Spring Boot微服务异常：服务X重复调用服务Y问题求助

When using Feign with Eureka, repeated calls to a slow microservice almost always stem from default retry mechanisms in Feign/Ribbon or misconfigured timeout settings. Let’s break down the causes and actionable fixes:

1. Feign's Default Retry Behavior #

Feign includes a built-in retryer that automatically retries requests on certain failures—including perceived timeouts if not properly configured. This is likely triggering multiple calls when Microservice Y takes too long to respond.

Fix: Disable or Customize Feign Retries #

You can turn off retries entirely for your Feign client, or adjust the retry logic to fit your needs:

Option 1: Use a Configuration Class

Create a dedicated config class for your Feign client:

import feign.Retryer;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;

@Configuration
public class FeignClientConfig {
    @Bean
    public Retryer feignRetryer() {
        // Disable all retries
        return Retryer.NEVER_RETRY;
    }
}

Reference this config in your Feign client annotation:

@FeignClient(name = "XYZ", configuration = FeignClientConfig.class)
public interface YourFeignClient {
    @RequestMapping(path="/catalogs/getCatalogList",method = RequestMethod.GET)
    ResponseEntity<RestResponse> getCatalogList() throws RestException;
}

Option 2: Configure via Properties

Add these lines to your application.properties or application.yml:

# Disable retries for all Feign clients
feign.client.config.default.retryer=feign.Retryer.NEVER_RETRY

# Or disable only for the XYZ service client
feign.client.config.XYZ.retryer=feign.Retryer.NEVER_RETRY

2. Ribbon Retry Policies #

If you’re using Ribbon (common with Eureka for client-side load balancing), it has its own set of retry rules that might be kicking in for slow responses.

Fix: Adjust Ribbon Retry Settings #

Override these properties to limit or disable retries:

# No retries on the same server instance
ribbon.MaxAutoRetries=0
# No retries on alternative server instances
ribbon.MaxAutoRetriesNextServer=0
# Don’t retry all operation types (stick to safe ones like GET only, or disable entirely)
ribbon.OkToRetryOnAllOperations=false

3. Timeout Mismatch #

If your client’s read timeout is shorter than Microservice Y’s actual response time, the client will interpret the delay as a failure and trigger retries.

Fix: Increase Timeout Values #

Set timeout limits that match or exceed Y’s expected response time:

Feign Timeout Configuration

# For all Feign clients
feign.client.config.default.connectTimeout=5000 # 5 seconds for connection
feign.client.config.default.readTimeout=30000 # 30 seconds for response

# For the XYZ client specifically
feign.client.config.XYZ.connectTimeout=5000
feign.client.config.XYZ.readTimeout=30000

Ribbon Timeout Configuration

ribbon.ConnectTimeout=5000
ribbon.ReadTimeout=30000

4. Check for Custom Retry Logic #

Double-check your code in Microservice X to ensure there’s no manual retry logic (like Spring Retry annotations or custom loops) that could be causing repeated calls.

Final Testing #

After applying these changes, test the scenario again with a slow response from Microservice Y. You should now see only one call being made instead of multiple retries.

内容的提问来源于stack exchange，提问作者gaurav singh