# OpenTelemetry .NET Metrics <details> <summary>Table of Contents</summary> * [Best Practices](#best-practices) * [Package Version](#package-version) * [Metrics API](#metrics-api) * [Meter](#meter) * [Instruments](#instruments) * [MeterProvider Management](#meterprovider-management) * [Memory Management](#memory-management) * [Example](#example) * [Pre-Aggregation](#pre-aggregation) * [Cardinality Limits](#cardinality-limits) * [Memory Preallocation](#memory-preallocation) * [Metrics Correlation](#metrics-correlation) * [Metrics Enrichment](#metrics-enrichment) * [Common issues that lead to missing metrics](#common-issues-that-lead-to-missing-metrics) </details> ## Best Practices The following tutorials have demonstrated the best practices for using metrics with OpenTelemetry .NET: * [Getting Started - ASP.NET Core Application](./getting-started-aspnetcore/README.md) * [Getting Started - Console Application](./getting-started-console/README.md) ## Package Version :heavy_check_mark: You should always use the [System.Diagnostics.Metrics](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics) APIs from the latest stable version of [System.Diagnostics.DiagnosticSource](https://www.nuget.org/packages/System.Diagnostics.DiagnosticSource/) package, regardless of the .NET runtime version being used: * If you are using the latest stable version of [OpenTelemetry .NET SDK](../../src/OpenTelemetry/README.md), you do not have to worry about the version of `System.Diagnostics.DiagnosticSource` package because it is already taken care of for you via [package dependency](../../Directory.Packages.props). * The .NET runtime team is holding a high bar for backward compatibility on `System.Diagnostics.DiagnosticSource` even during major version bumps, so compatibility is not a concern here. * Refer to the [.NET official document](https://learn.microsoft.com/dotnet/core/diagnostics/compare-metric-apis#systemdiagnosticsmetrics) for more information about `System.Diagnostics.Metrics`. ## Metrics API ### Meter :stop_sign: You should avoid creating [`System.Diagnostics.Metrics.Meter`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.meter) too frequently. `Meter` is fairly expensive and meant to be reused throughout the application. For most applications, it can be modeled as static readonly field (e.g. [Program.cs](./getting-started-console/Program.cs)) or singleton via dependency injection (e.g. [InstrumentationSource.cs](../../examples/AspNetCore/InstrumentationSource.cs)). :heavy_check_mark: You should use dot-separated [UpperCamelCase](https://en.wikipedia.org/wiki/Camel_case) as the [`Meter.Name`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.meter.name). In many cases, using the fully qualified class name might be a good option. ```csharp static readonly Meter MyMeter = new("MyCompany.MyProduct.MyLibrary", "1.0"); ``` ### Instruments :heavy_check_mark: You should understand and pick the right instrument type. > [!NOTE] > .NET runtime has provided several instrument types based on the [OpenTelemetry Specification](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#instrument). Picking the right instrument type for your use case is crucial to ensure the correct semantics and performance. Check the [Instrument Selection](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/supplementary-guidelines.md#instrument-selection) section from the supplementary guidelines for more information. | OpenTelemetry Specification | .NET Instrument Type | | --------------------------- | -------------------- | | [Asynchronous Counter](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#asynchronous-counter) | [`ObservableCounter<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.observablecounter-1) | | [Asynchronous Gauge](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#asynchronous-gauge) | [`ObservableGauge<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.observablegauge-1) | | [Asynchronous UpDownCounter](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#asynchronous-updowncounter) | [`ObservableUpDownCounter<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.observableupdowncounter-1) | | [Counter](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#counter) | [`Counter<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.counter-1) | | [Gauge](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#gauge) | [`Gauge<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.gauge-1) | | [Histogram](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#histogram) | [`Histogram<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.histogram-1) | | [UpDownCounter](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#updowncounter) | [`UpDownCounter<T>`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.updowncounter-1) | :stop_sign: You should avoid creating instruments (e.g. `Counter<T>`) too frequently. Instruments are fairly expensive and meant to be reused throughout the application. For most applications, instruments can be modeled as static readonly fields (e.g. [Program.cs](./getting-started-console/Program.cs)) or singleton via dependency injection (e.g. [InstrumentationSource.cs](../../examples/AspNetCore/InstrumentationSource.cs)). :stop_sign: You should avoid invalid instrument names. > [!NOTE] > OpenTelemetry will not collect metrics from instruments that are using invalid names. Refer to the [OpenTelemetry Specification](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#instrument-name-syntax) for the valid syntax. :stop_sign: You should avoid changing the order of tags while reporting measurements. > [!WARNING] > The last line of code has bad performance since the tags are not following the same order: ```csharp counter.Add(2, new("name", "apple"), new("color", "red")); counter.Add(3, new("name", "lime"), new("color", "green")); counter.Add(5, new("name", "lemon"), new("color", "yellow")); counter.Add(8, new("color", "yellow"), new("name", "lemon")); // bad perf ``` :heavy_check_mark: You should use TagList properly to achieve the best performance. There are two different ways of passing tags to an instrument API: * Pass the tags directly to the instrument API: ```csharp counter.Add(100, new("Key1", "Value1"), new("Key2", "Value2")); ``` * Use [`TagList`](https://learn.microsoft.com/dotnet/api/system.diagnostics.taglist): ```csharp var tags = new TagList { { "DimName1", "DimValue1" }, { "DimName2", "DimValue2" }, { "DimName3", "DimValue3" }, { "DimName4", "DimValue4" }, }; counter.Add(100, tags); ``` Here is the rule of thumb: * When reporting measurements with 3 tags or less, pass the tags directly to the instrument API. * When reporting measurements with 4 to 8 tags (inclusive), use [`TagList`](https://learn.microsoft.com/dotnet/api/system.diagnostics.taglist?#remarks) to avoid heap allocation if avoiding GC pressure is a primary performance goal. For high performance code which consider reducing CPU utilization more important (e.g. to reduce latency, to save battery, etc.) than optimizing memory allocations, use profiler and stress test to determine which approach is better. Here are some [metrics benchmark results](../../test/Benchmarks/Metrics/MetricsBenchmarks.cs) for reference. * When reporting measurements with more than 8 tags, the two approaches share very similar CPU performance and heap allocation. `TagList` is recommended due to its better readability and maintainability. > [!NOTE] > When reporting measurements with more than 8 tags, the API allocates memory on the hot code path. You SHOULD try to keep the number of tags less than or equal to 8. If you are exceeding this, check if you can model some of the tags as Resource, as [shown here](#metrics-enrichment). ## MeterProvider Management :stop_sign: You should avoid creating `MeterProvider` instances too frequently, `MeterProvider` is fairly expensive and meant to be reused throughout the application. For most applications, one `MeterProvider` instance per process would be sufficient. ```mermaid graph LR subgraph Meter A InstrumentX end subgraph Meter B InstrumentY InstrumentZ end subgraph Meter Provider 2 MetricReader2 MetricExporter2 MetricReader3 MetricExporter3 end subgraph Meter Provider 1 MetricReader1 MetricExporter1 end InstrumentX --> | Measurements | MetricReader1 InstrumentY --> | Measurements | MetricReader1 --> MetricExporter1 InstrumentZ --> | Measurements | MetricReader2 --> MetricExporter2 InstrumentZ --> | Measurements | MetricReader3 --> MetricExporter3 ``` :heavy_check_mark: You should properly manage the lifecycle of `MeterProvider` instances if they are created by you. Here is the rule of thumb when managing the lifecycle of `MeterProvider`: * If you are building an application with [dependency injection (DI)](https://learn.microsoft.com/dotnet/core/extensions/dependency-injection) (e.g. [ASP.NET Core](https://learn.microsoft.com/aspnet/core) and [.NET Worker](https://learn.microsoft.com/dotnet/core/extensions/workers)), in most cases you should create the `MeterProvider` instance and let DI manage its lifecycle. Refer to the [Getting Started with OpenTelemetry .NET Metrics in 5 Minutes - ASP.NET Core Application](./getting-started-aspnetcore/README.md) tutorial to learn more. * If you are building an application without DI, create a `MeterProvider` instance and manage the lifecycle explicitly. Refer to the [Getting Started with OpenTelemetry .NET Metrics in 5 Minutes - Console Application](./getting-started-console/README.md) tutorial to learn more. * If you forget to dispose the `MeterProvider` instance before the application ends, metrics might get dropped due to the lack of proper flush. * If you dispose the `MeterProvider` instance too early, any subsequent measurements will not be collected. ## Memory Management In OpenTelemetry, [measurements](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#measurement) are reported via the metrics API. The SDK [aggregates](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#aggregation) metrics using certain algorithms and memory management strategies to achieve good performance and efficiency. Here are the rules which OpenTelemetry .NET follows while implementing the metrics aggregation logic: 1. [**Pre-Aggregation**](#pre-aggregation): aggregation occurs within the SDK. 2. [**Cardinality Limits**](#cardinality-limits): the aggregation logic respects [cardinality limits](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#cardinality-limits), so the SDK does not use indefinite amount of memory when there is cardinality explosion. 3. [**Memory Preallocation**](#memory-preallocation): the memory used by aggregation logic is allocated during the SDK initialization, so the SDK does not have to allocate memory on-the-fly. This is to avoid garbage collection being triggered on the hot code path. ### Example Let us take the following example: * During the time range (T0, T1]: * value = 1, name = `apple`, color = `red` * value = 2, name = `lemon`, color = `yellow` * During the time range (T1, T2]: * no fruit has been received * During the time range (T2, T3]: * value = 5, name = `apple`, color = `red` * value = 2, name = `apple`, color = `green` * value = 4, name = `lemon`, color = `yellow` * value = 2, name = `lemon`, color = `yellow` * value = 1, name = `lemon`, color = `yellow` * value = 3, name = `lemon`, color = `yellow` If we aggregate and export the metrics using [Cumulative Aggregation Temporality](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/data-model.md#temporality): * (T0, T1] * attributes: {name = `apple`, color = `red`}, count: `1` * attributes: {verb = `lemon`, color = `yellow`}, count: `2` * (T0, T2] * attributes: {name = `apple`, color = `red`}, count: `1` * attributes: {verb = `lemon`, color = `yellow`}, count: `2` * (T0, T3] * attributes: {name = `apple`, color = `red`}, count: `6` * attributes: {name = `apple`, color = `green`}, count: `2` * attributes: {verb = `lemon`, color = `yellow`}, count: `12` If we aggregate and export the metrics using [Delta Aggregation Temporality](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/data-model.md#temporality): * (T0, T1] * attributes: {name = `apple`, color = `red`}, count: `1` * attributes: {verb = `lemon`, color = `yellow`}, count: `2` * (T1, T2] * nothing since we do not have any measurement received * (T2, T3] * attributes: {name = `apple`, color = `red`}, count: `5` * attributes: {name = `apple`, color = `green`}, count: `2` * attributes: {verb = `lemon`, color = `yellow`}, count: `10` ### Pre-Aggregation Taking the [fruit example](#example), there are 6 measurements reported during `(T2, T3]`. Instead of exporting every individual measurement event, the SDK aggregates them and only exports the summarized results. This approach, as illustrated in the following diagram, is called pre-aggregation: ```mermaid graph LR subgraph SDK Instrument --> | Measurements | Pre-Aggregation[Pre-Aggregation] end subgraph Collector Aggregation end Pre-Aggregation --> | Metrics | Aggregation ``` Pre-aggregation brings several benefits: 1. Although the amount of calculation remains the same, the amount of data transmitted can be significantly reduced using pre-aggregation, thus improving the overall efficiency. 2. Pre-aggregation makes it possible to apply [cardinality limits](#cardinality-limits) during SDK initialization, combined with [memory preallocation](#memory-preallocation), they make the metrics data collection behavior more predictable (e.g. a server under denial-of-service attack would still produce a constant volume of metrics data, rather than flooding the observability system with large volume of measurement events). There are cases where users might want to export raw measurement events instead of using pre-aggregation, as illustrated in the following diagram. OpenTelemetry does not support this scenario at the moment, if you are interested, please join the discussion by replying to this [feature ask](https://github.com/open-telemetry/opentelemetry-specification/issues/617). ```mermaid graph LR subgraph SDK Instrument end subgraph Collector Aggregation end Instrument --> | Measurements | Aggregation ``` ### Cardinality Limits The number of unique combinations of attributes is called cardinality. Taking the [fruit example](#example), if we know that we can only have apple/lemon as the name, red/yellow/green as the color, then we can say the cardinality is 6. No matter how many apples and lemons we have, we can always use the following table to summarize the total number of fruits based on the name and color. | Name | Color | Count | | ----- | ------ | ----- | | apple | red | 6 | | apple | yellow | 0 | | apple | green | 2 | | lemon | red | 0 | | lemon | yellow | 12 | | lemon | green | 0 | In other words, we know how much storage and network are needed to collect and transmit these metrics, regardless of the traffic pattern. In real world applications, the cardinality can be extremely high. Imagine if we have a long running service and we collect metrics with 7 attributes and each attribute can have 30 different values. We might eventually end up having to remember the complete set of all 21,870,000,000 combinations! This cardinality explosion is a well-known challenge in the metrics space. For example, it can cause surprisingly high costs in the observability system, or even be leveraged by hackers to launch a denial-of-service attack. [Cardinality limit](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#cardinality-limits) is a throttling mechanism which allows the metrics collection system to have a predictable and reliable behavior when excessive cardinality happens, whether it was due to a malicious attack or developer making mistakes while writing code. OpenTelemetry has a default cardinality limit of `2000` per metric. This limit can be configured at the individual metric level using the [View API](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#view) and the `MetricStreamConfiguration.CardinalityLimit` setting. Refer to this [doc](../../docs/metrics/customizing-the-sdk/README.md#changing-the-cardinality-limit-for-a-metric) for more information. As of `1.10.0` once a metric has reached the cardinality limit, any new measurement that could not be independently aggregated will be automatically aggregated using the [overflow attribute](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#overflow-attribute). > [!NOTE] > In SDK versions `1.6.0` - `1.9.0` the overflow attribute was an experimental feature that could be enabled by setting the environment variable `OTEL_DOTNET_EXPERIMENTAL_METRICS_EMIT_OVERFLOW_ATTRIBUTE=true`. As of `1.10.0` when [Delta Aggregation Temporality](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/data-model.md#temporality) is used, it is possible to choose a smaller cardinality limit because the SDK will reclaim unused metric points. > [!NOTE] > In SDK versions `1.7.0` - `1.9.0`, metric point reclaim was an experimental feature that could be enabled by setting the environment variable `OTEL_DOTNET_EXPERIMENTAL_METRICS_RECLAIM_UNUSED_METRIC_POINTS=true`. ### Memory Preallocation OpenTelemetry .NET SDK aims to avoid memory allocation on the hot code path. When this is combined with [proper use of Metrics API](#metrics-api), heap allocation can be avoided on the hot code path. Refer to the [metrics benchmark results](../../test/Benchmarks/Metrics/MetricsBenchmarks.cs) to learn more. :heavy_check_mark: You should measure memory allocation on hot code path, and ideally avoid any heap allocation while using the metrics API and SDK, especially when you use metrics to measure the performance of your application (for example, you do not want to spend 2 seconds doing [garbage collection](https://learn.microsoft.com/dotnet/standard/garbage-collection/) while measuring an operation which normally takes 10 milliseconds). ## Metrics Correlation In OpenTelemetry, metrics can be correlated to [traces](../trace/README.md) via [exemplars](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#exemplar). Check the [Exemplars](./exemplars/README.md) tutorial to learn more. ## Metrics Enrichment When metrics are being collected, they normally get stored in a [time series database](https://en.wikipedia.org/wiki/Time_series_database). From storage and consumption perspective, metrics can be multi-dimensional. Taking the [fruit example](#example), there are two dimensions - "name" and "color". For basic scenarios, all the dimensions can be reported during the [Metrics API](#metrics-api) invocation, however, for less trivial scenarios, the dimensions can come from different sources: * [Measurements](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#measurement) reported via the [Metrics API](#metrics-api). * Additional tags provided at instrument creation time. For example, the [`Meter.CreateCounter<T>(name, unit, description, tags)`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.meter.createcounter) overload. * Additional tags provided at meter creation time. For example, the [`Meter(name, version, tags, scope)`](https://learn.microsoft.com/dotnet/api/system.diagnostics.metrics.meter.-ctor) overload. * [Resources](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/resource/sdk.md) configured at the `MeterProvider` level. Refer to this [doc](./customizing-the-sdk/README.md#resource) for details and examples. * Additional attributes provided by the exporter or collector. For example, [jobs and instances](https://prometheus.io/docs/concepts/jobs_instances/) in Prometheus. > [!NOTE] > Instrument level tags support is not yet implemented in OpenTelemetry .NET since the [OpenTelemetry Specification](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/api.md#instrument) does not support it. Here is the rule of thumb when modeling the dimensions: * If the dimension is static throughout the process lifetime (e.g. the name of the machine, data center): * If the dimension applies to all metrics, model it as Resource, or even better, let the collector add these dimensions if feasible (e.g. a collector running in the same data center should know the name of the data center, rather than relying on / trusting each service instance to report the data center name). * If the dimension applies to a subset of metrics (e.g. the version of a client library), model it as meter level tags. * If the dimension value is dynamic, report it via the [Metrics API](#metrics-api). > [!NOTE] > There were discussions around adding a new concept called `MeasurementProcessor`, which allows dimensions to be added to / removed from measurements dynamically. This idea did not get traction due to the complexity and performance implications, refer to this [pull request](https://github.com/open-telemetry/opentelemetry-specification/pull/1938) for more context. ## Common issues that lead to missing metrics * The `Meter` used to create the instruments is not added to the `MeterProvider`. Use `AddMeter` method to enable the processing for the required metrics.