问题解决过程

线上一个界面发现老是没有数据，排查下来时生产者没有成功发送消息所致，报错如下：

org.springframework.kafka.core.KafkaProducerException: Failed to send; nested exception is org.apache.kafka.common.errors.TimeoutException: Failed to update metadata after 0 ms.

我们发现配置里有一个很奇怪的参数

    config.put(ProducerConfig.MAX_BLOCK_MS_CONFIG, 0);

我们找到相关源码：

private Future<RecordMetadata> doSend(ProducerRecord<K, V> record, Callback callback) {
        TopicPartition tp = null;
        try {
            throwIfProducerClosed();
            // first make sure the metadata for the topic is available
            ClusterAndWaitTime clusterAndWaitTime;
            try {
            // -------------------------maxBlockTimeMs最终会传入ProducerConfig.MAX_BLOCK_MS_CONFIG对应值0
                clusterAndWaitTime = waitOnMetadata(record.topic(), record.partition(), maxBlockTimeMs);
        .......  

private ClusterAndWaitTime waitOnMetadata(String topic, Integer partition, long maxWaitMs) throws InterruptedException {
       ....
        do {
            log.trace("Requesting metadata update for topic {}.", topic);
            metadata.add(topic);
            int version = metadata.requestUpdate();
            sender.wakeup();
            try {
                metadata.awaitUpdate(version, remainingWaitMs);
            } catch (TimeoutException ex) {
            // ------------------主要是在这里报错了------------
                // Rethrow with original maxWaitMs to prevent logging exception with remainingWaitMs
                throw new TimeoutException("Failed to update metadata after " + maxWaitMs + " ms.");
            }
            ....

我们找到max.block.ms的作用

public static final String MAX_BLOCK_MS_CONFIG = "max.block.ms";
    private static final String MAX_BLOCK_MS_DOC = "The configuration controls how long <code>KafkaProducer.send()</code> and <code>KafkaProducer.partitionsFor()</code> will block."
                                                    + "These methods can be blocked either because the buffer is full or metadata unavailable."
                                                    + "Blocking in the user-supplied serializers or partitioner will not be counted against this timeout.";

问题很清晰了, send时发现要更新metadata，更新metadata时0ms超时报错了！

故把该值改为1000即可

    config.put(ProducerConfig.MAX_BLOCK_MS_CONFIG, 1000);

深入思考这个问题

这个功能是由界面点击时触发的，但运营人员很少会去使用这个功能，故猜测太长时间没用，比如超过9分钟（connections.max.idle.ms), 导致连接关闭，下次send时需要重新拉取metadata导致，但实际是这样的吗？no

我们把connections.max.idle.ms=1000，期待1s后send失败，但实际上还是需要空闲好几分钟后send才会失败，说明connections.max.idle.ms这个参数不是报错的直接原因！

通过DEBUG日志, 我们发现超时连接断开时，会触发metadata的更新；当metadata.max.age.ms超时时也会触发metadate更新。我们再深入研究下去

调试时发现抛出报错的日志前面，有如下所示日志

[2021-08-25 22:28:15,039] [kafka-producer-network-thread | producer-1] [DEBUG] org.apache.kafka.clients.Metadata - Removing unused topic InnerProcess from the metadata list, expiryMs 1629901694499 now 1629901695039

这句话是在下面这个地方打印的

org.apache.kafka.clients.Metadata#update

public static final long TOPIC_EXPIRY_MS = 5 * 60 * 1000;

public synchronized void update(Cluster newCluster, Set<String> unavailableTopics, long now) {
       ....

// producer默认为true，consumer默认为false
        if (topicExpiryEnabled) {
            // Handle expiry of topics from the metadata refresh set.
            for (Iterator<Map.Entry<String, Long>> it = topics.entrySet().iterator(); it.hasNext(); ) {
                Map.Entry<String, Long> entry = it.next();
                long expireMs = entry.getValue();
                // 第一次需要立即更新为-1
                if (expireMs == TOPIC_EXPIRY_NEEDS_UPDATE)
                // 第一次设置了5分钟后失效
                    entry.setValue(now + TOPIC_EXPIRY_MS);
                else if (expireMs <= now) {
                // 达到失效时间后，topic就会删除，下次send需要重新更新metadata了
                    it.remove();
                    log.debug("Removing unused topic {} from the metadata list, expiryMs {} now {}", entry.getKey(), expireMs, now);
                }
            }
        }
        .....

我们可以看到topic其实是设置了失效时间的，默认5分钟，当第一次调用send()，失效时间为-1，并且会触发更新metadata，即会修改为entry.setValue(now + TOPIC_EXPIRY_MS)，5分钟后失效，但5分钟后不会立即失效，若等到9分钟连接断开时，会走到上面的代码：

	 if (expireMs <= now) {
                // 达到失效时间后，topic就会删除，下次send需要重新更新metadata了
                    it.remove();
                    log.debug("Removing unused topic {} from the metadata list, expiryMs {} now {}", entry.getKey(), expireMs, now);
                }

即会将topic移除了，故9分钟后再次send时，需要更新metadata，但max.block.ms=0导致超时报错，即发送失败！

我们可以做这样的试验, 修改下面2个参数

        // 在topic失效之前停掉
        config.put(ProducerConfig.CONNECTIONS_MAX_IDLE_MS_CONFIG, 4 * 60 * 1000);
        // 10小时才更新一次metadata
        config.put(ProducerConfig.METADATA_MAX_AGE_CONFIG, 10 * 60 * 60 * 1000);

然后每大于5分钟的时间间隔，send一次

@Scheduled(fixedDelay = 9 * 60 * 1000 + 10000)
    public void sendKafkaASync() {
        kafkaService.sendKafkaASync();
    }

发现除了第一次失败后，就不会失败了！！！

因此，试验下来，是由于max.block.ms设置过小，导致更新metadata超时，导致send失败；更深入的：
1）首次发送没有topic数据，send调用时需要首先更新metadata时间不够，导致超时；
2）后面发送失败，是因为topic已超时，并在之后，connections.max.idle.ms或metadata.max.age.ms到时间触发metadata更新，而把超时的topic移除了。topic移除后，再进行send，有需要先更新metadata，进而还是会失败；

总体上而言，send的间隔不要超过5分钟最好，超过5分钟不调用send, 存在一定的可能导致topic被移除，从而下次send需要重新拉去metadata

附录：相关代码

pom.xml

<parent>
        <artifactId>spring-boot-starter-parent</artifactId>
        <groupId>org.springframework.boot</groupId>
        <version>2.1.5.RELEASE</version>
    </parent>

    <dependencies>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
            <exclusions>
                <!-- logback to log4j2 -->
                <exclusion>
                    <groupId>org.springframework.boot</groupId>
                    <artifactId>spring-boot-starter-logging</artifactId>
                </exclusion>
            </exclusions>
        </dependency>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-log4j2</artifactId>
        </dependency>
        <dependency>
            <groupId>org.springframework.kafka</groupId>
            <artifactId>spring-kafka</artifactId>
            <version>2.2.5.RELEASE</version>
        </dependency>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-test</artifactId>
        </dependency>

        <dependency>
            <groupId>org.projectlombok</groupId>
            <artifactId>lombok</artifactId>
            <version>1.18.12</version>
        </dependency>

    </dependencies>

log4j2.xml

<?xml version="1.0" encoding="UTF-8"?>
<!-- status指定log4j本身的日志打印级别 -->
<!-- 配置文件发生修改，在monitorInterval时间即5s内，重新加载，无需重启 -->
<Configuration status="DEBUG" monitorInterval="5">
        .....

    <Loggers>
        <Logger name="org.springframework" level="INFO" />
        <Logger name="com.ydfind" level="DEBUG" />
        <Logger name="org.apache.kafka.clients.consumer" level="WARN"/>
        <Logger name="org.apache.kafka.clients.FetchSessionHandler" level="WARN" />
        <Logger name="org.apache.kafka.clients.producer.KafkaProducer" level="TRACE" />
    </Loggers>
</Configuration>

application.yml

logging:
  config: classpath:log4j2.xml

KafkaConfig.java

package com.ydfind.config;

import org.apache.kafka.clients.CommonClientConfigs;
import org.apache.kafka.clients.consumer.ConsumerConfig;
import org.apache.kafka.clients.producer.ProducerConfig;
import org.apache.kafka.common.config.SaslConfigs;
import org.apache.kafka.common.security.auth.SecurityProtocol;
import org.apache.kafka.common.security.scram.ScramLoginModule;
import org.apache.kafka.common.serialization.StringDeserializer;
import org.apache.kafka.common.serialization.StringSerializer;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import org.springframework.kafka.annotation.EnableKafka;
import org.springframework.kafka.config.ConcurrentKafkaListenerContainerFactory;
import org.springframework.kafka.config.KafkaListenerContainerFactory;
import org.springframework.kafka.core.DefaultKafkaConsumerFactory;
import org.springframework.kafka.core.DefaultKafkaProducerFactory;
import org.springframework.kafka.core.KafkaTemplate;
import org.springframework.kafka.core.ProducerFactory;
import org.springframework.kafka.listener.ConcurrentMessageListenerContainer;
import org.springframework.web.context.request.FacesRequestAttributes;

import java.awt.datatransfer.StringSelection;
import java.util.HashMap;
import java.util.Map;

@Configuration
@EnableKafka
public class KafkaConfig {

    @Bean
    public KafkaTemplate<String, String> kafkaTemplate() {
        return new KafkaTemplate<>(producerFactory());
    }

    private ProducerFactory<String, String> producerFactory() {
        return new DefaultKafkaProducerFactory<>(producerConfigs());
    }

    private Map<String, Object> producerConfigs() {
        Map<String, Object> config = new HashMap<>();
        config.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        config.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
        config.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
//
//        config.put(CommonClientConfigs.SECURITY_PROTOCOL_CONFIG, SecurityProtocol.SASL_PLAINTEXT.name);
//        config.put(SaslConfigs.SASL_JAAS_CONFIG, "org.apache.kafka.common.security.scram.ScramLoginModule required username=\"xxx\" password=\"xxx\"");
//        config.put(SaslConfigs.SASL_MECHANISM, "SCRAM-SHA-256");

        config.put(ProducerConfig.MAX_BLOCK_MS_CONFIG, 0);
//        config.put(ProducerConfig.CONNECTIONS_MAX_IDLE_MS_CONFIG, 1 * 60 * 1000);
//        config.put(ProducerConfig.METADATA_MAX_AGE_CONFIG, 30 * 60 * 1000);

//        config.put(ProducerConfig.CONNECTIONS_MAX_IDLE_MS_CONFIG, 100 * 60 * 1000);
//        config.put(ProducerConfig.METADATA_MAX_AGE_CONFIG, 1 * 60 * 1000);


        // ----貌似下面会10小时内都成功！！！！---------------------------
        // 4min时connect断掉，5min时topic失效，6min时send()成功，10min时把topic失效时间弄从15min！！！！！
        // 是send()导致entry.getValue() == TOPIC_EXPIRY_NEEDS_UPDATE，还是connect断开的更新topic导致呢！！！
        // topics.put(topic, TOPIC_EXPIRY_NEEDS_UPDATE)-------------send时会把topic的失效时间弄成-1----相当于续命，不会失效

        // 在topic失效之前停掉
        config.put(ProducerConfig.CONNECTIONS_MAX_IDLE_MS_CONFIG, 4 * 60 * 1000);
        // 10小时才更新一次metadata
        config.put(ProducerConfig.METADATA_MAX_AGE_CONFIG, 10 * 60 * 60 * 1000);
        return config;
    }
}

KafkaService

package com.ydfind.service;

import lombok.extern.slf4j.Slf4j;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.kafka.core.KafkaTemplate;
import org.springframework.kafka.support.SendResult;
import org.springframework.stereotype.Component;
import org.springframework.util.concurrent.ListenableFuture;

import java.util.concurrent.ExecutionException;
import java.util.concurrent.atomic.AtomicInteger;

@Component
@Slf4j
public class KafkaService {

    @Autowired
    private KafkaTemplate<String, String> kafkaTemplate;

    public String sendKafkaSync() {
        ListenableFuture<SendResult<String, String>> send = kafkaTemplate.send("InnerProcess", "sendKafkaSync");
        String msg = "";
        try {
            SendResult<String, String> stringStringSendResult = send.get();
            msg += stringStringSendResult.getRecordMetadata().offset();
        } catch (InterruptedException e) {
            e.printStackTrace();
            return e.getMessage();
        } catch (ExecutionException e) {
            e.printStackTrace();
            return e.getMessage();
        }
        return msg;
    }

    private AtomicInteger total = new AtomicInteger();
    private AtomicInteger success = new AtomicInteger();

    public void sendKafkaASync() {
        ListenableFuture<SendResult<String, String>> send = kafkaTemplate.send("InnerProcess", "sendKafkaSync");
        send.completable().whenCompleteAsync((n, e) -> {
            total.getAndIncrement();
            if (null != e) {
                e.printStackTrace();
            } else {
                success.getAndIncrement();
            }
            log.info("\n-------------------total = {}, success = {}", total.get(), success.get());
        });
    }
}

MainApp.java

package com.ydfind;


import com.ydfind.service.KafkaService;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.scheduling.annotation.EnableScheduling;
import org.springframework.scheduling.annotation.Scheduled;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RestController;

@SpringBootApplication
@RestController
@EnableScheduling
public class MainApp {

    public static void main(String[] args) {
        SpringApplication.run(MainApp.class, args);
    }

    @Autowired
    private KafkaService kafkaService;

    @GetMapping("sendKafka")
    public String sendKafka() {
        return kafkaService.sendKafkaSync();
    }

    @Scheduled(fixedDelay = 9 * 60 * 1000 + 10000)
    public void sendKafkaASync() {
        kafkaService.sendKafkaASync();
    }
}
# kafka # java
Logo
华为云开发者联盟

为开发者提供学习成长、分享交流、生态实践、资源工具等服务，帮助开发者快速成长。

更多推荐

cover

通过HPA+CronHPA组合应对业务复杂弹性伸缩场景

avatar 华为云开发者联盟
cover

FT-FMEA融合混沌演练，零售运营系统韧性架构在线验证实践

avatar 华为云开发者联盟
cover

如何使用Python和Plotly绘制3D图形

avatar 华为云开发者联盟