By Version

Edit this Page

SmallRye Fault Tolerance

One of the challenges brought by the distributed nature of microservices is that communication with external systems is inherently unreliable. This increases demand on resiliency of applications. To simplify making more resilient applications, Quarkus provides SmallRye Fault Tolerance, an implementation of the MicroProfile Fault Tolerance specification.

In this guide, we demonstrate usage of MicroProfile Fault Tolerance annotations such as @Timeout, @Fallback, @Retry, @CircuitBreaker, and @RateLimit.

Prerequisites

To complete this guide, you need:

Roughly 15 minutes
An IDE
JDK 17+ installed with JAVA_HOME configured appropriately
Apache Maven 3.9.9
Optionally the Quarkus CLI if you want to use it
Optionally Mandrel or GraalVM installed and configured appropriately if you want to build a native executable (or Docker if you use a native container build)

The Scenario

The application built in this guide simulates a simple backend for a gourmet coffee e-shop. It implements a REST endpoint providing information about coffee samples we have on store.

Let’s imagine, although it’s not implemented as such, that some methods in our endpoint require communication to external services like a database or an external microservice, which introduces a factor of unreliability.

Solution

We recommend that you follow the instructions in the next sections and create the application step by step. However, you can go right to the completed example.

Clone the Git repository: git clone https://github.com/quarkusio/quarkus-quickstarts.git, or download an archive.

The solution is located in the microprofile-fault-tolerance-quickstart directory.

Creating the Maven Project

First, we need a new project. Create a new project with the following command:

CLI

quarkus create app org.acme:microprofile-fault-tolerance-quickstart \
    --extension='smallrye-fault-tolerance,rest-jackson' \
    --no-code
cd microprofile-fault-tolerance-quickstart

To create a Gradle project, add the --gradle or --gradle-kotlin-dsl option.

For more information about how to install and use the Quarkus CLI, see the Quarkus CLI guide.

Maven

mvn io.quarkus.platform:quarkus-maven-plugin:3.16.1:create \
    -DprojectGroupId=org.acme \
    -DprojectArtifactId=microprofile-fault-tolerance-quickstart \
    -Dextensions='smallrye-fault-tolerance,rest-jackson' \
    -DnoCode
cd microprofile-fault-tolerance-quickstart

To create a Gradle project, add the -DbuildTool=gradle or -DbuildTool=gradle-kotlin-dsl option.

For Windows users:

If using cmd, (don’t use backward slash \ and put everything on the same line)
If using Powershell, wrap -D parameters in double quotes e.g. "-DprojectArtifactId=microprofile-fault-tolerance-quickstart"

This command generates a project, importing the extensions for Quarkus REST (formerly RESTEasy Reactive)/Jakarta REST and SmallRye Fault Tolerance.

If you already have your Quarkus project configured, you can add the smallrye-fault-tolerance extension to your project by running the following command in your project base directory:

CLI

quarkus extension add smallrye-fault-tolerance

Maven

./mvnw quarkus:add-extension -Dextensions='smallrye-fault-tolerance'

Gradle

./gradlew addExtension --extensions='smallrye-fault-tolerance'

This will add the following to your build file:

pom.xml

<dependency>
    <groupId>io.quarkus</groupId>
    <artifactId>quarkus-smallrye-fault-tolerance</artifactId>
</dependency>

build.gradle

implementation("io.quarkus:quarkus-smallrye-fault-tolerance")

Preparing an Application: REST Endpoint and CDI Bean

In this section we create a skeleton of our application, so that we have something that we can extend and to which we can add fault tolerance features later on.

First, create a simple entity representing a coffee sample in our store:

package org.acme.microprofile.faulttolerance;

public class Coffee {

    public Integer id;
    public String name;
    public String countryOfOrigin;
    public Integer price;

    public Coffee() {
    }

    public Coffee(Integer id, String name, String countryOfOrigin, Integer price) {
        this.id = id;
        this.name = name;
        this.countryOfOrigin = countryOfOrigin;
        this.price = price;
    }
}

Let’s continue with a simple CDI bean, that would work as a repository of our coffee samples.

package org.acme.microprofile.faulttolerance;

import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.stream.Collectors;
import jakarta.enterprise.context.ApplicationScoped;

@ApplicationScoped
public class CoffeeRepositoryService {

    private Map<Integer, Coffee> coffeeList = new HashMap<>();

    public CoffeeRepositoryService() {
        coffeeList.put(1, new Coffee(1, "Fernandez Espresso", "Colombia", 23));
        coffeeList.put(2, new Coffee(2, "La Scala Whole Beans", "Bolivia", 18));
        coffeeList.put(3, new Coffee(3, "Dak Lak Filter", "Vietnam", 25));
    }

    public List<Coffee> getAllCoffees() {
        return new ArrayList<>(coffeeList.values());
    }

    public Coffee getCoffeeById(Integer id) {
        return coffeeList.get(id);
    }

    public List<Coffee> getRecommendations(Integer id) {
        if (id == null) {
            return Collections.emptyList();
        }
        return coffeeList.values().stream()
                .filter(coffee -> !id.equals(coffee.id))
                .limit(2)
                .collect(Collectors.toList());
    }
}

Finally, create the org.acme.microprofile.faulttolerance.CoffeeResource class as follows:

package org.acme.microprofile.faulttolerance;

import java.util.List;
import java.util.Random;
import java.util.concurrent.atomic.AtomicLong;
import jakarta.inject.Inject;
import jakarta.ws.rs.GET;
import jakarta.ws.rs.Path;

import org.jboss.logging.Logger;

@Path("/coffee")
public class CoffeeResource {

    private static final Logger LOGGER = Logger.getLogger(CoffeeResource.class);

    @Inject
    CoffeeRepositoryService coffeeRepository;

    private AtomicLong counter = new AtomicLong(0);

    @GET
    public List<Coffee> coffees() {
        final Long invocationNumber = counter.getAndIncrement();

        maybeFail(String.format("CoffeeResource#coffees() invocation #%d failed", invocationNumber));

        LOGGER.infof("CoffeeResource#coffees() invocation #%d returning successfully", invocationNumber);
        return coffeeRepository.getAllCoffees();
    }

    private void maybeFail(String failureLogMessage) {
        if (new Random().nextBoolean()) {
            LOGGER.error(failureLogMessage);
            throw new RuntimeException("Resource failure.");
        }
    }
}

At this point, we expose a single REST method that will show a list of coffee samples in a JSON format. Note that we introduced some fault making code in our CoffeeResource#maybeFail() method, which is going to cause failures in the CoffeeResource#coffees() endpoint method in about 50 % of requests.

Why not check that our application works? Run the Quarkus development server with:

CLI

quarkus dev

Maven

./mvnw quarkus:dev

Gradle

./gradlew --console=plain quarkusDev

and open http://localhost:8080/coffee in your browser. Make a couple of requests (remember, some of them are expected to fail). At least some requests should show us the list of our coffee samples in JSON, the rest will fail with a RuntimeException thrown in CoffeeResource#maybeFail().

Congratulations, you’ve just made a working (although somewhat unreliable) Quarkus application!

Adding Resiliency: Retries

Let the Quarkus development server running and in your IDE add the @Retry annotation to the CoffeeResource#coffees() method as follows and save the file:

import org.eclipse.microprofile.faulttolerance.Retry;
...

public class CoffeeResource {
    ...
    @GET
    @Retry(maxRetries = 4)
    public List<Coffee> coffees() {
        ...
    }
    ...
}

Hit refresh in your browser. The Quarkus development server will automatically detect the changes and recompile the app for you, so there’s no need to restart it.

You can hit refresh couple more times. Practically all requests should now be succeeding. The CoffeeResource#coffees() method is still in fact failing in about 50 % of time, but every time it happens, the platform will automatically retry the call!

To see that the failures still happen, check the output of the development server. The log messages should be similar to these:

2019-03-06 12:17:41,725 INFO  [org.acm.fau.CoffeeResource] (XNIO-1 task-1) CoffeeResource#coffees() invocation #5 returning successfully
2019-03-06 12:17:44,187 INFO  [org.acm.fau.CoffeeResource] (XNIO-1 task-1) CoffeeResource#coffees() invocation #6 returning successfully
2019-03-06 12:17:45,166 ERROR [org.acm.fau.CoffeeResource] (XNIO-1 task-1) CoffeeResource#coffees() invocation #7 failed
2019-03-06 12:17:45,172 ERROR [org.acm.fau.CoffeeResource] (XNIO-1 task-1) CoffeeResource#coffees() invocation #8 failed
2019-03-06 12:17:45,176 INFO  [org.acm.fau.CoffeeResource] (XNIO-1 task-1) CoffeeResource#coffees() invocation #9 returning successfully

You can see that every time an invocation fails, it’s immediately followed by another invocation, until one succeeds. Since we allowed 4 retries, it would require 5 invocations to fail in a row, in order for the user to be actually exposed to a failure. Which is fairly unlikely to happen.

Adding Resiliency: Timeouts

So what else have we got in MicroProfile Fault Tolerance? Let’s look into timeouts.

Add following two methods to our CoffeeResource endpoint. Again, no need to restart the server, just paste the code and save the file.

import org.eclipse.microprofile.faulttolerance.Timeout;
...
public class CoffeeResource {
    ...
    @GET
    @Path("/{id}/recommendations")
    @Timeout(250)
    public List<Coffee> recommendations(int id) {
        long started = System.currentTimeMillis();
        final long invocationNumber = counter.getAndIncrement();

        try {
            randomDelay();
            LOGGER.infof("CoffeeResource#recommendations() invocation #%d returning successfully", invocationNumber);
            return coffeeRepository.getRecommendations(id);
        } catch (InterruptedException e) {
            LOGGER.errorf("CoffeeResource#recommendations() invocation #%d timed out after %d ms",
                    invocationNumber, System.currentTimeMillis() - started);
            return null;
        }
    }

    private void randomDelay() throws InterruptedException {
        Thread.sleep(new Random().nextInt(500));
    }
}

We added some new functionality. We want to be able to recommend some related coffees based on a coffee that a user is currently looking at. It’s not a critical functionality, it’s a nice-to-have. When the system is overloaded and the logic behind obtaining recommendations takes too long to execute, we would rather time out and render the UI without recommendations.

Note that the timeout was configured to 250 ms, and a random artificial delay between 0 and 500 ms was introduced into the CoffeeResource#recommendations() method.

In your browser, go to http://localhost:8080/coffee/2/recommendations and hit refresh a couple of times.

You should see some requests time out with org.eclipse.microprofile.faulttolerance.exceptions.TimeoutException. Requests that do not time out should show two recommended coffee samples in JSON.

Adding Resiliency: Fallbacks

Let’s make our recommendations feature even better by providing a fallback (and presumably faster) way of getting related coffees.

Add a fallback method to CoffeeResource and a @Fallback annotation to CoffeeResource#recommendations() method as follows:

import java.util.Collections;
import org.eclipse.microprofile.faulttolerance.Fallback;

...
public class CoffeeResource {
    ...
    @Fallback(fallbackMethod = "fallbackRecommendations")
    public List<Coffee> recommendations(int id) {
        ...
    }

    public List<Coffee> fallbackRecommendations(int id) {
        LOGGER.info("Falling back to RecommendationResource#fallbackRecommendations()");
        // safe bet, return something that everybody likes
        return Collections.singletonList(coffeeRepository.getCoffeeById(1));
    }
    ...
}

Hit refresh several times on http://localhost:8080/coffee/2/recommendations. The TimeoutException should not appear anymore. Instead, in case of a timeout, the page will display a single recommendation that we hardcoded in our fallback method fallbackRecommendations(), rather than two recommendations returned by the original method.

Check the server output to see that fallback is really happening:

2020-01-09 13:21:34,250 INFO  [org.acm.fau.CoffeeResource] (executor-thread-1) CoffeeResource#recommendations() invocation #1 returning successfully
2020-01-09 13:21:36,354 ERROR [org.acm.fau.CoffeeResource] (executor-thread-1) CoffeeResource#recommendations() invocation #2 timed out after 250 ms
2020-01-09 13:21:36,355 INFO  [org.acm.fau.CoffeeResource] (executor-thread-1) Falling back to RecommendationResource#fallbackRecommendations()

The fallback method is required to have the same parameters as the original method.

Adding Resiliency: Circuit Breaker

A circuit breaker is useful for limiting number of failures happening in the system, when part of the system becomes temporarily unstable. The circuit breaker records successful and failed invocations of a method, and when the ratio of failed invocations reaches the specified threshold, the circuit breaker opens and blocks all further invocations of that method for a given time.

Add the following code into the CoffeeRepositoryService bean, so that we can demonstrate a circuit breaker in action:

import java.util.concurrent.atomic.AtomicLong;
import org.eclipse.microprofile.faulttolerance.CircuitBreaker;
...

public class CoffeeRepositoryService {
    ...

    private AtomicLong counter = new AtomicLong(0);

    @CircuitBreaker(requestVolumeThreshold = 4)
    public Integer getAvailability(Coffee coffee) {
        maybeFail();
        return new Random().nextInt(30);
    }

    private void maybeFail() {
        // introduce some artificial failures
        final Long invocationNumber = counter.getAndIncrement();
        if (invocationNumber % 4 > 1) { // alternate 2 successful and 2 failing invocations
            throw new RuntimeException("Service failed.");
        }
    }
}

And inject the code below into the CoffeeResource endpoint:

public class CoffeeResource {
    ...
    @Path("/{id}/availability")
    @GET
    public Response availability(int id) {
        final Long invocationNumber = counter.getAndIncrement();

        Coffee coffee = coffeeRepository.getCoffeeById(id);
        // check that coffee with given id exists, return 404 if not
        if (coffee == null) {
            return Response.status(Response.Status.NOT_FOUND).build();
        }

        try {
            Integer availability = coffeeRepository.getAvailability(coffee);
            LOGGER.infof("CoffeeResource#availability() invocation #%d returning successfully", invocationNumber);
            return Response.ok(availability).build();
        } catch (RuntimeException e) {
            String message = e.getClass().getSimpleName() + ": " + e.getMessage();
            LOGGER.errorf("CoffeeResource#availability() invocation #%d failed: %s", invocationNumber, message);
            return Response.status(Response.Status.INTERNAL_SERVER_ERROR)
                    .entity(message)
                    .type(MediaType.TEXT_PLAIN_TYPE)
                    .build();
        }
    }
    ...
}

We added another functionality - the application can return the amount of remaining packages of given coffee on our store (just a random number).

This time an artificial failure was introduced in the CDI bean: the CoffeeRepositoryService#getAvailability() method is going to alternate between two successful and two failed invocations.

We also added a @CircuitBreaker annotation with requestVolumeThreshold = 4. CircuitBreaker.failureRatio is by default 0.5, and CircuitBreaker.delay is by default 5 seconds. That means that a circuit breaker will open when 2 of the last 4 invocations failed, and it will stay open for 5 seconds.

To test this out, do the following:

Go to http://localhost:8080/coffee/2/availability in your browser. You should see a number being returned.
Hit refresh, this second request should again be successful and return a number.
Refresh two more times. Both times you should see text "RuntimeException: Service failed.", which is the exception thrown by CoffeeRepositoryService#getAvailability().
Refresh a couple more times. Unless you waited too long, you should again see exception, but this time it’s "CircuitBreakerOpenException: getAvailability". This exception indicates that the circuit breaker opened and the CoffeeRepositoryService#getAvailability() method is not being called anymore.
Give it 5 seconds during which circuit breaker should close, and you should be able to make two successful requests again.

Adding Resiliency: Rate Limits

This is an additional feature of SmallRye Fault Tolerance and is not specified by MicroProfile Fault Tolerance.

It is possible to prevent an operation from being executed too often using a rate limit. Rate limits enforce a maximum number of permitted invocations in a time window of some length. For example, with a rate limit, one can make sure that a method may only be called 50 times per minute. Invocations that would exceed the limit are rejected with an exception of type RateLimitException.

Additionally, it is possible to define minimum spacing between invocations. For example, with a minimum spacing of 1 second, if a second invocation happens 500 millis after the first, it is rejected even if the limit would not be exceeded yet.

Rate limit is superficially similar to a bulkhead (concurrency limit), but is in fact quite different. Bulkhead limits the number of executions happening concurrently at any point in time. Rate limit limits the number of executions in a time window of some length, without considering concurrency.

Rate limits need to maintain some state between invocations: the number of recent invocations, the time stamp of last invocation, and so on. This state is a singleton, irrespective of the lifecycle of the bean that uses the @RateLimit annotation.

More specifically, the rate limit state is uniquely identified by the combination of the bean class (java.lang.Class) and the method object (java.lang.reflect.Method) representing the guarded method.

Let the Quarkus development mode run and in your IDE add the @RateLimit annotation to the CoffeeResource#coffees() method as follows and save the file:

import java.time.temporal.ChronoUnit;
import io.smallrye.faulttolerance.api.RateLimit;

...
public class CoffeeResource {
    ...
    @GET
    @RateLimit(value = 2, window = 10, windowUnit = ChronoUnit.SECONDS)
    public List<Coffee> coffees() {
        ...
    }
    ...
}

Hit refresh in your browser. The Quarkus development server will automatically detect the changes and recompile the app for you, so there’s no need to restart it.

You can hit refresh a couple more times. After 2 requests within a 10 second interval you should start seeing errors in the logs, similar to these:

INFO  [org.acm.mic.fau.CoffeeResource] (executor-thread-1) CoffeeResource#coffees() invocation #1 returning successfully
ERROR [io.qua.ver.htt.run.QuarkusErrorHandler] (executor-thread-1) HTTP Request to /coffee failed, error id: d3e59090-fd45-4c67-844e-80a8f7fa6ee0-4: io.smallrye.faulttolerance.api.RateLimitException: org.acme.microprofile.faulttolerance.CoffeeResource#coffees rate limit exceeded
        at io.smallrye.faulttolerance.core.rate.limit.RateLimit.doApply(RateLimit.java:58)
        at io.smallrye.faulttolerance.core.rate.limit.RateLimit.apply(RateLimit.java:44)
        at io.smallrye.faulttolerance.FaultToleranceInterceptor.syncFlow(FaultToleranceInterceptor.java:255)
        at io.smallrye.faulttolerance.FaultToleranceInterceptor.intercept(FaultToleranceInterceptor.java:182)
        at io.smallrye.faulttolerance.FaultToleranceInterceptor_Bean.intercept(Unknown Source)
        at io.quarkus.arc.impl.InterceptorInvocation.invoke(InterceptorInvocation.java:42)
        at io.quarkus.arc.impl.AroundInvokeInvocationContext.perform(AroundInvokeInvocationContext.java:30)
        at io.quarkus.arc.impl.InvocationContexts.performAroundInvoke(InvocationContexts.java:27)
        at org.acme.microprofile.faulttolerance.CoffeeResource_Subclass.coffees(Unknown Source)
        at org.acme.microprofile.faulttolerance.CoffeeResource$quarkusrestinvoker$coffees_73d7590ab944adfa130e4ad57c30282f825b2d18.invoke(Unknown Source)
        at org.jboss.resteasy.reactive.server.handlers.InvocationHandler.handle(InvocationHandler.java:29)
        at io.quarkus.resteasy.reactive.server.runtime.QuarkusResteasyReactiveRequestContext.invokeHandler(QuarkusResteasyReactiveRequestContext.java:141)
        at org.jboss.resteasy.reactive.common.core.AbstractResteasyReactiveContext.run(AbstractResteasyReactiveContext.java:147)
        at io.quarkus.vertx.core.runtime.VertxCoreRecorder$14.runWith(VertxCoreRecorder.java:599)
        at org.jboss.threads.EnhancedQueueExecutor$Task.doRunWith(EnhancedQueueExecutor.java:2516)
        at org.jboss.threads.EnhancedQueueExecutor$Task.run(EnhancedQueueExecutor.java:2495)
        at org.jboss.threads.EnhancedQueueExecutor$ThreadBody.run(EnhancedQueueExecutor.java:1521)
        at org.jboss.threads.DelegatingRunnable.run(DelegatingRunnable.java:11)
        at org.jboss.threads.ThreadLocalResettingRunnable.run(ThreadLocalResettingRunnable.java:11)
        at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
        at java.base/java.lang.Thread.run(Thread.java:1583)

If @Fallback is used with @RateLimit, the fallback method or handler may be invoked if a RateLimitException is thrown, depending on the fallback configuration.

If @Retry is used with @RateLimit, each retry attempt is processed by the rate limit as an independent invocation. If RateLimitException is thrown, the execution may be retried, depending on how the retry is configured.

If @CircuitBreaker is used with @RateLimit, the circuit breaker is checked before enforcing the rate limit. If rate limiting results in RateLimitException, this may be counted as a failure, depending on how the circuit breaker is configured.

Runtime configuration

You can override the annotations parameters at runtime inside your application.properties file.

If we take the retry example that we already saw:

package org.acme;

import org.eclipse.microprofile.faulttolerance.Retry;
...

public class CoffeeResource {
    ...
    @GET
    @Retry(maxRetries = 4)
    public List<Coffee> coffees() {
        ...
    }
    ...
}

We can override the maxRetries parameter with 6 retries instead of 4 by the following configuration item:

org.acme.CoffeeResource/coffees/Retry/maxRetries=6

The format is fully-qualified-class-name/method-name/annotation-name/property-name=value. You can also configure a property for all the annotation via annotation-name/property-name=value.

Conclusion

SmallRye Fault Tolerance allows to improve resiliency of your application, without having an impact on the complexity of our business logic.

All that is needed to enable the fault tolerance features in Quarkus is:

adding the smallrye-fault-tolerance Quarkus extension to your project using the quarkus-maven-plugin:

CLI

quarkus extension add smallrye-fault-tolerance

Maven

./mvnw quarkus:add-extension -Dextensions='smallrye-fault-tolerance'

Gradle

./gradlew addExtension --extensions='smallrye-fault-tolerance'

or simply adding the following Maven dependency:

pom.xml

<dependency>
    <groupId>io.quarkus</groupId>
    <artifactId>quarkus-smallrye-fault-tolerance</artifactId>
</dependency>

build.gradle

implementation("io.quarkus:quarkus-smallrye-fault-tolerance")

Additional resources

SmallRye Fault Tolerance has more features than shown here. Please check the SmallRye Fault Tolerance documentation to learn about them.

In Quarkus, you can use the SmallRye Fault Tolerance optional features out of the box.

Support for Mutiny is present, so your asynchronous methods can return Uni in addition to CompletionStage.

MicroProfile Context Propagation is integrated with Fault Tolerance, so existing contexts are automatically propagated to your asynchronous methods.

This also applies to the CDI request context: if it is active on the original thread, it is propagated to the new thread, but if it’s not, then the new thread won’t have it either. This is contrary to MicroProfile Fault Tolerance specification, which states that the request context must be active during the @Asynchronous method invocation.

We believe that in presence of MicroProfile Context Propagation, this requirement should not apply. The entire point of context propagation is to make sure the new thread has the same contexts as the original thread.

Non-compatible mode is enabled by default. This means that methods that return CompletionStage (or Uni) have asynchronous fault tolerance applied without any @Asynchronous, @AsynchronousNonBlocking, @Blocking or @NonBlocking annotation. It also means that circuit breaker, fallback and retry automatically inspect the exception cause chain if the exception itself is insufficient to decide what should happen.

This mode is not compatible with the MicroProfile Fault Tolerance specification, albeit the incompatibility is very small. To restore full compatibility, add this configuration property:

smallrye.faulttolerance.mp-compatibility=true

The programmatic API is present, including Mutiny support, and integrated with the declarative, annotation-based API. You can use the FaultTolerance and MutinyFaultTolerance APIs out of the box.

Support for Kotlin is present (assuming you use the Quarkus extension for Kotlin), so you can guard your suspend functions with fault tolerance annotations.

Metrics are automatically discovered and integrated. If your application uses the Quarkus extension for Micrometer, SmallRye Fault Tolerance will emit metrics to Micrometer. If your application uses the Quarkus extension for SmallRye Metrics, SmallRye Fault Tolerance will emit metrics to MicroProfile Metrics. Otherwise, SmallRye Fault Tolerance metrics will be disabled.

Prerequisites
The Scenario
Solution
Creating the Maven Project
Preparing an Application: REST Endpoint and CDI Bean
Adding Resiliency: Retries
Adding Resiliency: Timeouts
Adding Resiliency: Fallbacks
Adding Resiliency: Circuit Breaker
Adding Resiliency: Rate Limits
Runtime configuration
Conclusion
Additional resources