【Azure 事件中心】azure-spring-cloud-stream-binder-eventhubs客戶端組件問題, 實踐消息非順序可達

問題描述

查閱了Azure的官方文檔( 將事件發送到特定分區: https://docs.azure.cn/zh-cn/event-hubs/event-hubs-availability-and-consistency?tabs=java#send-events-to-a-specific-partition)，在工程里引用組件“azure-spring-cloud-stream-binder-eventhubs”來連接EventHub發送和消費消息事件。在發送端一個For循環中發送帶順序號的消息，編號從0開始，並且在消息的header中指定了 "Partition Key"，相同PartitionKey的消息會被發送到相同的Partition，來保證這些消息的順序。

但是在消費端的工程中消費這些消息時，看到打印到日志中的結果並不是從0遞增的。所以想知道是發送端在發送時就已經亂序發送了？還是消息到達EventHub后亂序保存了？還是消費端的消費方式的問題，導致打印出的結果是亂序的？

 

下面是發送端的代碼：

public void testPushMessages(int mcount, String partitionKey) {
String message = "Message ";
for (int i=0; i <mcount; i++) {
source.output().send(MessageBuilder.withPayload(partitionKey + mcount + i).setHeaderIfAbsent(AzureHeaders.PARTITION_KEY,partitionKey).build());
    }
}

下面是消費端代碼：

@StreamListener(Sink.INPUT)
public void onEvent(String message, @Header(AzureHeaders.CHECKPOINTER) Checkpointer checkpointer,
                    @Header(AzureHeaders.RAW_PARTITION_ID) String rawPartitionId,
                    @Header(AzureHeaders.PARTITION_KEY) String partitionKey) {
    checkpointer.success()
            .doOnSuccess(s -> log.info("Message '{}' successfully check pointed.rawPartitionId={},partitionKey={}", message, rawPartitionId, partitionKey))
            .doOnError(s -> log.error("Checkpoint message got exception."))
            .subscribe();

下面是打印的日志

......,"data":"Message 'testKey4testMessage1' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage29' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage27' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage26' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage25' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage28' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage14' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage13' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage15' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage5' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage7' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage20' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage19' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage18' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage0' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage9' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage12' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}
......,"data":"Message 'testKey5testMessage8' successfully check pointed.rawPartitionId=1,partition<*****>","xcptn":""}

從日志中可以看到，消息確實都被發送到了同一個分區（rawPartitionId=1），但是從消息體的序號上看，是亂序的

 

問題分析

這個是和這個配置相關的fixedDelay，指定默認輪詢器的固定延遲，是一個周期性觸發器，之前代碼會根據這個輪詢器進行發送和接受消息的。使用Send發送的方法，現在最新的SDK 不使用這個方法，所以需要使用新的sdk 發送數據測試一下。

 

新sdk 參考文檔您可以參考一下：https://github.com/Azure/azure-sdk-for-java/tree/master/sdk/spring/azure-spring-boot-samples/azure-spring-cloud-sample-eventhubs-binder

SDK版本為

<dependency>
    <groupId>com.azure.spring</groupId>
    <artifactId>azure-spring-cloud-stream-binder-eventhubs</artifactId>
    <version>2.4.0</version>
</dependency>

在參考官網的示例后，使用Supplier方法發送消息，代替Send。經過多次測試，指定partitionkey 之后，發送消息是順序發送的，消費的時候也是按照順序消費的，下面是測試的代碼和結果

發送端的代碼

// Copyright (c) Microsoft Corporation. All rights reserved.
// Licensed under the MIT License.

package com.azure.spring.sample.eventhubs.binder;

import com.azure.spring.integration.core.EventHubHeaders;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import org.springframework.context.annotation.Profile;
import org.springframework.messaging.Message;
import org.springframework.messaging.support.MessageBuilder;

import java.util.function.Supplier;

import static com.azure.spring.integration.core.EventHubHeaders.SEQUENCE_NUMBER;

@Configuration
public class EventProducerConfiguration {

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

    private int i = 0;

    @Bean
    public Supplier<Message<String>> supply() {
        return () -> {
            //LOGGER.info("Sending message, sequence " + i);
            String partitionKey="info";

            LOGGER.info("Send message " + MessageBuilder.withPayload("hello world, "+i).setHeaderIfAbsent(EventHubHeaders.PARTITION_KEY, partitionKey).build());
            return MessageBuilder.withPayload("hello world, "+ i++).
                    setHeaderIfAbsent(EventHubHeaders.PARTITION_KEY, partitionKey).build();

        };
    }

}

接收端的代碼

package com.ywt.demoEventhub;

import com.azure.spring.integration.core.EventHubHeaders;
import com.azure.spring.integration.core.api.reactor.Checkpointer;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;
import org.springframework.integration.annotation.ServiceActivator;
import org.springframework.messaging.Message;

import java.util.function.Consumer;

import static com.azure.spring.integration.core.AzureHeaders.CHECKPOINTER;

@Configuration
public class EventConsume {

    private static final Logger LOGGER = LoggerFactory.getLogger(EventConsume.class);
    @Bean
    public Consumer<Message<String>> consume() {
        return message -> {
            Checkpointer checkpointer = (Checkpointer) message.getHeaders().get(CHECKPOINTER);
            LOGGER.info("New message received: '{}', partition key: {}, sequence number: {}, offset: {}, enqueued time: {}",
                    message.getPayload(),
                    message.getHeaders().get(EventHubHeaders.PARTITION_KEY),
                    message.getHeaders().get(EventHubHeaders.SEQUENCE_NUMBER),
                    message.getHeaders().get(EventHubHeaders.OFFSET),
                    message.getHeaders().get(EventHubHeaders.ENQUEUED_TIME)
            );

            checkpointer.success()
                    .doOnSuccess(success -> LOGGER.info("Message '{}' successfully checkpointed number '{}' ", message.getPayload(), message.getHeaders().get(EventHubHeaders.CHECKPOINTER)))
                    .doOnError(error -> LOGGER.error("Exception found", error))
                    .subscribe();
        };
    }
}

發送消息的日志

消費消息的日志

 

 

 

參考資料

Azure Spring Cloud Stream Binder for Event Hub Code Sample shared library for Java：https://github.com/Azure/azure-sdk-for-java/tree/master/sdk/spring/azure-spring-boot-samples/azure-spring-cloud-sample-eventhubs-binder

How to create a Spring Cloud Stream Binder application with Azure Event Hubs - Add sample code to implement basic event hub functionality : https://docs.microsoft.com/en-us/azure/developer/java/spring-framework/configure-spring-cloud-stream-binder-java-app-azure-event-hub#add-sample-code-to-implement-basic-event-hub-functionality

 

[END]