README.md

Polly

Polly is a .NET resilience and transient-fault-handling library that allows developers to express policies such as Retry, Circuit Breaker, Timeout, Bulkhead Isolation, and Fallback in a fluent and thread-safe manner.

Polly targets .NET 4.0, .NET 4.5 and .NET Standard 1.1 (coverage: .NET Core, Mono, Xamarin.iOS, Xamarin.Android, UWP, WP8.1+).

We are now a member of the .NET Foundation!

Keep up to date with new feature announcements, tips & tricks, and other news through www.thepollyproject.org

Installing via NuGet

Install-Package Polly

You can install the Strongly Named version via:

Install-Package Polly-Signed

.NET4.0 support is provided via the packages:

Install-Package Polly.Net40Async
Install-Package Polly.Net40Async-Signed

Resilience policies

Polly offers multiple resilience policies:

Policy	Premise	Aka	How does the policy mitigate?
Retry (policy family) (quickstart ; deep)	Many faults are transient and may self-correct after a short delay.	"Maybe it's just a blip"	Allows configuring automatic retries.
Circuit-breaker (policy family) (quickstart ; deep)	When a system is seriously struggling, failing fast is better than making users/callers wait. Protecting a faulting system from overload can help it recover.	"Stop doing it if it hurts" "Give that system a break"	Breaks the circuit (blocks executions) for a period, when faults exceed some pre-configured threshold.
Timeout (quickstart ; deep)	Beyond a certain wait, a success result is unlikely.	"Don't wait forever"	Guarantees the caller won't have to wait beyond the timeout.
Bulkhead Isolation (quickstart ; deep)	When a process faults, multiple failing calls backing up can easily swamp resource (eg threads/CPU) in a host. A faulting downstream system can also cause 'backed-up' failing calls upstream. Both risk a faulting process bringing down a wider system.	"One fault shouldn't sink the whole ship"	Constrains the governed actions to a fixed-size resource pool, isolating their potential to affect others.
Cache (quickstart ; deep)	Some proportion of requests may be similar.	"You've asked that one before"	Provides a response from cache if known. Stores responses automatically in cache, when first retrieved.
Fallback (quickstart ; deep)	Things will still fail - plan what you will do when that happens.	"Degrade gracefully"	Defines an alternative value to be returned (or action to be executed) on failure.
PolicyWrap (quickstart ; deep)	Different faults require different strategies; resilience means using a combination.	"Defence in depth"	Allows any of the above policies to be combined flexibly.

In addition to the detailed pages on each policy, an introduction to the role of each policy in resilience engineering is also provided in the wiki.

Using Polly with HttpClient factory from ASPNET Core 2.1

For using Polly with HttpClient factory from ASPNET Core 2.1, see our detailed wiki page, then come back here or explore the wiki to learn more about the operation of each policy.

Usage – fault-handling policies

Fault-handling policies handle specific exceptions thrown by, or results returned by, the delegates you execute through the policy.

Step 1 : Specify the exceptions/faults you want the policy to handle

(for fault-handling policies: Retry family, CircuitBreaker family and Fallback)

// Single exception type
Policy
  .Handle<HttpRequestException>()

// Single exception type with condition
Policy
  .Handle<SqlException>(ex => ex.Number == 1205)

// Multiple exception types
Policy
  .Handle<HttpRequestException>()
  .Or<OperationCanceledException>()

// Multiple exception types with condition
Policy
  .Handle<SqlException>(ex => ex.Number == 1205)
  .Or<ArgumentException>(ex => ex.ParamName == "example")

// Inner exceptions of ordinary exceptions or AggregateException, with or without conditions
Policy
  .HandleInner<HttpRequestException>()
  .OrInner<OperationCanceledException>(ex => ex.CancellationToken != myToken)

Step 1b: (optionally) Specify return results you want to handle

From Polly v4.3.0 onwards, policies wrapping calls returning a TResult can also handle TResult return values:

// Handle return value with condition 
Policy
  .HandleResult<HttpResponseMessage>(r => r.StatusCode == HttpStatusCode.NotFound)

// Handle multiple return values 
Policy
  .HandleResult<HttpResponseMessage>(r => r.StatusCode == HttpStatusCode.InternalServerError)
  .OrResult<HttpResponseMessage>(r => r.StatusCode == HttpStatusCode.BadGateway)

// Handle primitive return values (implied use of .Equals())
Policy
  .HandleResult<HttpStatusCode>(HttpStatusCode.InternalServerError)
  .OrResult<HttpStatusCode>(HttpStatusCode.BadGateway)
 
// Handle both exceptions and return values in one policy
HttpStatusCode[] httpStatusCodesWorthRetrying = {
   HttpStatusCode.RequestTimeout, // 408
   HttpStatusCode.InternalServerError, // 500
   HttpStatusCode.BadGateway, // 502
   HttpStatusCode.ServiceUnavailable, // 503
   HttpStatusCode.GatewayTimeout // 504
}; 
HttpResponseMessage result = Policy
  .Handle<HttpRequestException>()
  .OrResult<HttpResponseMessage>(r => httpStatusCodesWorthRetrying.Contains(r.StatusCode))
  .RetryAsync(...)
  .ExecuteAsync( /* some Func<Task<HttpResponseMessage>> */ )

For more information, see Handling Return Values at foot of this readme.

Step 2 : Specify how the policy should handle those faults

Retry

// Retry once
Policy
  .Handle<SomeExceptionType>()
  .Retry()

// Retry multiple times
Policy
  .Handle<SomeExceptionType>()
  .Retry(3)

// Retry multiple times, calling an action on each retry 
// with the current exception and retry count
Policy
    .Handle<SomeExceptionType>()
    .Retry(3, (exception, retryCount) =>
    {
        // do something 
    });

// Retry multiple times, calling an action on each retry 
// with the current exception, retry count and context 
// provided to Execute()
Policy
    .Handle<SomeExceptionType>()
    .Retry(3, (exception, retryCount, context) =>
    {
        // do something 
    });

Retry forever (until succeeds)

// Retry forever
Policy
  .Handle<SomeExceptionType>()
  .RetryForever()

// Retry forever, calling an action on each retry with the 
// current exception
Policy
  .Handle<SomeExceptionType>()
  .RetryForever(exception =>
  {
        // do something       
  });

// Retry forever, calling an action on each retry with the
// current exception and context provided to Execute()
Policy
  .Handle<SomeExceptionType>()
  .RetryForever((exception, context) =>
  {
        // do something       
  });

Wait and retry

// Retry, waiting a specified duration between each retry
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(new[]
  {
    TimeSpan.FromSeconds(1),
    TimeSpan.FromSeconds(2),
    TimeSpan.FromSeconds(3)
  });

// Retry, waiting a specified duration between each retry, 
// calling an action on each retry with the current exception
// and duration
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(new[]
  {
    TimeSpan.FromSeconds(1),
    TimeSpan.FromSeconds(2),
    TimeSpan.FromSeconds(3)
  }, (exception, timeSpan) => {
    // do something    
  }); 

// Retry, waiting a specified duration between each retry, 
// calling an action on each retry with the current exception, 
// duration and context provided to Execute()
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(new[]
  {
    TimeSpan.FromSeconds(1),
    TimeSpan.FromSeconds(2),
    TimeSpan.FromSeconds(3)
  }, (exception, timeSpan, context) => {
    // do something    
  });

// Retry, waiting a specified duration between each retry, 
// calling an action on each retry with the current exception, 
// duration, retry count, and context provided to Execute()
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(new[]
  {
    TimeSpan.FromSeconds(1),
    TimeSpan.FromSeconds(2),
    TimeSpan.FromSeconds(3)
  }, (exception, timeSpan, retryCount, context) => {
    // do something    
  });

// Retry a specified number of times, using a function to 
// calculate the duration to wait between retries based on 
// the current retry attempt (allows for exponential backoff)
// In this case will wait for
//  2 ^ 1 = 2 seconds then
//  2 ^ 2 = 4 seconds then
//  2 ^ 3 = 8 seconds then
//  2 ^ 4 = 16 seconds then
//  2 ^ 5 = 32 seconds
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(5, retryAttempt => 
	TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)) 
  );

// Retry a specified number of times, using a function to 
// calculate the duration to wait between retries based on 
// the current retry attempt, calling an action on each retry 
// with the current exception, duration and context provided 
// to Execute()
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(
    5, 
    retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)), 
    (exception, timeSpan, context) => {
      // do something
    }
  );

// Retry a specified number of times, using a function to 
// calculate the duration to wait between retries based on 
// the current retry attempt, calling an action on each retry 
// with the current exception, duration, retry count, and context 
// provided to Execute()
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetry(
    5, 
    retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)), 
    (exception, timeSpan, retryCount, context) => {
      // do something
    }
  );

For WaitAndRetry policies handling Http Status Code 429 Retry-After, see wiki documentation.

Wait and retry forever (until succeeds)

// Wait and retry forever
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetryForever(retryAttempt => 
	TimeSpan.FromSeconds(Math.Pow(2, retryAttempt))
    );

// Wait and retry forever, calling an action on each retry with the 
// current exception and the time to wait
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetryForever(
    retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)),    
    (exception, timespan) =>
    {
        // do something       
    });

// Wait and retry forever, calling an action on each retry with the
// current exception, time to wait, and context provided to Execute()
Policy
  .Handle<SomeExceptionType>()
  .WaitAndRetryForever(
    retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)),    
    (exception, timespan, context) =>
    {
        // do something       
    });

For more detail see: Retry policy documentation on wiki.

Circuit Breaker

// Break the circuit after the specified number of consecutive exceptions
// and keep circuit broken for the specified duration.
Policy
    .Handle<SomeExceptionType>()
    .CircuitBreaker(2, TimeSpan.FromMinutes(1));

// Break the circuit after the specified number of consecutive exceptions
// and keep circuit broken for the specified duration,
// calling an action on change of circuit state.
Action<Exception, TimeSpan> onBreak = (exception, timespan) => { ... };
Action onReset = () => { ... };
CircuitBreakerPolicy breaker = Policy
    .Handle<SomeExceptionType>()
    .CircuitBreaker(2, TimeSpan.FromMinutes(1), onBreak, onReset);

// Break the circuit after the specified number of consecutive exceptions
// and keep circuit broken for the specified duration,
// calling an action on change of circuit state,
// passing a context provided to Execute().
Action<Exception, TimeSpan, Context> onBreak = (exception, timespan, context) => { ... };
Action<Context> onReset = context => { ... };
CircuitBreakerPolicy breaker = Policy
    .Handle<SomeExceptionType>()
    .CircuitBreaker(2, TimeSpan.FromMinutes(1), onBreak, onReset);

// Monitor the circuit state, for example for health reporting.
CircuitState state = breaker.CircuitState;

/*
CircuitState.Closed - Normal operation. Execution of actions allowed.
CircuitState.Open - The automated controller has opened the circuit. Execution of actions blocked.
CircuitState.HalfOpen - Recovering from open state, after the automated break duration has expired. Execution of actions permitted. Success of subsequent action/s controls onward transition to Open or Closed state.
CircuitState.Isolated - Circuit held manually in an open state. Execution of actions blocked.
*/

// Manually open (and hold open) a circuit breaker - for example to manually isolate a downstream service.
breaker.Isolate(); 
// Reset the breaker to closed state, to start accepting actions again.
breaker.Reset(); 

For more detail see: Circuit-Breaker documentation on wiki.

Advanced Circuit Breaker

// Break the circuit if, within any period of duration samplingDuration, 
// the proportion of actions resulting in a handled exception exceeds failureThreshold, 
// provided also that the number of actions through the circuit in the period
// is at least minimumThroughput.

Policy
    .Handle<SomeExceptionType>()
    .AdvancedCircuitBreaker(
        failureThreshold: 0.5, // Break on >=50% actions result in handled exceptions...
        samplingDuration: TimeSpan.FromSeconds(10), // ... over any 10 second period
        minimumThroughput: 8, // ... provided at least 8 actions in the 10 second period.
        durationOfBreak: TimeSpan.FromSeconds(30) // Break for 30 seconds.
                );

// Configuration overloads taking state-change delegates are
// available as described for CircuitBreaker above.

// Circuit state monitoring and manual controls are
// available as described for CircuitBreaker above.

For more detail see: Advanced Circuit-Breaker documentation on wiki.

For more information on the Circuit Breaker pattern in general see:

• Making the Netflix API More Resilient
• Circuit Breaker (Martin Fowler)
• Circuit Breaker Pattern (Microsoft)
• Original Circuit Breaking Link

Fallback

// Provide a substitute value, if an execution faults.
Policy
   .Handle<Whatever>()
   .Fallback<UserAvatar>(UserAvatar.Blank)

// Specify a func to provide a substitute value, if execution faults.
Policy
   .Handle<Whatever>()
   .Fallback<UserAvatar>(() => UserAvatar.GetRandomAvatar()) // where: public UserAvatar GetRandomAvatar() { ... }

// Specify a substitute value or func, calling an action (eg for logging) if the fallback is invoked.
Policy
   .Handle<Whatever>()
   .Fallback<UserAvatar>(UserAvatar.Blank, onFallback: (exception, context) => 
    {
        // do something
    });

For more detail see: Fallback policy documentation on wiki.

Step 3 : Execute the policy

// Execute an action
var policy = Policy
              .Handle<SomeExceptionType>()
              .Retry();

policy.Execute(() => DoSomething());

// Execute an action passing arbitrary context data
var policy = Policy
    .Handle<SomeExceptionType>()
    .Retry(3, (exception, retryCount, context) =>
    {
        var methodThatRaisedException = context["methodName"];
		Log(exception, methodThatRaisedException);
    });

policy.Execute(
	() => DoSomething(),
	new Dictionary<string, object>() {{ "methodName", "some method" }}
);

// Execute a function returning a result
var policy = Policy
              .Handle<SomeExceptionType>()
              .Retry();

var result = policy.Execute(() => DoSomething());

// Execute a function returning a result passing arbitrary context data
var policy = Policy
    .Handle<SomeExceptionType>()
    .Retry(3, (exception, retryCount, context) =>
    {
        object methodThatRaisedException = context["methodName"];
        Log(exception, methodThatRaisedException)
    });

var result = policy.Execute(
    () => DoSomething(),
    new Dictionary<string, object>() {{ "methodName", "some method" }}
);

// You can of course chain it all together
Policy
  .Handle<SqlException>(ex => ex.Number == 1205)
  .Or<ArgumentException>(ex => ex.ParamName == "example")
  .Retry()
  .Execute(() => DoSomething());

The above examples show policy definition immediately followed by policy execution, for simplicity. Policy definition and execution may just as often be separated in the codebase and application lifecycle. You may choose for example to define policies on start-up, then provide them to point-of-use by dependency injection (perhaps using PolicyRegistry).

Usage – general resilience policies

The general resilience policies add resilience strategies that are not explicitly centred around handling faults which delegates may throw or return.

Step 1 : Configure

Timeout

Optimistic timeout

Optimistic timeout operates via CancellationToken and assumes delegates you execute support co-operative cancellation. You must use Execute/Async(...) overloads taking a CancellationToken, and the executed delegate must honor that CancellationToken.

// Timeout and return to the caller after 30 seconds, if the executed delegate has not completed.  Optimistic timeout: Delegates should take and honour a CancellationToken.
Policy
  .Timeout(30)

// Configure timeout as timespan.
Policy
  .Timeout(TimeSpan.FromMilliseconds(2500))

// Configure variable timeout via a func provider.
Policy
  .Timeout(() => myTimeoutProvider)) // Func<TimeSpan> myTimeoutProvider

// Timeout, calling an action if the action times out
Policy
  .Timeout(30, onTimeout: (context, timespan, task) => 
    {
        // do something 
    });

// Eg timeout, logging that the execution timed out:
Policy
  .Timeout(30, onTimeout: (context, timespan, task) => 
    {
        logger.Warn($"{context.PolicyKey} at {context.ExecutionKey}: execution timed out after {timespan.TotalSeconds} seconds.");
    });

// Eg timeout, capturing any exception from the timed-out task when it completes:
Policy
  .Timeout(30, onTimeout: (context, timespan, task) => 
    {
        task.ContinueWith(t => {
            if (t.IsFaulted) logger.Error($"{context.PolicyKey} at {context.ExecutionKey}: execution timed out after {timespan.TotalSeconds} seconds, with: {t.Exception}.");
        });
    });

Example execution:

Policy timeoutPolicy = Policy.TimeoutAsync(30);
HttpResponseMessage httpResponse = await timeoutPolicy
    .ExecuteAsync(
      async ct => await httpClient.GetAsync(endpoint, ct), // Execute a delegate which responds to a CancellationToken input parameter.
      CancellationToken.None // In this case, CancellationToken.None is passed into the execution, indicating you have no independent cancellation control you wish to add to the cancellation provided by TimeoutPolicy.  Your own indepdent CancellationToken can also be passed - see wiki for examples.
      );

Pessimistic timeout

Pessimistic timeout allows calling code to 'walk away' from waiting for an executed delegate to complete, even if it does not support cancellation. In synchronous executions this is at the expense of an extra thread; see deep doco on wiki for more detail.

// Timeout after 30 seconds, if the executed delegate has not completed.  Enforces this timeout even if the executed code has no cancellation mechanism. 
Policy
  .Timeout(30, TimeoutStrategy.Pessimistic)

// (All syntax variants outlined for optimistic timeout above also exist for pessimistic timeout.)

Example execution:

Policy timeoutPolicy = Policy.TimeoutAsync(30, TimeoutStrategy.Pessimistic);
var response = await timeoutPolicy
    .ExecuteAsync(
      async () => await FooNotHonoringCancellationAsync(), // Execute a delegate which takes no CancellationToken and does not respond to cancellation.
      );

For more detail see: Timeout policy documentation on wiki.

Bulkhead

// Restrict executions through the policy to a maximum of twelve concurrent actions.
Policy
  .Bulkhead(12)

// Restrict executions through the policy to a maximum of twelve concurrent actions, 
// with up to two actions waiting for an execution slot in the bulkhead if all slots are taken.
Policy
  .Bulkhead(12, 2)

// Restrict concurrent executions, calling an action if an execution is rejected
Policy
  .Bulkhead(12, context => 
    {
        // do something 
    });

// Monitor the bulkhead available capacity, for example for health/load reporting.
var bulkhead = Policy.Bulkhead(12, 2);
// ...
int freeExecutionSlots = bulkhead.BulkheadAvailableCount;
int freeQueueSlots     = bulkhead.QueueAvailableCount;

For more detail see: Bulkhead policy documentation on wiki.

Cache

// Define a cache Policy in the .NET Framework, using the Polly.Caching.MemoryCache nuget package.
var memoryCacheProvider = new Polly.Caching.MemoryCache.MemoryCacheProvider(MemoryCache.Default);
var cachePolicy = Policy.Cache(memoryCacheProvider, TimeSpan.FromMinutes(5));

// Define a cache policy with absolute expiration at midnight tonight.
var cachePolicy = Policy.Cache(memoryCacheProvider, new AbsoluteTtl(DateTimeOffset.Now.Date.AddDays(1));

// Define a cache policy with sliding expiration: items remain valid for another 5 minutes each time the cache item is used.
var cachePolicy = Policy.Cache(memoryCacheProvider, new SlidingTtl(TimeSpan.FromMinutes(5));

// Execute through the cache as a read-through cache: check the cache first; if not found, execute underlying delegate and store the result in the cache. 
TResult result = cachePolicy.Execute(() => getFoo(), new Context("FooKey")); // "FooKey" is the cache key used in this execution.

// Define a cache Policy, and catch any cache provider errors for logging. 
var cachePolicy = Policy.Cache(myCacheProvider, TimeSpan.FromMinutes(5), 
   (context, key, ex) => { 
       logger.Error($"Cache provider, for key {key}, threw exception: {ex}."); // (for example) 
   }
);

For richer options and details of using further cache providers see: Cache policy documentation on wiki.

PolicyWrap

// Define a combined policy strategy, built of previously-defined policies.
var policyWrap = Policy
  .Wrap(fallback, cache, retry, breaker, timeout, bulkhead);
// (wraps the policies around any executed delegate: fallback outermost ... bulkhead innermost)
policyWrap.Execute(...)

// Define a standard resilience strategy ...
PolicyWrap commonResilience = Policy.Wrap(retry, breaker, timeout);

// ... then wrap in extra policies specific to a call site, at that call site:
Avatar avatar = Policy
   .Handle<Whatever>()
   .Fallback<Avatar>(Avatar.Blank)
   .Wrap(commonResilience)
   .Execute(() => { /* get avatar */ });

// Share commonResilience, but wrap different policies in at another call site:
Reputation reps = Policy
   .Handle<Whatever>()
   .Fallback<Reputation>(Reputation.NotAvailable)
   .Wrap(commonResilience)
   .Execute(() => { /* get reputation */ });  

For more detail see: PolicyWrap documentation on wiki.

NoOp

// Define a policy which will simply cause delegates passed for execution to be executed 'as is'.
// This is useful for stubbing-out Polly in unit tests, 
// or in application situations where your code architecture might expect a policy
// but you simply want to pass the execution through without policy intervention.
NoOpPolicy noOp = Policy.NoOp();

For more detail see: NoOp documentation on wiki.

Step 2 : Execute the policy

As for fault-handling policies above.

Post-execution: capturing the result, or any final exception

Using the ExecuteAndCapture(...) methods you can capture the outcome of an execution: the methods return a PolicyResult instance which describes whether the outcome was a successful execution or a fault.

var policyResult = Policy
              .Handle<HttpRequestException>()
              .RetryAsync()
              .ExecuteAndCaptureAsync(() => DoSomethingAsync());
/*              
policyResult.Outcome - whether the call succeeded or failed         
policyResult.FinalException - the final exception captured, will be null if the call succeeded
policyResult.ExceptionType - was the final exception an exception the policy was defined to handle (like HttpRequestException above) or an unhandled one (say Exception). Will be null if the call succeeded.
policyResult.Result - if executing a func, the result if the call succeeded or the type's default value
*/

Handing return values, and Policy<TResult>

As described at step 1b, from Polly v4.3.0 onwards, policies can handle return values and exceptions in combination:

// Handle both exceptions and return values in one policy
HttpStatusCode[] httpStatusCodesWorthRetrying = {
   HttpStatusCode.RequestTimeout, // 408
   HttpStatusCode.InternalServerError, // 500
   HttpStatusCode.BadGateway, // 502
   HttpStatusCode.ServiceUnavailable, // 503
   HttpStatusCode.GatewayTimeout // 504
}; 
HttpResponseMessage result = Policy
  .Handle<HttpRequestException>()
  .OrResult<HttpResponseMessage>(r => httpStatusCodesWorthRetrying.Contains(r.StatusCode))
  .RetryAsync(...)
  .ExecuteAsync( /* some Func<Task<HttpResponseMessage>> */ )

The exceptions and return results to handle can be expressed fluently in any order.

Strongly-typed Policy<TResult>

Configuring a policy with .HandleResult<TResult>(...) or .OrResult<TResult>(...) generates a strongly-typed Policy<TResult> of the specific policy type, eg Retry<TResult>, AdvancedCircuitBreaker<TResult>.

These policies must be used to execute delegates returning TResult, ie:

• Execute(Func<TResult>) (and related overloads)
• ExecuteAsync(Func<CancellationToken, Task<TResult>>) (and related overloads)

ExecuteAndCapture<TResult>()

.ExecuteAndCapture(...) on non-generic policies returns a PolicyResult with properties:

policyResult.Outcome - whether the call succeeded or failed         
policyResult.FinalException - the final exception captured; will be null if the call succeeded
policyResult.ExceptionType - was the final exception an exception the policy was defined to handle (like HttpRequestException above) or an unhandled one (say Exception)? Will be null if the call succeeded.
policyResult.Result - if executing a func, the result if the call succeeded; otherwise, the type's default value

.ExecuteAndCapture<TResult>(Func<TResult>) on strongly-typed policies adds two properties:

policyResult.FaultType - was the final fault handled an exception or a result handled by the policy? Will be null if the delegate execution succeeded. 
policyResult.FinalHandledResult - the final fault result handled; will be null or the type's default value, if the call succeeded

State-change delegates on Policy<TResult> policies

In non-generic policies handling only exceptions, state-change delegates such as onRetry and onBreak take an Exception parameter.

In generic-policies handling TResult return values, state-change delegates are identical except they take a DelegateResult<TResult> parameter in place of Exception. DelegateResult<TResult> has two properties:

• Exception // The exception just thrown if policy is in process of handling an exception (otherwise null)
• Result // The TResult just raised, if policy is in process of handling a result (otherwise default(TResult))

BrokenCircuitException<TResult>

Non-generic CircuitBreaker policies throw a BrokenCircuitException when the circuit is broken. This BrokenCircuitException contains the last exception (the one which caused the circuit to break) as the InnerException.

For CircuitBreakerPolicy<TResult> policies:

• A circuit broken due to an exception throws a BrokenCircuitException with InnerException set to the exception which triggered the break (as previously).
• A circuit broken due to handling a result throws a BrokenCircuitException<TResult> with the Result property set to the result which caused the circuit to break.

Policy Keys and Context data

// Identify policies with a PolicyKey, using the WithPolicyKey() extension method
// (for example, for correlation in logs or metrics)

var policy = Policy
    .Handle<DataAccessException>()
    .Retry(3, onRetry: (exception, retryCount, context) =>
       {
           logger.Error($"Retry {retryCount} of {context.PolicyKey} at {context.ExecutionKey}, due to: {exception}.");
       })
    .WithPolicyKey("MyDataAccessPolicy");

// Identify call sites with an ExecutionKey, by passing in a Context
var customerDetails = policy.Execute(myDelegate, new Context("GetCustomerDetails"));

// "MyDataAccessPolicy" -> context.PolicyKey 
// "GetCustomerDetails  -> context.ExecutionKey


// Pass additional custom information from call site into execution context 
var policy = Policy
    .Handle<DataAccessException>()
    .Retry(3, onRetry: (exception, retryCount, context) =>
       {
           logger.Error($"Retry {retryCount} of {context.PolicyKey} at {context.ExecutionKey}, getting {context["Type"]} of id {context["Id"]}, due to: {exception}.");
       })
    .WithPolicyKey("MyDataAccessPolicy");

int id = ... // customer id from somewhere
var customerDetails = policy.Execute(() => GetCustomer(id), 
    new Context("GetCustomerDetails", new Dictionary<string, object>() {{"Type","Customer"},{"Id",id}}
    ));

For more detail see: Keys and Context Data on wiki.

PolicyRegistry

// Create a policy registry (for example on application start-up) 
PolicyRegistry registry = new PolicyRegistry();

// Populate the registry with policies
registry.Add("StandardHttpResilience", myStandardHttpResiliencePolicy);
// Or:
registry["StandardHttpResilience"] = myStandardHttpResiliencePolicy;

// Pass the registry instance to usage sites by DI, perhaps
public class MyServiceGateway 
{
    public void MyServiceGateway(..., IReadOnlyPolicyRegistry<string> registry, ...)
    {
       ...
    } 
}
// (Or if you prefer ambient-context pattern, use a thread-safe singleton)

// Use a policy from the registry
registry.Get<IAsyncPolicy<HttpResponseMessage>>("StandardHttpResilience")
    .ExecuteAsync<HttpResponseMessage>(...)

PolicyRegistry has a range of further dictionary-like semantics such as .ContainsKey(...), .TryGet<TPolicy>(...), .Count, .Clear(), and Remove(...).

Available from v5.2.0. For more detail see: PolicyRegistry on wiki.

Asynchronous Support

Polly fully supports asynchronous executions, using the asynchronous methods:

• RetryAsync
• WaitAndRetryAsync
• CircuitBreakerAsync
• (etc)
• ExecuteAsync
• ExecuteAndCaptureAsync

In place of their synchronous counterparts:

• Retry
• WaitAndRetry
• CircuitBreaker
• (etc)
• Execute
• ExecuteAndCapture

Async overloads exist for all policy types and for all Execute() and ExecuteAndCapture() overloads.

Usage example:

await Policy
  .Handle<SqlException>(ex => ex.Number == 1205)
  .Or<ArgumentException>(ex => ex.ParamName == "example")
  .RetryAsync()
  .ExecuteAsync(() => DoSomethingAsync());

SynchronizationContext

Async continuations and retries by default do not run on a captured synchronization context. To change this, use .ExecuteAsync(...) overloads taking a boolean continueOnCapturedContext parameter.

Cancellation support

Async policy execution supports cancellation via .ExecuteAsync(...) overloads taking a CancellationToken.

The token you pass as the cancellationToken parameter to the ExecuteAsync(...) call serves three purposes:

• It cancels Policy actions such as further retries, waits between retries or waits for a bulkhead execution slot.
• It is passed by the policy as the CancellationToken input parameter to any delegate executed through the policy, to support cancellation during delegate execution.
• In common with the Base Class Library implementation in Task.Run(…) and elsewhere, if the cancellation token is cancelled before execution begins, the user delegate is not executed at all.

// Try several times to retrieve from a uri, but support cancellation at any time.
CancellationToken cancellationToken = // ...
var policy = Policy
    .Handle<HttpRequestException>()
    .WaitAndRetryAsync(new[] { 
        TimeSpan.FromSeconds(1), 
        TimeSpan.FromSeconds(2), 
        TimeSpan.FromSeconds(4) 
    });
var response = await policy.ExecuteAsync(ct => httpClient.GetAsync(uri, ct), cancellationToken);

From Polly v5.0, synchronous executions also support cancellation via CancellationToken.

Thread safety

All Polly policies are fully thread-safe. You can safely re-use policies at multiple call sites, and execute through policies concurrently on different threads.

While the internal operation of the policy is thread-safe, this does not magically make delegates you execute through the policy thread-safe: if delegates you execute through the policy are not thread-safe, they remain not thread-safe.

Interfaces

Polly v5.2.0 adds interfaces intended to support PolicyRegistry and to group Policy functionality by the interface segregation principle. Polly's interfaces are not intended for coding your own policy implementations against.

Execution interfaces: ISyncPolicy etc

Execution interfaces ISyncPolicy, IAsyncPolicy, ISyncPolicy<TResult> and IAsyncPolicy<TResult> define the execution overloads available to policies targeting sync/async, and non-generic / generic calls respectively.

See blog posts for why Polly has both non-generic and generic policies and separate sync and async policies.

Policy-kind interfaces: ICircuitBreakerPolicy etc

Orthogonal to the execution interfaces, interfaces specific to the kind of Policy define properties and methods common to that type of policy.

For example, ICircuitBreakerPolicy defines

• CircuitState CircuitState
• Exception LastException
• void Isolate()
• void Reset()

with ICircuitBreakerPolicy<TResult> : ICircuitBreakerPolicy adding:

• TResult LastHandledResult.

This allows collections of similar kinds of policy to be treated as one - for example, for monitoring all your circuit-breakers as described here.

For more detail see: Polly and interfaces on wiki.

.NET4.0 support

The .NET4.0 package uses Microsoft.Bcl.Async to add async support. To minimise dependencies on the main Polly nuget package, the .NET4.0 version is available as separate Nuget packages Polly.Net40Async and Polly.Net40Async-signed.

.NET3.5 support

The last version of Polly to support .NET3.5 is v4.3.0. This includes all of the 4.x features, as it was the last version before v5.0. This package may be installed via the NuGet package manager by specifying its version: Install-Package Polly -Version 4.3.0.

Release notes

For details of changes by release see the change log. We also tag relevant Pull Requests and Issues with milestones, which match to nuget package release numbers.

3rd Party Libraries and Contributions

• Fluent Assertions - A set of .NET extension methods that allow you to more naturally specify the expected outcome of a TDD or BDD-style test | Apache License 2.0 (Apache)
• xUnit.net - Free, open source, community-focused unit testing tool for the .NET Framework | Apache License 2.0 (Apache)
• Ian Griffith's TimedLock
• Steven van Deursen's ReadOnlyDictionary (until v5.0.6)
• Stephen Cleary's AsyncEx library for AsyncSemaphore (supports BulkheadAsync policy for .NET4.0 only) | MIT license
• @theraot's ExceptionDispatchInfo implementation for .NET4.0 (supports Timeout policy for .NET4.0 only) including also a contribution by @migueldeicaza | Licensed under and distributed under Creative Commons Attribution Share Alike license per StackExchange Terms of Service
• Build powered by Cake and GitVersionTask.

Acknowledgements

• lokad-shared-libraries - Helper assemblies originally for .NET 3.5 and Silverlight 2.0 which were developed as part of the Open Source effort by Lokad.com (discontinued) | New BSD License
• @michael-wolfenden - The creator and mastermind of Polly!
• @ghuntley - Portable Class Library implementation.
• @mauricedb - Initial async implementation.
• @robgibbens - Added existing async files to PCL project
• Hacko - Added extra NotOnCapturedContext call to prevent potential deadlocks when blocking on asynchronous calls
• @ThomasMentzel - Added ability to capture the results of executing a policy via ExecuteAndCapture
• @yevhen - Added full control of whether to continue on captured synchronization context or not
• @reisenberger - Added full async cancellation support
• @reisenberger - Added async support for ContextualPolicy
• @reisenberger - Added ContextualPolicy support for circuit-breaker
• @reisenberger - Extended circuit-breaker for public monitoring and control
• @reisenberger - Added ExecuteAndCapture support with arbitrary context data
• @kristianhald and @reisenberger - Added AdvancedCircuitBreaker
• @reisenberger - Allowed async onRetry delegates to async retry policies
• @Lumirris - Add new Polly.Net40Async project/package supporting async for .NET40 via Microsoft.Bcl.Async
• @SteveCote - Added overloads to WaitAndRetry and WaitAndRetryAsync methods that accept an onRetry delegate which includes the attempt count.
• @reisenberger - Allowed policies to handle returned results; added strongly-typed policies Policy<TResult>;.
• @christopherbahr - Added optimisation for circuit-breaker hot path.
• @Finity - Fixed circuit-breaker threshold bug.
• @reisenberger - Add some missing ExecuteAndCapture/Async overloads.
• @brunolauze - Add CancellationToken support to synchronous executions (to support TimeoutPolicy).
• @reisenberger - Add PolicyWrap.
• @reisenberger - Add Fallback policy.
• @reisenberger - Add PolicyKeys and context to all policy executions, as bedrock for policy events and metrics tracking executions.
• @reisenberger, and contributions from @brunolauze - Add Bulkhead Isolation policy.
• @reisenberger - Add Timeout policy.
• @reisenberger - Fix .NETStandard 1.0 targeting. Remove PCL259 target. PCL259 support is provided via .NETStandard1.0 target, going forward.
• @reisenberger - Fix CircuitBreaker HalfOpen state and cases when breakDuration is shorter than typical call timeout. Thanks to @vgouw and @kharos for the reports and insightful thinking.
• @lakario - Tidy CircuitBreaker LastException property.
• @lakario - Add NoOpPolicy.
• @Julien-Mialon - Fixes, support and examples for .NETStandard compatibility with Xamarin PCL projects
• @reisenberger - Add mutable Context and extra overloads taking Context. Allows different parts of a policy execution to exchange data via the mutable Context travelling with each execution.
• @ankitbko - Add PolicyRegistry for storing and retrieving policies.
• @reisenberger - Add interfaces by policy type and execution type.
• @seanfarrow - Add IReadOnlyPolicyRegistry interface.
• @kesmy - Migrate solution to msbuild15, banish project.json and packages.config
• @hambudi - Ensure sync TimeoutPolicy with TimeoutStrategy.Pessimistic rethrows delegate exceptions without additional AggregateException.
• @jiimaho and @Extremo75 - Provide public factory methods for PolicyResult, to support testing.
• @Extremo75 - Allow fallback delegates to take handled fault as input parameter.
• @reisenberger and @seanfarrow - Add CachePolicy, with interfaces for pluggable cache providers and serializers.
• Thanks to the awesome devs at @tretton37 who delivered the following as part of a one-day in-company hackathon led by @reisenberger, sponsored by @tretton37 and convened by @thecodejunkie
  • @matst80 - Allow WaitAndRetry to take handled fault as an input to the sleepDurationProvider, allowing WaitAndRetry to take account of systems which specify a duration to wait as part of a fault response; eg Azure CosmosDB may specify this in x-ms-retry-after-ms headers or in a property to an exception thrown by the Azure CosmosDB SDK.
  • @MartinSStewart - Add GetPolicies() extension methods to IPolicyWrap.
  • @jbergens37 - Parallelize test running where possible, to improve overall build speed.
• @reisenberger - Add new .HandleInner(...) syntax for handling inner exceptions natively.
• @rjongeneelen and @reisenberger - Allow PolicyWrap configuration to configure policies via interfaces.
• @reisenberger - Performance improvements.
• @awarrenlove - Add ability to calculate cache Ttl based on item to cache.
• @erickhouse - Add a new onBreak overload that provides the prior state on a transition to an open state.
• @benagain - Bug fix: RelativeTtl in CachePolicy now always returns a ttl relative to time item is cached.
• @urig - Allow TimeoutPolicy to be configured with Timeout.InfiniteTimeSpan.

Sample Projects

Polly-Samples contains practical examples for using various implementations of Polly. Please feel free to contribute to the Polly-Samples repository in order to assist others who are either learning Polly for the first time, or are seeking advanced examples and novel approaches provided by our generous community.

Instructions for Contributing

Please check out our Wiki for contributing guidelines. We are following the excellent GitHub Flow process, and would like to make sure you have all of the information needed to be a world-class contributor!

Since Polly is part of the .NET Foundation, we ask our contributors to abide by their Code of Conduct. To contribute (beyond trivial typo corrections), review and sign the .Net Foundation Contributor License Agreement. This ensures the community is free to use your contributions. The registration process can be completed entirely online.

Also, we've stood up a Slack channel for easier real-time discussion of ideas and the general direction of Polly as a whole. Be sure to join the conversation today!

License

Licensed under the terms of the New BSD License

Blog posts, podcasts and videos about Polly

When we discover an interesting write-up on Polly, we'll add it to this list. If you have a blog post you'd like to share, please submit a PR!

Blog posts

• Adding resilience and Transient Fault handling to your .NET Core HttpClient with Polly - by Scott Hanselman
• Reliable Event Processing in Azure Functions - by Jeff Hollan
• Resilient network connectivity in Xamarin Forms - by Adam Pedley
• Policy recommendations for Azure Cognitive Services - by Joel Hulen
• Using Polly with F# async workflows - by Mark Seemann
• Azure SQL transient errors - by Mattias Karlsson
• Polly series on No Dogma blog - by Bryan Hogan
• Polly 5.0 - a wider resilience framework! - by Dylan Reisenberger
• Implementing the retry pattern in c sharp using Polly - by Alastair Crabtree
• NuGet Package of the Week: Polly wanna fluently express transient exception handling policies in .NET? - by Scott Hanselman
• Circuit Breaking with Polly - by J. Conway
• Exception handling policies with Polly - by Mark Timmings
• When you use the Polly circuit-breaker, make sure you share your Policy instances! - by Andrew Lock
• Polly is Repetitive, and I love it! - by Joel Hulen

Podcasts

• Dylan Reisenberger sits down virtually with Bryan Hogan of NoDogmaBlog for an Introduction to Polly podcast. Why do I need Polly? History of the Polly project. What do we mean by resilience and transient faults? How retry and circuit-breaker help. Exponential backoff. Stability patterns. Bulkhead isolation. Future directions (as at April 2017).

PluralSight course

• Bryan Hogan of the NoDogmaBlog has authored a PluralSight course on Polly. The course takes you through all the major features of Polly as at v5.6.1. The course examples are based around using Polly for fault tolerance when calling remote web services, but the principles and techniques are applicable to any context in which Polly may be used.

Twitter

• Follow Dylan Reisenberger on twitter for notification of new Polly releases, advance notice of new proposals, tweets of interesting resilience articles (no junk).

Videos

• From MVP Houssem Dellai, a youtube video on How to use Polly with Xamarin Apps, covering wait-and-retry and discussing circuit-breaker policy with a demonstration in Xamarin Forms. Here is the source code of the application demonstrated in the video. Draws on the ResilientHttpClient from Microsoft's eShopOnContainers project.
• In the video, .NET Rocks Live with Jon Skeet and Bill Wagner, Bill Wagner discusses Polly.
• Scott Allen discusses Polly during his Building for Resiliency and Scale in the Cloud presentation at NDC.