dubbo实战篇：dubbo超时重试

dubbo超时重试概述

consumer端和provider端都可以设置timeout。

超时优先级：consumer方法级>provider方法级>consumer接口级>provider接口级>consumer级>provider级

 

当consumer端调用超时，会触发重试调用。

重试对应的配置属性是retries。默认的重试次数是2。就是说，当调用超时，会最多重试2次，如果仍然失败，会提示异常。

 

对于查询或删除来说，接口重试是幂等的。

对于新增数据，如果retries>0，则要做幂等处理，否则会造成重复数据入库而产生bug。安全起见，可单独设置retries=0。

 

【说明】在直连的情况下，是不会触发重试的。

 

代码

§1. 服务接口定义：

package dubbodemo.contract;

public interface HelloWord {
    String say();

    String add(String str);
}

 

§2. 服务提供方

Provider application.yml dubbo配置timeout和retries：

dubbo:
  application:
    name: zhanggz-dubbodemo
  registry:
#    address: N/A
    address: zookeeper://127.0.0.1:2181
  protocol:
    port: 28088
    name: dubbo
  scan:
    base-packages: dubbodemo.provider
  provider:
    timeout: 2200
    retries: 3

Provider代码：

package dubbodemo.provider;

import dubbodemo.contract.HelloWord;
import lombok.extern.slf4j.Slf4j;
import org.apache.commons.lang3.RandomUtils;
import org.apache.dubbo.config.annotation.Service;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;

@Slf4j
@Service(retries = 3)
@SpringBootApplication
public class ProviderApplication implements HelloWord {
    public static void main(String[] args) {
        SpringApplication.run(ProviderApplication.class);
    }

    public String add(String str) {
        log.info("add入参:{}", str);
        long start = System.currentTimeMillis();
        try {
//int i=1/0;
            try {
                int r = RandomUtils.nextInt(1000, 2000);
                log.info("sleep time={}", r);
                Thread.sleep(r);
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
            return "test retry:" + str;
        } finally {
            log.info("duration={}", System.currentTimeMillis() - start);
        }
    }

    public String say() {
        return ("hello");
    }
}

 

§3. 服务消费方

Consumer application.yml dubbo配置timeout和retries：

dubbo:
  application:
    name: zhanggz-dubbodemo-consumer
  registry:
    address: zookeeper://127.0.0.1:2181
  #    address: N/A
  protocol:
    name: dubbo
  consumer:
    timeout: 1000
#    retries: 0

Consumer代码：

package dubbodemo.consumer;

import dubbodemo.contract.HelloWord;
import lombok.extern.slf4j.Slf4j;
import org.apache.dubbo.config.annotation.Reference;
import org.junit.Test;
import org.junit.runner.RunWith;
import org.springframework.boot.test.context.SpringBootTest;
import org.springframework.test.context.junit4.SpringRunner;

@Slf4j
@RunWith(SpringRunner.class)
@SpringBootTest
public class DubboConsumerTest {
    @Reference(/*url = "dubbo://localhost:28088",*/)
    private HelloWord helloWord;

    @Test
    public void testAddRetry() {
        long start = System.currentTimeMillis();
        log.info("请求开始...");
        try {
            String str = helloWord.add("str");
            log.info("返回值={}", str);
        } finally {
            log.info("duration=={}", System.currentTimeMillis() - start);
        }
    }
}

 

测试结论

--------直连的情况下，测试dubbo调用超时。

将Provider方法利用Thread.sleep(2000)来模拟超时。

consumer不设置timeout（默认值=1000ms），改变provider的timeout值

§1.【provider】设置provider.timeout=1000 ,因为方法执行耗时是2000ms，会有WARN提示。

2020-07-23 18:53:57.427 INFO 13356 --- [:20880-thread-2] dubbodemo.provider.ProviderApplication : duration=2000
2020-07-23 18:53:57.432 WARN 13356 --- [:20880-thread-2] o.apache.dubbo.rpc.filter.TimeoutFilter : [DUBBO] invoke time out. 
method: add arguments: [str] , 
url is dubbo://192.168.40.69:20880/dubbodemo.contract.HelloWord?anyhost=true&application=zhanggz-dubbodemo&bean.name=ServiceBean:dubbodemo.contract.HelloWord&bind.ip=192.168.40.69&bind.port=20880&deprecated=false&dubbo=2.0.2&dynamic=true&generic=false&interface=dubbodemo.contract.HelloWord&methods=add,say&pid=13356&qos.enable=false&register=true&release=2.7.3&side=provider&timeout=1000&timestamp=1595501612605, 
invoke elapsed 2005 ms., dubbo version: 2.7.3, current host: 192.168.40.69

 

§2. 【provider】设置provider.timeout=2200, 方法耗时2000<2200，不会提示invoke timeout。

2020-07-23 18:57:16.949 INFO 12280 --- [:20880-thread-2] dubbodemo.provider.ProviderApplication : duration=2000

 

§3. 不管Provider如何设置timeout，因为Consumer的timeout=1000ms，所以consumer都会收到timeout异常。

2020-07-23 18:53:56.273 INFO 13216 --- [ main] dubbodemo.consumer.DubboConsumerTest : duration==1040

org.apache.dubbo.rpc.RpcException: Invoke remote method timeout. 
method: add, 
provider: dubbo://localhost:20880/dubbodemo.contract.HelloWord?application=zhanggz-dubbodemo-consumer&interface=dubbodemo.contract.HelloWord&lazy=false&pid=13216&qos.enable=false&register.ip=192.168.40.69&remote.application=&side=consumer&sticky=false, 
cause: org.apache.dubbo.remoting.TimeoutException: Waiting server-side response timeout by scan timer. start time: 2020-07-23 18:53:55.250, end time: 2020-07-23 18:53:56.272, client elapsed: 106 ms, server elapsed: 916 ms, timeout: 1000 ms, 
request: Request [id=0, version=2.0.2, twoway=true, event=false, broken=false, data=RpcInvocation [methodName=add, parameterTypes=[class java.lang.String], arguments=[str], attachments={path=dubbodemo.contract.HelloWord, interface=dubbodemo.contract.HelloWord, version=0.0.0}]], 
channel: /192.168.40.69:1813 -> /192.168.40.69:20880

 

 

§1. Provider服务停止后，consumer端无法启动

org.springframework.beans.factory.BeanCreationException: Error creating bean with name 'dubbodemo.consumer.DubboConsumerTest': 
Injection of @Reference dependencies is failed; 
nested exception is org.apache.dubbo.rpc.RpcException: Fail to create remoting client for service(
dubbo://localhost:20880/dubbodemo.contract.HelloWord?application=zhanggz-dubbodemo-consumer&codec=dubbo&heartbeat=60000&interface=dubbodemo.contract.HelloWord&lazy=false&pid=5236&qos.enable=false&register.ip=192.168.40.69&remote.application=&side=consumer&sticky=false&timeout=2200
): client(url: 
dubbo://localhost:20880/dubbodemo.contract.HelloWord?application=zhanggz-dubbodemo-consumer&codec=dubbo&heartbeat=60000&interface=dubbodemo.contract.HelloWord&lazy=false&pid=5236&qos.enable=false&register.ip=192.168.40.69&remote.application=&side=consumer&sticky=false&timeout=2200
) failed to connect to server localhost/127.0.0.1:20880, error message is:Connection refused: no further information: /192.168.40.69:20880

 

 

§2. Provider服务停止后，consumer端设置dubbo.consumer.check=false。则可以启动consumer服务。

check属性说明：Check if service provider exists, if not exists, it will be fast fail

不过，当调用远程服务的时候，由于远程服务不可用，所以当然会报错。

• 错误提示（连接了注册中心的场景）：

org.apache.dubbo.rpc.RpcException: No provider available from registry 192.168.40.84:2181 
for service dubbodemo.contract.HelloWord on consumer 192.168.40.69 use dubbo version 2.7.3, 
please check status of providers(disabled, not registered or in blacklist).

• 错误提示（直连的场景）：

org.apache.dubbo.rpc.RpcException: Failed to invoke remote method: add, 
provider: dubbo://localhost:20880/dubbodemo.contract.HelloWord?application=zhanggz-dubbodemo-consumer&check=false&interface=dubbodemo.contract.HelloWord&lazy=false&pid=12504&qos.enable=false&register.ip=192.168.40.69&remote.application=&side=consumer&sticky=false&timeout=2200, 
cause: message can not send, because channel is closed . 
url:dubbo://localhost:20880/dubbodemo.contract.HelloWord?application=zhanggz-dubbodemo-consumer&check=false&codec=dubbo&heartbeat=60000&interface=dubbodemo.contract.HelloWord&lazy=false&pid=12504&qos.enable=false&register.ip=192.168.40.69&remote.application=&side=consumer&sticky=false&timeout=2200

 

----------------------------------

consumer端，设置retries=3

2020-07-24 17:12:57.389 INFO 7468 --- [ main] dubbodemo.consumer.DubboConsumerTest : 请求开始...
2020-07-24 17:13:00.202 INFO 7468 --- [ main] dubbodemo.consumer.DubboConsumerTest : 返回值=test retry:str
2020-07-24 17:13:00.202 INFO 7468 --- [ main] dubbodemo.consumer.DubboConsumerTest : duration==2813

 

provider端：最多会被执行1+retries=4次

2020-07-24 17:12:57.540 INFO 2372 --- [28088-thread-17] dubbodemo.provider.ProviderApplication : add入参:str
2020-07-24 17:12:57.541 INFO 2372 --- [28088-thread-17] dubbodemo.provider.ProviderApplication : r=1169
2020-07-24 17:12:58.444 INFO 2372 --- [28088-thread-18] dubbodemo.provider.ProviderApplication : add入参:str
2020-07-24 17:12:58.444 INFO 2372 --- [28088-thread-18] dubbodemo.provider.ProviderApplication : r=1934
2020-07-24 17:12:58.711 INFO 2372 --- [28088-thread-17] dubbodemo.provider.ProviderApplication : duration=1170
2020-07-24 17:12:59.461 INFO 2372 --- [28088-thread-19] dubbodemo.provider.ProviderApplication : add入参:str
2020-07-24 17:12:59.461 INFO 2372 --- [28088-thread-19] dubbodemo.provider.ProviderApplication : r=736
2020-07-24 17:13:00.198 INFO 2372 --- [28088-thread-19] dubbodemo.provider.ProviderApplication : duration=737
2020-07-24 17:13:00.379 INFO 2372 --- [28088-thread-18] dubbodemo.provider.ProviderApplication : duration=1935

 

优先级测试

§1. Provider设置retries=1 consumer设置retries=3，服务端会重试1次。 

注意，log里的Tried 2 times 指的是一共调用了2次。即首次+重试的那1次。

org.apache.dubbo.rpc.RpcException: Failed to invoke the method add in the service dubbodemo.contract.HelloWord. 
Tried 2 times of the providers [192.168.40.84:28088, 192.168.40.69:28088] (2/2) from the registry 192.168.40.84:2181 on the consumer 192.168.40.69 using the dubbo version 2.7.3. 
Last error is: Invoke remote method timeout. method: add, 
provider: dubbo://192.168.40.84:28088/dubbodemo.contract.HelloWord?anyhost=true&application=zhanggz-dubbodemo-consumer&bean.name=ServiceBean:dubbodemo.contract.HelloWord&check=false&deprecated=false&dubbo=2.0.2&dynamic=true&generic=false&interface=dubbodemo.contract.HelloWord&lazy=false&methods=add,say&pid=23324&qos.enable=false&register=true&register.ip=192.168.40.69&release=2.7.3&remote.application=zhanggz-dubbodemo&retries=1&revision=1.0-SNAPSHOT&side=consumer&sticky=false&timeout=1000&timestamp=1596179853357, 
cause: org.apache.dubbo.remoting.TimeoutException: Waiting server-side response timeout by scan timer. start time: 2020-07-31 15:33:52.809, end time: 2020-07-31 15:33:53.826, client elapsed: 1 ms, server elapsed: 1016 ms, timeout: 1000 ms, 
request: Request [id=1, version=2.0.2, twoway=true, event=false, broken=false, data=RpcInvocation [methodName=add, parameterTypes=[class java.lang.String], arguments=[str], attachments={path=dubbodemo.contract.HelloWord, interface=dubbodemo.contract.HelloWord, version=0.0.0, timeout=1000}]], 
channel: /192.168.40.69:3666 -> /192.168.40.84:28088

 

§2. Provider设置retries=0 consumer设置retries=3，服务端不会重试。

org.apache.dubbo.rpc.RpcException: Failed to invoke the method add in the service dubbodemo.contract.HelloWord. 
Tried 1 times of the providers [192.168.40.69:28088] (1/1) from the registry 127.0.0.1:2181 on the consumer 192.168.40.69 using the dubbo version 2.7.3. 
Last error is: Invoke remote method timeout. method: add, 
provider: dubbo://192.168.40.69:28088/dubbodemo.contract.HelloWord?anyhost=true&application=zhanggz-dubbodemo-consumer&bean.name=ServiceBean:dubbodemo.contract.HelloWord&check=false&deprecated=false&dubbo=2.0.2&dynamic=true&generic=false&interface=dubbodemo.contract.HelloWord&lazy=false&methods=add,say&pid=9436&qos.enable=false&register=true&register.ip=192.168.40.69&release=2.7.3&remote.application=zhanggz-dubbodemo&retries=0&side=consumer&sticky=false&timeout=1000&timestamp=1596170589487, 
cause: org.apache.dubbo.remoting.TimeoutException: Waiting server-side response timeout by scan timer. start time: 2020-07-31 12:43:36.395, end time: 2020-07-31 12:43:37.419, client elapsed: 111 ms, server elapsed: 912 ms, timeout: 1000 ms, 
request: Request [id=0, version=2.0.2, twoway=true, event=false, broken=false, data=RpcInvocation [methodName=add, parameterTypes=[class java.lang.String], arguments=[str], attachments={path=dubbodemo.contract.HelloWord, interface=dubbodemo.contract.HelloWord, version=0.0.0, timeout=1000}]], 
channel: /192.168.40.69:2295 -> /192.168.40.69:28088