c#开源消息队列中间件EQueue 教程

一、简介

EQueue是一个参照RocketMQ实现的开源消息队列中间件,兼容Mono，具体可以参看作者的文章《分享一个c#写的开源分布式消息队列equeue》。项目开源地址：https://github.com/tangxuehua/equeue，项目中包含了队列的全部源代码以及如何使用的示例。

二、安装EQueue

Producer、Consumer、Broker支持分布式部署，安装EQueue需要.NET 4, Visual Studio 2010/2012/2013. 目前EQueue是个类库，需要自己实现Broker的宿主，可以参照QuickStart，创建一个QuickStart.BrokerServer项目，通过Visual Studio的Nuget 查找equeue

using System;
using System.Text;
using ECommon.Autofac;
using ECommon.Configurations;
using ECommon.JsonNet;
using ECommon.Log4Net;
using EQueue.Broker;
using EQueue.Configurations;
using EQueue.Protocols;

namespace QuickStart.BrokerServer
{
    class Program
    {
        static void Main(string[] args)
        {
            InitializeEQueue();
            var setting = new BrokerSetting();
            setting.NotifyWhenMessageArrived = false;
            setting.DeleteMessageInterval = 1000;
            new BrokerController(setting).Initialize().Start();
            Console.ReadLine();
        }

        static void InitializeEQueue()
        {
            Configuration
                .Create()
                .UseAutofac()
                .RegisterCommonComponents()
                .UseLog4Net()
                .UseJsonNet()
                .RegisterEQueueComponents();
        }
    }
}

InitializeEQueue方法初始化EQueue的环境，使用了Autofac作为IOC容器，使用log4Net记录日志, 我们看一下RegisterEQueueComponents方法：

   public static class ConfigurationExtensions
    {
        public static Configuration RegisterEQueueComponents(this Configuration configuration)
        {
            configuration.SetDefault<IAllocateMessageQueueStrategy, AverageAllocateMessageQueueStrategy>();
            configuration.SetDefault<IQueueSelector, QueueHashSelector>();
            configuration.SetDefault<ILocalOffsetStore, DefaultLocalOffsetStore>();
            configuration.SetDefault<IMessageStore, InMemoryMessageStore>();
            configuration.SetDefault<IMessageService, MessageService>();
            configuration.SetDefault<IOffsetManager, InMemoryOffsetManager>();
            return configuration;
        }
    }

代码中涉及到6个组件：

• IAllocateMessageQueueStrategy
• IQueueSelector
• ILocalOffsetStore
• IMessageStore
• IMessageService
• IOffsetManager

DeleteMessageInterval 这个属性是用来设置equeue的定时删除间隔，单位为毫秒,默认值是一个小时。另外还有ProducerSocketSetting 和 ConsumerSocketSetting 分别用于设置Producer连接Broker和Consumer连接Broker的IP和端口，默认端口是5000和5001。

 public class BrokerSetting
    {
        public SocketSetting ProducerSocketSetting { get; set; }
        public SocketSetting ConsumerSocketSetting { get; set; }
        public bool NotifyWhenMessageArrived { get; set; }
        public int DeleteMessageInterval { get; set; }

        public BrokerSetting()
        {
            ProducerSocketSetting = new SocketSetting { Address = SocketUtils.GetLocalIPV4().ToString(), Port = 5000, Backlog = 5000 };
            ConsumerSocketSetting = new SocketSetting { Address = SocketUtils.GetLocalIPV4().ToString(), Port = 5001, Backlog = 5000 };
            NotifyWhenMessageArrived = true;
            DeleteMessageInterval = 1000 * 60 * 60;
        }
    }

运行项目，如果显示下面类似内容，说明Broker启动成功：

2014-03-23 20:10:30,255  INFO BrokerController - Broker started, producer:[169.254.80.80:5000], consumer:[169.254.80.80:5001]

三、在Visual Studio中开发测试

1.创建一个VS项目 QuickStart.ProducerClient，通过Nuget引用EQueue，编写下面Producer代码

 using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.Text;
using System.Threading;
using System.Threading.Tasks;
using ECommon.Autofac;
using ECommon.Configurations;
using ECommon.IoC;
using ECommon.JsonNet;
using ECommon.Log4Net;
using ECommon.Scheduling;
using EQueue.Clients.Producers;
using EQueue.Configurations;
using EQueue.Protocols;

namespace QuickStart.ProducerClient
{
    class Program
    {
        static void Main(string[] args)
        {
            InitializeEQueue();

            var scheduleService = ObjectContainer.Resolve<IScheduleService>();
            var producer = new Producer().Start();
            var total = 1000;
            var parallelCount = 10;
            var finished = 0;
            var messageIndex = 0;
            var watch = Stopwatch.StartNew();

            var action = new Action(() =>
            {
                for (var index = 1; index <= total; index++)
                {
                    var message = "message" + Interlocked.Increment(ref messageIndex);
                    producer.SendAsync(new Message("SampleTopic", Encoding.UTF8.GetBytes(message)), index.ToString()).ContinueWith(sendTask =>
                    {
                        var finishedCount = Interlocked.Increment(ref finished);
                        if (finishedCount % 1000 == 0)
                        {
                            Console.WriteLine(string.Format("Sent {0} messages, time spent:{1}", finishedCount, watch.ElapsedMilliseconds));
                        }
                    });
                }
            });

            var actions = new List<Action>();
            for (var index = 0; index < parallelCount; index++)
            {
                actions.Add(action);
            }

            Parallel.Invoke(actions.ToArray());

            Console.ReadLine();
        }

        static void InitializeEQueue()
        {
            Configuration
                .Create()
                .UseAutofac()
                .RegisterCommonComponents()
                .UseLog4Net()
                .UseJsonNet()
                .RegisterEQueueComponents();
        }
    }
}

Producer对象在使用之前必须要调用Start初始化，初始化一次即可， 注意：切记不可以在每次发送消息时，都调用Start方法。Producer 默认连接本机的5000端口，可以通过ProducerSetting 进行设置，可以参看下面的代码：

 public class ProducerSetting
    {
        public string BrokerAddress { get; set; }
        public int BrokerPort { get; set; }
        public int SendMessageTimeoutMilliseconds { get; set; }
        public int UpdateTopicQueueCountInterval { get; set; }

        public ProducerSetting()
        {
            BrokerAddress = SocketUtils.GetLocalIPV4().ToString();
            BrokerPort = 5000;
            SendMessageTimeoutMilliseconds = 1000 * 10;
            UpdateTopicQueueCountInterval = 1000 * 5;
        }

2、创建一个VS项目 QuickStart.ConsumerClient，通过Nuget引用EQueue，编写下面Consumer代码

using System;
using System.Linq;
using System.Text;
using System.Threading;
using ECommon.Autofac;
using ECommon.Configurations;
using ECommon.IoC;
using ECommon.JsonNet;
using ECommon.Log4Net;
using ECommon.Scheduling;
using EQueue.Broker;
using EQueue.Clients.Consumers;
using EQueue.Configurations;
using EQueue.Protocols;

namespace QuickStart.ConsumerClient
{
    class Program
    {
        static void Main(string[] args)
        {
            InitializeEQueue();

            var messageHandler = new MessageHandler();
            var consumer1 = new Consumer("Consumer1", "group1").Subscribe("SampleTopic").Start(messageHandler);
            var consumer2 = new Consumer("Consumer2", "group1").Subscribe("SampleTopic").Start(messageHandler);
            var consumer3 = new Consumer("Consumer3", "group1").Subscribe("SampleTopic").Start(messageHandler);
            var consumer4 = new Consumer("Consumer4", "group1").Subscribe("SampleTopic").Start(messageHandler);

            Console.WriteLine("Start consumer load balance, please wait for a moment.");
            var scheduleService = ObjectContainer.Resolve<IScheduleService>();
            var waitHandle = new ManualResetEvent(false);
            var taskId = scheduleService.ScheduleTask(() =>
            {
                var c1AllocatedQueueIds = consumer1.GetCurrentQueues().Select(x => x.QueueId);
                var c2AllocatedQueueIds = consumer2.GetCurrentQueues().Select(x => x.QueueId);
                var c3AllocatedQueueIds = consumer3.GetCurrentQueues().Select(x => x.QueueId);
                var c4AllocatedQueueIds = consumer4.GetCurrentQueues().Select(x => x.QueueId);
                if (c1AllocatedQueueIds.Count() == 1 && c2AllocatedQueueIds.Count() == 1 && c3AllocatedQueueIds.Count() == 1 && c4AllocatedQueueIds.Count() == 1)
                {
                    Console.WriteLine(string.Format("Consumer load balance finished. Queue allocation result: c1:{0}, c2:{1}, c3:{2}, c4:{3}",
                        string.Join(",", c1AllocatedQueueIds),
                        string.Join(",", c2AllocatedQueueIds),
                        string.Join(",", c3AllocatedQueueIds),
                        string.Join(",", c4AllocatedQueueIds)));
                    waitHandle.Set();
                }
            }, 1000, 1000);

            waitHandle.WaitOne();
            scheduleService.ShutdownTask(taskId);

            Console.ReadLine();
        }

        static void InitializeEQueue()
        {
            Configuration
                .Create()
                .UseAutofac()
                .RegisterCommonComponents()
                .UseLog4Net()
                .UseJsonNet()
                .RegisterEQueueComponents();
        }
    }

    class MessageHandler : IMessageHandler
    {
        private int _handledCount;

        public void Handle(QueueMessage message, IMessageContext context)
        {
            var count = Interlocked.Increment(ref _handledCount);
            if (count % 1000 == 0)
            {
                Console.WriteLine("Total handled {0} messages.", count);
            }
            context.OnMessageHandled(message);
        }
    }
}

使用方式给用户感觉是消息从EQueue服务器推到了应用客户端。 但是实际Consumer内部是使用长轮询Pull方式从EQueue服务器拉消息，然后再回调用户Listener方法。Consumer默认连接本机的5001端口，可以通过ConsumerSetting 进行设置，可以参看下面的代码：

    public class ConsumerSetting
    {
        public string BrokerAddress { get; set; }
        public int BrokerPort { get; set; }
        public int RebalanceInterval { get; set; }
        public int UpdateTopicQueueCountInterval { get; set; }
        public int HeartbeatBrokerInterval { get; set; }
        public int PersistConsumerOffsetInterval { get; set; }
        public PullRequestSetting PullRequestSetting { get; set; }
        public MessageModel MessageModel { get; set; }
        public MessageHandleMode MessageHandleMode { get; set; }

        public ConsumerSetting()
        {
            BrokerAddress = SocketUtils.GetLocalIPV4().ToString();
            BrokerPort = 5001;
            RebalanceInterval = 1000 * 5;
            HeartbeatBrokerInterval = 1000 * 5;
            UpdateTopicQueueCountInterval = 1000 * 5;
            PersistConsumerOffsetInterval = 1000 * 5;
            PullRequestSetting = new PullRequestSetting();
            MessageModel = MessageModel.Clustering;
            MessageHandleMode = MessageHandleMode.Parallel;
        }

EQueue兼容Linux/Mono，下面是CentOS 6.4/Mono 3.2.3 环境下的运行结果：

Kafka/Metaq设计思想学习笔记

EQueue - 详细谈一下消息持久化以及消息堆积的设计