples that OpenTelemetry API authors might consider:

# Python

http_server_duration = meter.create_histogram(
    name="http.server.duration",
    description="measures the duration of the inbound HTTP request",
    unit="ms",
    value_type=float)

// C#

var httpServerDuration = meter.CreateHistogram<double>(
    "http.server.duration",
    description: "measures the duration of the inbound HTTP request",
    unit: "ms"
    );

Histogram operations

Record

Updates the statistics with the specified amount.

This API SHOULD NOT return a value (it MAY return a dummy value if required by certain programming languages or systems, for example null, undefined).

This API MUST accept the following parameter:

• A numeric value to record.

  The value needs to be provided by a user. If possible, this API SHOULD be structured so a user is obligated to provide this parameter. If it is not possible to structurally enforce this obligation, this API MUST be documented in a way to communicate to users that this parameter is needed.

  The value is expected to be non-negative. This API SHOULD be documented in a way to communicate to users that this value is expected to be non-negative. This API SHOULD NOT validate this value, that is left to implementations of the API.

• Attributes to associate with the value.

  Users can provide attributes to associate with the value, but it is up to their discretion. Therefore, this API MUST be structured to accept a variable number of attributes, including none.

OpenTelemetry API authors MAY decide to allow flexible attributes to be passed in as individual arguments. OpenTelemetry API authors MAY allow attribute values to be passed in using a more efficient way (e.g. strong typed struct allocated on the callstack, tuple). Here are some examples that OpenTelemetry API authors might consider:

# Python

http_server_duration.Record(50, {"http.request.method": "POST", "url.scheme": "https"})
http_server_duration.Record(100, http_method="GET", http_scheme="http")

// C#

httpServerDuration.Record(50, ("http.request.method", "POST"), ("url.scheme", "https"));
httpServerDuration.Record(100, new HttpRequestAttributes { method = "GET", scheme = "http" });

Gauge

Status: Experimental

Gauge is a synchronous Instrument which can be used to record non-additive value(s) (e.g. the background noise level - it makes no sense to record the background noise level value from multiple rooms and sum them up) when changes occur.

Note: If the values are additive (e.g. the process heap size - it makes sense to report the heap size from multiple processes and sum them up, so we get the total heap usage), use UpDownCounter.

Note: Synchronous Gauge is normally used when the measurements are exposed via a subscription to change events ( i.e. backgroundNoiseLevel.onChange(value -> gauge.record(value))). If the measurement is exposed via an accessor, use Asynchronous Gauge to invoke the accessor in a callback function ( i.e. createObservableGauge(observable -> observable.record(backgroundNoiseLevel.getCurrentValue())).

Example uses for Gauge:

• subscribe to change events for the background noise level
• subscribe to change events for the CPU fan speed

Gauge creation

There MUST NOT be any API for creating a Gauge other than with a Meter. This MAY be called CreateGauge. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateUInt64Gauge, CreateDoubleGauge, CreateGauge<UInt64>, CreateGauge<double>.

See the general requirements for synchronous instruments.

Here are some examples that OpenTelemetry API authors might consider:

// Java

DoubleGauge backgroundNoiseLevel = meter.gaugeBuilder("facility.noise.level")
    .setDescription("Background noise level of rooms")
    .setUnit("B")
    .build();

Gauge operations

Record

Record the Gauge current value.

This API SHOULD NOT return a value (it MAY return a dummy value if required by certain programming languages or systems, for example null, undefined).

This API MUST accept the following parameter:

• A numeric value. The current absolute value.

  The value needs to be provided by a user. If possible, this API SHOULD be structured so a user is obligated to provide this parameter. If it is not possible to structurally enforce this obligation, this API MUST be documented in a way to communicate to users that this parameter is needed.

• Attributes to associate with the value.

  Users can provide attributes to associate with the value, but it is up to their discretion. Therefore, this API MUST be structured to accept a variable number of attributes, including none.

The OpenTelemetry API authors MAY decide to allow flexible attributes to be passed in as arguments. If the attribute names and types are provided during the gauge creation, the OpenTelemetry API authors MAY allow attribute values to be passed in using a more efficient way (e.g. strong typed struct allocated on the callstack, tuple). The API MUST allow callers to provide flexible attributes at invocation time rather than having to register all the possible attribute names during the instrument creation. Here are some examples that OpenTelemetry API authors might consider:

// Java
Attributes roomA = Attributes.builder().put("room.id", "Rack A");
Attributes roomB = Attributes.builder().put("room.id", "Rack B");

backgroundNoiseLevel.record(4.3, roomA);
backgroundNoiseLevel.record(2.5, roomB);

Asynchronous Gauge

Asynchronous Gauge is an asynchronous Instrument which reports non-additive value(s) (e.g. the room temperature - it makes no sense to report the temperature value from multiple rooms and sum them up) when the instrument is being observed.

Note: if the values are additive (e.g. the process heap size - it makes sense to report the heap size from multiple processes and sum them up, so we get the total heap usage), use Asynchronous Counter or Asynchronous UpDownCounter.

Example uses for Asynchronous Gauge:

• the current room temperature
• the CPU fan speed

Asynchronous Gauge creation

There MUST NOT be any API for creating an Asynchronous Gauge other than with a Meter. This MAY be called CreateObservableGauge. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateUInt64ObservableGauge, CreateDoubleObservableGauge, CreateObservableGauge<UInt64>, CreateObservableGauge<double>.

It is highly recommended that implementations use the name ObservableGauge (or any language idiomatic variation, e.g. observable_gauge) unless there is a strong reason not to do so. Please note that the name has nothing to do with asynchronous pattern and observer pattern.

See the general requirements for asynchronous instruments.

Here are some examples that OpenTelemetry API authors might consider:

# Python

def cpu_frequency_callback():
    # Note: in the real world these would be retrieved from the operating system
    return (
        (3.38, ("cpu", 0), ("core", 0)),
        (3.51, ("cpu", 0), ("core", 1)),
        (0.57, ("cpu", 1), ("core", 0)),
        (0.56, ("cpu", 1), ("core", 1)),
    )

meter.create_observable_gauge(
    name="cpu.frequency",
    description="the real-time CPU clock speed",
    callback=cpu_frequency_callback,
    unit="GHz",
    value_type=float)

# Python

def cpu_frequency_callback(result):
    # Note: in the real world these would be retrieved from the operating system
    result.Observe(3.38, ("cpu", 0), ("core", 0))
    result.Observe(3.51, ("cpu", 0), ("core", 1))
    result.Observe(0.57, ("cpu", 1), ("core", 0))
    result.Observe(0.56, ("cpu", 1), ("core", 1))

meter.create_observable_gauge(
    name="cpu.frequency",
    description="the real-time CPU clock speed",
    callback=cpu_frequency_callback,
    unit="GHz",
    value_type=float)

// C#

// A simple scenario where only one value is reported

meter.CreateObservableGauge<double>("temperature", () => sensor.GetTemperature());

Asynchronous Gauge operations

Asynchronous Gauge uses an idiomatic interface for reporting measurements through a callback, which is registered during Asynchronous Gauge creation.

For callback functions registered after an asynchronous instrument is created, the API is required to support a mechanism for unregistration. For example, the object returned from register_callback can support an unregister() method directly.

# Python
class Device:
    """A device with one gauge"""

    def __init__(self, meter, x):
        self.x = x
        gauge = meter.create_observable_gauge(name="pressure", description="force/area")
        self.cb = gauge.register_callback(self.gauge_callback)

    def gauge_callback(self, result):
        result.Observe(self.read_gauge(), {'x', self.x})

    def read_gauge(self):
        return 100  # ...

    def stop(self):
        self.cb.unregister()

UpDownCounter

UpDownCounter is a synchronous Instrument which supports increments and decrements.

Note: if the value is monotonically increasing, use Counter instead.

Example uses for UpDownCounter:

• the number of active requests
• the number of items in a queue

An UpDownCounter is intended for scenarios where the absolute values are not pre-calculated, or fetching the "current value" requires extra effort. If the pre-calculated value is already available or fetching the snapshot of the "current value" is straightforward, use Asynchronous UpDownCounter instead.

UpDownCounter supports counting the size of a collection incrementally, e.g. reporting the number of items in a concurrent bag by the "color" and "material" properties as they are added and removed.

Color	Material	Count
Red	Aluminum	1
Red	Steel	2
Blue	Aluminum	0
Blue	Steel	5
Yellow	Aluminum	0
Yellow	Steel	3

# Python

items_counter = meter.create_up_down_counter(
    name="store.inventory",
    description="the number of the items available")

def restock_item(color, material):
    inventory.add_item(color=color, material=material)
    items_counter.add(1, {"color": color, "material": material})
    return true

def sell_item(color, material):
    succeeded = inventory.take_item(color=color, material=material)
    if succeeded:
        items_counter.add(-1, {"color": color, "material": material})
    return succeeded

UpDownCounter creation

There MUST NOT be any API for creating an UpDownCounter other than with a Meter. This MAY be called CreateUpDownCounter. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateInt64UpDownCounter, CreateDoubleUpDownCounter, CreateUpDownCounter<Int64>, CreateUpDownCounter<double>.

See the general requirements for synchronous instruments.

Here are some examples that OpenTelemetry API authors might consider:

# Python

customers_in_store = meter.create_up_down_counter(
    name="grocery.customers",
    description="measures the current customers in the grocery store",
    value_type=int)

// C#

var customersInStore = meter.CreateUpDownCounter<int>(
    "grocery.customers",
    description: "measures the current customers in the grocery store",
    );

UpDownCounter operations

Add

Increment or decrement the UpDownCounter by a fixed amount.

This API SHOULD NOT return a value (it MAY return a dummy value if required by certain programming languages or systems, for example null, undefined).

This API MUST accept the following parameter:

• A numeric value to add.

  The value needs to be provided by a user. If possible, this API SHOULD be structured so a user is obligated to provide this parameter. If it is not possible to structurally enforce this obligation, this API MUST be documented in a way to communicate to users that this parameter is needed.

• Attributes to associate with the value.

  Users can provide attributes to associate with the value, but it is up to their discretion. Therefore, this API MUST be structured to accept a variable number of attributes, including none.

OpenTelemetry API authors MAY decide to allow flexible attributes to be passed in as individual arguments. OpenTelemetry API authors MAY allow attribute values to be passed in using a more efficient way (e.g. strong typed struct allocated on the callstack, tuple). Here are some examples that OpenTelemetry API authors might consider:

# Python
customers_in_store.add(1, {"account.type": "commercial"})
customers_in_store.add(-1, account_type="residential")

// C#
customersInStore.Add(1, ("account.type", "commercial"));
customersInStore.Add(-1, new Account { Type = "residential" });

Asynchronous UpDownCounter

Asynchronous UpDownCounter is an asynchronous Instrument which reports additive value(s) (e.g. the process heap size - it makes sense to report the heap size from multiple processes and sum them up, so we get the total heap usage) when the instrument is being observed.

Note: if the value is monotonically increasing, use Asynchronous Counter instead; if the value is non-additive, use Asynchronous Gauge instead.

Example uses for Asynchronous UpDownCounter:

• the process heap size
• the approximate number of items in a lock-free circular buffer

Asynchronous UpDownCounter creation

There MUST NOT be any API for creating an Asynchronous UpDownCounter other than with a Meter. This MAY be called CreateObservableUpDownCounter. If strong type is desired, OpenTelemetry API authors MAY decide the language idiomatic name(s), for example CreateUInt64ObservableUpDownCounter, CreateDoubleObservableUpDownCounter, CreateObservableUpDownCounter<UInt64>, CreateObservableUpDownCounter<double>.

It is highly recommended that implementations use the name ObservableUpDownCounter (or any language idiomatic variation, e.g. observable_up_down_counter) unless there is a strong reason not to do so. Please note that the name has nothing to do with asynchronous pattern and observer pattern.

See the general requirements for asynchronous instruments.

Note: Unlike UpDownCounter.Add() which takes the increment/delta value, the callback function reports the absolute value of the Asynchronous UpDownCounter. To determine the reported rate the Asynchronous UpDownCounter is changing, the difference between successive measurements is used.

Here are some examples that OpenTelemetry API authors might consider:

# Python

def ws_callback():
    # Note: in the real world these would be retrieved from the operating system
    return (
        (8,      ("pid", 0),   ("bitness", 64)),
        (20,     ("pid", 4),   ("bitness", 64)),
        (126032, ("pid", 880), ("bitness", 32)),
    )

meter.create_observable_up_down_counter(
    name="process.workingset",
    description="process working set",
    callback=ws_callback,
    unit="kB",
    value_type=int)

# Python

def ws_callback(result):
    # Note: in the real world these would be retrieved from the operating system
    result.Observe(8,      ("pid", 0),   ("bitness", 64))
    result.Observe(20,     ("pid", 4),   ("bitness", 64))
    result.Observe(126032, ("pid", 880), ("bitness", 32))

meter.create_observable_up_down_counter(
    name="process.workingset",
    description="process working set",
    callback=ws_callback,
    unit="kB",
    value_type=int)

// C#

// A simple scenario where only one value is reported

meter.CreateObservableUpDownCounter<UInt64>("memory.physical.free", () => WMI.Query("FreePhysicalMemory"));

Asynchronous UpDownCounter operations

Asynchronous UpDownCounter uses an idiomatic interface for reporting measurements through a callback, which is registered during Asynchronous Updowncounter creation.

For callback functions registered after an asynchronous instrument is created, the API is required to support a mechanism for unregistration. For example, the object returned from register_callback can support an unregister() method directly.

# Python
class Device:
    """A device with one up_down_counter"""

    def __init__(self, meter, x):
        self.x = x
        updowncounter = meter.create_observable_up_down_counter(name="queue_size", description="items in process")
        self.cb = updowncounter.register_callback(self.up_down_counter_callback)

    def up_down_counter_callback(self, result):
        result.Observe(self.read_up_down_counter(), {'x', self.x})

    def read_up_down_counter(self):
        return 100  # ...

    def stop(self):
        self.cb.unregister()

Measurement

A Measurement represents a data point reported via the metrics API to the SDK. Please refer to the Metrics Programming Model for the interaction between the API and SDK.

Measurements encapsulate:

• A value
• Attributes

Multiple-instrument callbacks

The Metrics API MAY support an interface allowing the use of multiple instruments from a single registered Callback. The API to register a new Callback SHOULD accept:

• A callback function
• A list (or tuple, etc.) of Instruments used in the callback function.

It is RECOMMENDED that the API authors use one of the following forms for the callback function:

• The list (or tuple, etc.) returned by the callback function contains (Instrument, Measurement) pairs.
• the Observable Result parameter receives an additional (Instrument, Measurement) pairs

This interface is typically a more performant way to report multiple measurements when they are obtained through an expensive process, such as reading /proc files or probing the garbage collection subsystem.

For example,

# Python
class Device:
    """A device with two instruments"""

    def __init__(self, meter, property):
        self.property = property
        self.usage = meter.create_observable_counter(name="usage", description="count of items used")
        self.pressure = meter.create_observable_gauge(name="pressure", description="force per unit area")

        # Note the two associated instruments are passed to the callback.
        meter.register_callback([self.usage, self.pressure], self.observe)

    def observe(self, result):
        usage, pressure = expensive_system_call()
        result.observe(self.usage, usage, {'property', self.property})
        result.observe(self.pressure, pressure, {'property', self.property})

Compatibility requirements

All the metrics components SHOULD allow new APIs to be added to existing components without introducing breaking changes.

All the metrics APIs SHOULD allow optional parameter(s) to be added to existing APIs without introducing breaking changes, if possible.

Concurrency requirements

For languages which support concurrent execution the Metrics APIs provide specific guarantees and safeties.

MeterProvider - all methods are safe to be called concurrently.

Meter - all methods are safe to be called concurrently.

Instrument - All methods of any Instrument are safe to be called concurrently.