Skip to content

Latest commit

 

History

History
 
 

riptide-failsafe

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
src
src
 
 
README.md
README.md
 
 
pom.xml
pom.xml
 
 

README.md

Riptide: Failsafe

Valves

Javadoc Maven Central

Riptide: Failsafe adds Failsafe support to Riptide. It offers retries and a circuit breaker to every remote call.

Example

Http.builder().requestFactory(new HttpComponentsClientHttpRequestFactory())
    .plugin(new FailsafePlugin()
        .withPolicy(circuitBreaker)
        .withPolicy(new RetryRequestPolicy(retryPolicy)))
    .build();

Features

  • seamlessly integrates Riptide with Failsafe

Dependencies

  • Riptide Core
  • Failsafe

Installation

Add the following dependency to your project:

<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>riptide-failsafe</artifactId>
    <version>${riptide.version}</version>
</dependency>

Configuration

The failsafe plugin will not perform retries nor apply circuit breakers unless they were explicitly configured:

Http.builder().requestFactory(new HttpComponentsClientHttpRequestFactory())
    .plugin(new FailsafePlugin()
        .withPolicy(
            new RetryRequestPolicy(
                RetryPolicy.<ClientHttpResponse>builder()
                    .withDelay(Duration.ofMillis(25))
                    .withDelayFn(new RetryAfterDelayFunction(clock))
                    .withMaxRetries(4)
                    .build())
                .withListener(myRetryListener))
        .withPolicy(
            CircuitBreaker.<ClientHttpResponse>builder()
                .withFailureThreshold(3, 10)
                .withSuccessThreshold(5)
                .withDelay(Duration.ofMinutes(1))
                .build()))
    .build();

Please visit the Failsafe readme in order to see possible configurations.

Retries

Beware when using retryOn to retry conditionally on certain exception types. You'll need to register RetryException in order for the retry() route to work:

RetryPolicy.<ClientHttpResponse>builder()
    .handle(SocketTimeoutException.class)
    .handle(RetryException.class)
    .build();

Failsafe supports dynamically computed delays using a custom function.

Riptide: Failsafe offers implementations that understand:

Http.builder().requestFactory(new HttpComponentsClientHttpRequestFactory())
    .plugin(new FailsafePlugin()
        .withPolicy(RetryPolicy.<ClientHttpResponse>builder()
            .withDelayFn(new CompositeDelayFunction<>(Arrays.asList(
                new RetryAfterDelayFunction(clock),
                new RateLimitResetDelayFunction(clock)
            )))
            .withMaxDuration(Duration.ofSeconds(5))
            .build()))
    .build();

⚠️ Make sure you you specify a max duration otherwise any value coming from a server that is further ahead in the future will make your retry block practically forever.

Make sure you check out zalando/failsafe-actuator for a seamless integration of Failsafe and Spring Boot.

Backup Requests

The BackupRequest policy implements the backup request pattern, also known as hedged requests:

Http.builder().requestFactory(new HttpComponentsClientHttpRequestFactory())
    .plugin(new FailsafePlugin()
        .withPolicy(new BackupRequest(1, SECONDS)))
    .build();

Custom executor

The withExecutor method allows to specify a custom ExecutorService being used to perform asynchronous executions and listen for callbacks:

Http.builder().requestFactory(new HttpComponentsClientHttpRequestFactory())
    .plugin(new FailsafePlugin()
        .withPolicy(
            CircuitBreaker.<ClientHttpResponse>builder()
                .withFailureThreshold(3, 10)
                .withSuccessThreshold(5)
                .withDelay(Duration.ofMinutes(1))
                .build())
        .withExecutor(Executors.newFixedThreadPool(2)))
    .build();

If no executor is specified, the default executor configured by Failsafe is used. See Failsafe DelegatingScheduler class, and also Failsafe documentation for more information.

Beware when specifying a custom ExecutorService:

  1. The ExecutorService should have a core pool size or parallelism of at least 2 in order for timeouts to work
  2. In general, it is not recommended to specify the same ExecutorService for multiple Http clients

Usage

Given the failsafe plugin was configured as shown in the last section: A regular call like the following will now be retried up to 4 times if the server did not respond within the socket timeout.

http.get("/users/me")
    .dispatch(series(),
        on(SUCCESSFUL).call(User.class, this::greet),
        anySeries().call(problemHandling()))

Handling certain technical issues automatically, like socket timeouts, is quite useful. But there might be cases where the server did respond, but the response indicates something that is worth retrying, e.g. a 409 Conflict or a 503 Service Unavailable. Use the predefined retry route that comes with the failsafe plugin:

http.get("/users/me")
    .dispatch(series(),
        on(SUCCESSFUL).call(User.class, this::greet),
        on(CLIENT_ERROR).dispatch(status(),
            on(CONFLICT).call(retry())),
        on(SERVER_ERROR).dispatch(status(),
            on(SERVICE_UNAVAILABLE).call(retry())),
        anySeries().call(problemHandling()))

Safe and Idempotent methods

Only safe and idempontent methods are retried by default. The following request methods can be detected:

You also have the option to declare any request to be idempotent by setting the respective request attribute. This is useful in situation where none of the options are above would detect it but based on the contract of the API you may know that a certain operation is in fact idempotent.

http.post("/subscriptions/{id}/cursors", subscriptionId)
    .attribute(MethodDetector.IDEMPOTENT, true)
    .header("X-Nakadi-StreamId", streamId)
    .body(cursors)
    .dispatch(series(),
        on(SUCCESSFUL).call(pass()),
        anySeries().call(problemHandling()))

In case those options are insufficient you may specify your own method detector:

Http.builder().requestFactory(new HttpComponentsClientHttpRequestFactory())
    .plugin(new FailsafePlugin()
        .withPolicy(retryPolicy)
        .withDecorator(new CustomIdempotentMethodDetector()))
    .build();

Getting Help

If you have questions, concerns, bug reports, etc., please file an issue in this repository's Issue Tracker.

Getting Involved/Contributing

To contribute, simply open a pull request and add a brief description (1-2 sentences) of your addition or change. For more details, check the contribution guidelines.

Credits and references