Table of Contents

Dependency injection

Starting with version 8, Polly provides features that make the integration of Polly with the .NET IServiceCollection Dependency Injection (DI) container more streamlined. This is a thin layer atop the resilience pipeline registry which manages resilience pipelines.

Usage

To use the DI functionality, add the Polly.Extensions package to your project:

dotnet add package Polly.Extensions

Afterwards, you can use the AddResiliencePipeline(...) extension method to set up your pipeline:

var services = new ServiceCollection();

// Define a resilience pipeline
services.AddResiliencePipeline("my-key", builder =>
{
    // Add strategies to your pipeline here, timeout for example
    builder.AddTimeout(TimeSpan.FromSeconds(10));
});

// You can also access IServiceProvider by using the alternate overload
services.AddResiliencePipeline("my-key", (builder, context) =>
{
    // Resolve any service from DI
    var loggerFactory = context.ServiceProvider.GetRequiredService<ILoggerFactory>();

    // Add strategies to your pipeline here
    builder.AddTimeout(TimeSpan.FromSeconds(10));
});

// Resolve the resilience pipeline
ServiceProvider serviceProvider = services.BuildServiceProvider();
ResiliencePipelineProvider<string> pipelineProvider = serviceProvider.GetRequiredService<ResiliencePipelineProvider<string>>();
ResiliencePipeline pipeline = pipelineProvider.GetPipeline("my-key");

// Use it
await pipeline.ExecuteAsync(
    static async cancellation => await Task.Delay(100, cancellation));

The AddResiliencePipeline extension method also registers the following services into the DI container:

  • ResiliencePipelineRegistry<string>: Allows adding and retrieving resilience pipelines.
  • ResiliencePipelineProvider<string>: Allows retrieving resilience pipelines.
  • IOptions<ResiliencePipelineRegistryOptions<string>>: Options for ResiliencePipelineRegistry<string>.
Note

The generic string is inferred since the pipeline was defined using the "my-key" value.

If you only need the registry without defining a pipeline, use the AddResiliencePipelineRegistry(...) method.

Generic resilience pipelines

You can also define generic resilience pipelines (ResiliencePipeline<T>), as demonstrated below:

var services = new ServiceCollection();

// Define a generic resilience pipeline
// First parameter is the type of key, second one is the type of the results the generic pipeline works with
services.AddResiliencePipeline<string, HttpResponseMessage>("my-pipeline", builder =>
{
    builder.AddRetry(new()
    {
        MaxRetryAttempts = 2,
        ShouldHandle = new PredicateBuilder<HttpResponseMessage>()
            .Handle<HttpRequestException>()
            .Handle<TimeoutRejectedException>()
            .HandleResult(response => response.StatusCode == System.Net.HttpStatusCode.InternalServerError)
    })
    .AddTimeout(TimeSpan.FromSeconds(2));
});

// Resolve the resilience pipeline
ServiceProvider serviceProvider = services.BuildServiceProvider();
ResiliencePipelineProvider<string> pipelineProvider = serviceProvider.GetRequiredService<ResiliencePipelineProvider<string>>();
ResiliencePipeline<HttpResponseMessage> pipeline = pipelineProvider.GetPipeline<HttpResponseMessage>("my-key");

// Use it
await pipeline.ExecuteAsync(
    async cancellation => await client.GetAsync(endpoint, cancellation),
    cancellationToken);

Keyed services

.NET 8 introduced support for keyed services. Starting from version 8.3.0, Polly supports the retrieval of ResiliencePipeline or ResiliencePipeline<T> using keyed services.

To begin, define your resilience pipeline:

// Define a resilience pipeline
services.AddResiliencePipeline<string, HttpResponseMessage>("my-pipeline", builder =>
{
    // Configure the pipeline
});

// Define a generic resilience pipeline
services.AddResiliencePipeline("my-pipeline", builder =>
{
    // Configure the pipeline
});

Following the definition above, you can resolve the resilience pipelines using keyed services as shown in the example below:

public class MyApi
{
    private readonly ResiliencePipeline _pipeline;
    private readonly ResiliencePipeline<HttpResponseMessage> _genericPipeline;

    public MyApi(
        [FromKeyedServices("my-pipeline")]
        ResiliencePipeline pipeline,
        [FromKeyedServices("my-pipeline")]
        ResiliencePipeline<HttpResponseMessage> genericPipeline)
    {
        // Although the pipelines are registered with the same key, they are distinct instances.
        // One is generic, the other is not.
        _pipeline = pipeline;
        _genericPipeline = genericPipeline;
    }
}
Note

The resilience pipelines are registered in the DI container as transient services. This enables the resolution of multiple instances of ResiliencePipeline when complex pipeline keys are used. The resilience pipeline is retrieved and registered using ResiliencePipelineProvider that is responsible for lifetime management of resilience pipelines.

Deferred addition of pipelines

If you want to use a key for a resilience pipeline that may not be available immediately you can use the AddResiliencePipelines() method to defer adding them until just prior to the ResiliencePipelineProvider<TKey> is instantiated by the DI container, allowing the IServiceProvider to be used if required.

services
    .AddResiliencePipelines<string>((ctx) =>
    {
        var config = ctx.ServiceProvider.GetRequiredService<IConfiguration>();

        var configSection = config.GetSection("ResiliencePipelines");
        if (configSection is not null)
        {
            foreach (var pipelineConfig in configSection.GetChildren())
            {
                var pipelineName = pipelineConfig.GetValue<string>("Name");
                if (!string.IsNullOrEmpty(pipelineName))
                {
                    ctx.AddResiliencePipeline(pipelineName, (builder, context) =>
                    {
                        // Load configuration and configure pipeline...
                    });
                }
            }
        }
    });
Note

The AddResiliencePipelines method does not support keyed services. To enable the resolution of a resilience pipeline using keyed services, you should use the AddResiliencePipeline extension method, which adds a single resilience pipeline and registers it into the keyed services.

Dynamic reloads

Dynamic reloading is a feature of the pipeline registry that is also surfaced when using the AddResiliencePipeline(...) extension method. Use an overload that provides access to AddResiliencePipelineContext:

services
    .Configure<RetryStrategyOptions>("my-retry-options", configurationSection) // Configure the options
    .AddResiliencePipeline("my-pipeline", (builder, context) =>
    {
        // Enable the reloads whenever the named options change
        context.EnableReloads<RetryStrategyOptions>("my-retry-options");

        // Utility method to retrieve the named options
        var retryOptions = context.GetOptions<RetryStrategyOptions>("my-retry-options");

        // Add retries using the resolved options
        builder.AddRetry(retryOptions);
    });
  • EnableReloads<T>(...) activates the dynamic reloading of my-pipeline.
  • RetryStrategyOptions are fetched using context.GetOptions(...) utility method.
  • A retry strategy is added.

During a reload:

  • The callback re-executes.
  • The previous pipeline is discarded.

If an error occurs during reloading, the old pipeline remains, and dynamic reloading stops.

Resource disposal

Like dynamic reloading, the pipeline registry's resource disposal feature lets you register callbacks. These callbacks run when the pipeline is discarded, reloaded, or the registry is disposed at application shutdown.

See the example below:

services.AddResiliencePipeline("my-pipeline", (builder, context) =>
{
    // Create disposable resource
    var limiter = new ConcurrencyLimiter(new ConcurrencyLimiterOptions { PermitLimit = 100, QueueLimit = 100 });

    // Use it
    builder.AddRateLimiter(limiter);

    // Dispose the resource created in the callback when the pipeline is discarded
    context.OnPipelineDisposed(() => limiter.Dispose());
});

This feature ensures that resources are properly disposed when a pipeline reloads, discarding the old version.

Complex pipeline keys

The AddResiliencePipeline(...) method supports complex pipeline keys. This capability allows you to define the structure of your pipeline and dynamically resolve and cache multiple instances of the pipeline with different keys.

Start by defining your complex key:

public record struct MyPipelineKey(string PipelineName, string InstanceName)
{
}

Next, register your pipeline:

services.AddResiliencePipeline(new MyPipelineKey("my-pipeline", string.Empty), builder =>
{
    // Circuit breaker is a stateful strategy. To isolate the builder across different pipelines,
    // we must use multiple instances.
    builder.AddCircuitBreaker(new CircuitBreakerStrategyOptions());
});

The "my-pipeline" pipeline is now registered. Note that the InstanceName is an empty string. While we're registering the builder action for a specific pipeline, the InstanceName parameter isn't used during the pipeline's registration. Some further modifications are required for this to function.

Introduce the PipelineNameComparer:

public sealed class PipelineNameComparer : IEqualityComparer<MyPipelineKey>
{
    public bool Equals(MyPipelineKey x, MyPipelineKey y) => x.PipelineName == y.PipelineName;

    public int GetHashCode(MyPipelineKey obj) => obj.PipelineName.GetHashCode(StringComparison.Ordinal);
}

Then, configure the registry behavior:

services
    .AddResiliencePipelineRegistry<MyPipelineKey>(options =>
    {
        options.BuilderComparer = new PipelineNameComparer();

        options.InstanceNameFormatter = key => key.InstanceName;

        options.BuilderNameFormatter = key => key.PipelineName;
    });

Let's summarize our actions:

  • We assigned the PipelineNameComparer instance to the BuilderComparer property. This action changes the default registry behavior, ensuring that only the PipelineName is used to find the associated builder.
  • We used the InstanceNameFormatter delegate to represent the MyPipelineKey as an instance name for telemetry purposes, keeping the instance name as it is.
  • Likewise, the BuilderNameFormatter delegate represents the MyPipelineKey as a builder name in telemetry.

Finally, use the ResiliencePipelineProvider<MyPipelineKey> to dynamically create and cache multiple instances of the same pipeline:

ResiliencePipelineProvider<MyPipelineKey> pipelineProvider = serviceProvider.GetRequiredService<ResiliencePipelineProvider<MyPipelineKey>>();

// The registry dynamically creates and caches instance-A using the associated builder action
ResiliencePipeline instanceA = pipelineProvider.GetPipeline(new MyPipelineKey("my-pipeline", "instance-A"));

// The registry creates and caches instance-B
ResiliencePipeline instanceB = pipelineProvider.GetPipeline(new MyPipelineKey("my-pipeline", "instance-B"));

Anti-patterns

Over the years, many developers have used Polly in various ways. Some of these recurring patterns may not be ideal. The sections below highlight anti-patterns to avoid.

Accessing the IServiceCollection instead of IServiceProvider

❌ DON'T

Capture IServiceCollection inside AddResiliencePipeline():

var services = new ServiceCollection();
services.AddResiliencePipeline("myFavoriteStrategy", builder =>
{
    builder.AddRetry(new()
    {
        OnRetry = args =>
        {
            var serviceProvider = services.BuildServiceProvider();
            var logger = serviceProvider.GetService<ILogger>();
            // ...
            return default;
        }
    });
});

Reasoning:

This approach builds a new ServiceProvider before each retry attempt unnecessarily.

✅ DO

Use another overload of AddResiliencePipeline() which allows access to IServiceProvider:

var services = new ServiceCollection();
services.AddResiliencePipeline("myFavoriteStrategy", static (builder, context) =>
{
    builder.AddRetry(new()
    {
        OnRetry = args =>
        {
            var logger = context.ServiceProvider.GetService<ILogger>();
            // ...
            return default;
        }
    });
});

Reasoning:

This approach uses the already built ServiceProvider and uses the same instance before every retry attempts.

Edit this page