Auto‑Optimizing Resilience4j Circuit Breaker Settings with Spring AI

1. Introduction

In microservice systems, frequent service failures and timeouts can cause cascading failures. The default static configuration of Resilience4j circuit breakers does not adapt to changing traffic and runtime conditions. This tutorial shows how to integrate Spring AI to automatically analyze runtime metrics and generate optimal circuit‑breaker parameters, achieving adaptive tuning and higher fault‑tolerance.

2. Environment

Spring Boot 3.5.0

<dependency>
  <groupId>org.springframework.boot</groupId>
  <artifactId>spring-boot-starter-aop</artifactId>
</dependency>
<dependency>
  <groupId>io.github.resilience4j</groupId>
  <artifactId>resilience4j-spring-boot3</artifactId>
  <version>2.3.0</version>
</dependency>
<dependency>
  <groupId>com.alibaba.cloud.ai</groupId>
  <artifactId>spring-ai-alibaba-starter-dashscope</artifactId>
  <version>1.1.2.0</version>
</dependency>

Basic YAML configuration:

resilience4j:
  circuitbreaker:
    instances:
      businessService:
        registerHealthIndicator: true
        slidingWindowSize: 100
        minimumNumberOfCalls: 2
        failureRateThreshold: 10
        waitDurationInOpenState: 10s
        permittedNumberOfCallsInHalfOpenState: 3
        automaticTransitionFromOpenToHalfOpenEnabled: true
---
spring:
  ai:
    dashscope:
      api-key: ${DASHSCOPE_API_KEY}
      base-url: https://dashscope.aliyuncs.com
      chat:
        options:
          stream: true
          model: qwen-turbo
          temperature: 0.1

3. Basic circuit‑breaker usage

Service method with @CircuitBreaker annotation:

@CircuitBreaker(name = "businessService", fallbackMethod = "queryFallback")
public String query(Long id) {
    System.err.println("Executing business query [%s]".formatted(id));
    if (id == 666L) {
        throw new RuntimeException("Invalid parameter");
    }
    return "Apple";
}

public String queryFallback(Throwable e) {
    logger.error("Fallback: query error {}", e.getMessage());
    return "fallback-query";
}

Controller exposing the service:

private final BusinessService businessService;

@GetMapping("/{id}")
public ResponseEntity<String> query(@PathVariable Long id) {
    return ResponseEntity.ok(businessService.query(id));
}

When the endpoint is called with id = 666, the circuit breaker opens after two failed calls, preventing the actual method execution.

4. AI‑driven dynamic tuning

A scheduled component collects the current circuit‑breaker metrics, builds a prompt for the AI model, receives recommended parameters, and replaces the existing configuration at runtime.

@Component
public class AiControllerResilience {
    private static final String CB_NAME = "businessService";
    private final CircuitBreakerRegistry circuitBreakerRegistry;
    private final ChatClient chatClient;

    @Scheduled(fixedRate = 10000)
    public void tuneAiCircuitBreaker() throws Exception {
        // 1. Get current circuit‑breaker and its metrics
        CircuitBreaker cb = circuitBreakerRegistry.circuitBreaker(CB_NAME);
        CircuitBreaker.Metrics metrics = cb.getMetrics();
        // 2. Build AI prompt with current config and metrics
        CircuitBreakerConfig config = cb.getCircuitBreakerConfig();
        String prompt = String.format("""
            # 1 Analyze Resilience4j service metrics
            ## 1.1 Current circuit‑breaker config
            | Field | Current Value |
            | ----- | ------------- |
            | slidingWindowSize | %s |
            | minimumNumberOfCalls | %s |
            | failureRateThreshold | %s |
            | waitDurationInOpenState | %s |
            | permittedNumberOfCallsInHalfOpenState | %s |
            ## 1.2 Runtime metrics
            %s
            # 2 Desired result
            Provide a set of optimal parameters (slidingWindowSize, minimumNumberOfCalls, failureRateThreshold, waitDurationInOpenState, permittedNumberOfCallsInHalfOpenState) and a reason if any parameter changes.
            """,
            config.getSlidingWindowSize(),
            config.getMinimumNumberOfCalls(),
            config.getFailureRateThreshold(),
            config.getWaitIntervalFunctionInOpenState().apply(1),
            config.getPermittedNumberOfCallsInHalfOpenState(),
            new ObjectMapper().writeValueAsString(metrics));
        // 3. Call AI model
        AiConfigRecommend recommend = chatClient.prompt().user(prompt).call().entity(AiConfigRecommend.class);
        // 4. Build new config
        CircuitBreakerConfig newCfg = CircuitBreakerConfig.custom()
            .failureRateThreshold(recommend.getFailureRateThreshold())
            .slidingWindowSize(recommend.getSlidingWindowSize())
            .minimumNumberOfCalls(recommend.getMinimumNumberOfCalls())
            .waitDurationInOpenState(Duration.ofMillis(recommend.getWaitDurationInOpenState()))
            .permittedNumberOfCallsInHalfOpenState(recommend.getPermittedNumberOfCallsInHalfOpenState())
            .build();
        // 5. Replace configuration
        circuitBreakerRegistry.remove(CB_NAME);
        circuitBreakerRegistry.circuitBreaker(CB_NAME, newCfg);
        System.err.println("Old config: slidingWindowSize = %s, minimumNumberOfCalls = %s, failureRateThreshold = %s, waitDurationInOpenState = %s, permittedNumberOfCallsInHalfOpenState = %s"
            .formatted(config.getSlidingWindowSize(), config.getMinimumNumberOfCalls(), config.getFailureRateThreshold(), config.getWaitIntervalFunctionInOpenState().apply(1), config.getPermittedNumberOfCallsInHalfOpenState()));
        System.err.println("New config: slidingWindowSize = %s, minimumNumberOfCalls = %s, failureRateThreshold = %s, waitDurationInOpenState = %s, permittedNumberOfCallsInHalfOpenState = %s, reason = %s"
            .formatted(recommend.getSlidingWindowSize(), recommend.getMinimumNumberOfCalls(), recommend.getFailureRateThreshold(), recommend.getWaitDurationInOpenState(), recommend.getPermittedNumberOfCallsInHalfOpenState(), recommend.getReason()));
    }

    public static class AiConfigRecommend {
        private int slidingWindowSize;
        private int minimumNumberOfCalls;
        private float failureRateThreshold;
        private long waitDurationInOpenState;
        private int permittedNumberOfCallsInHalfOpenState;
        private String reason;
        // getters and setters omitted for brevity
    }
}

5. Result

After restarting the service and invoking the endpoint with both valid and error parameters, the console prints the old and new configuration values, confirming that the circuit‑breaker settings have been updated automatically.

Console example:

Final observation: the businessService circuit‑breaker configuration is now dynamically tuned by AI, eliminating manual parameter adjustments.