nodejs多核处理

前言
大家都知道nodejs是一个单进程单线程的服务器引擎，不管有多么的强大硬件，只能利用到单个CPU进行计算。所以，有人开发了第三方的cluster，让node可以利用多核CPU实现并行。

随着nodejs的发展，让nodejs上生产环境，就必须是支持多进程多核处理！在V0.6.0版本，Nodejs内置了cluster的特性。自此，Nodejs终于可以作为一个独立的应用开发解决方案，映入大家眼帘了。

cluster介绍
cluster的简单使用
cluster的工作原理
cluster的API
master和worker的通信
用cluster实现负载均衡(Load Balance) — win7失败
用cluster实现负载均衡(Load Balance) — ubuntu成功
cluster负载均衡策略的测试

1. cluster介绍

cluster是一个nodejs内置的模块，用于nodejs多核处理。cluster模块，可以帮助我们简化多进程并行化程序的开发难度，轻松构建一个用于负载均衡的集群。

2. cluster的简单使用

我的系统环境

win7 64bit
Nodejs:v0.10.5
Npm:1.2.19

在win的环境中，我们通过cluster启动多核的node提供web服务。

新建工程目录：



~ D:\workspace\javascript>mkdir nodejs-cluster && cd nodejs-cluster

新建文件：app.js



~ vi app.js

var cluster = require('cluster');

var http = require('http');

var numCPUs = require('os').cpus().length;

if (cluster.isMaster) {

    console.log("master start...");

    // Fork workers.

    for (var i = 0; i < numCPUs; i++) {

        cluster.fork();

    }

    cluster.on('listening',function(worker,address){

        console.log('listening: worker ' + worker.process.pid +', Address: '+address.address+":"+address.port);

    });

    cluster.on('exit', function(worker, code, signal) {

        console.log('worker ' + worker.process.pid + ' died');

    });

} else {

    http.createServer(function(req, res) {

        res.writeHead(200);

        res.end("hello world\n");

    }).listen(0);

}

在控制台启动node程序



~ D:\workspace\javascript\nodejs-cluster>node app.js

master start...

listening: worker 2368, Address: 0.0.0.0:57132

listening: worker 1880, Address: 0.0.0.0:57132

listening: worker 1384, Address: 0.0.0.0:57132

listening: worker 1652, Address: 0.0.0.0:57132

master是总控节点，worker是运行节点。然后根据CPU的数量，启动worker。我本地是双核双通道的CPU，所以被检测为4核，启动了4个worker。

3. cluster的工作原理

每个worker进程通过使用child_process.fork()函数，基于IPC（Inter-Process Communication，进程间通信），实现与master进程间通信。

当worker使用server.listen（...）函数时，将参数序列传递给master进程。如果master进程已经匹配workers，会将传递句柄给工人。如果master没有匹配好worker，那么会创建一个worker，再传递并句柄传递给worker。

在边界条件，有3个有趣的行为：
注：下面server.listen()，是对底层“http.Server-->net.Server”类的调用。

1. server.listen({fd: 7}):在master和worker通信过程，通过传递文件，master会监听“文件描述为7”，而不是传递“文件描述为7”的引用。
2. server.listen(handle):master和worker通信过程，通过handle函数进行通信，而不用进程联系
3. server.listen(0):在master和worker通信过程，集群中的worker会打开一个随机端口共用，通过socket通信，像上例中的57132

当多个进程都在 accept() 同样的资源的时候，操作系统的负载均衡非常高效。Node.js没有路由逻辑，worker之间没有共享状态。所以，程序要设计得简单一些，比如基于内存的session。

因为workers都是独力运行的，根据程序的需要，它们可以被独立删除或者重启，worker并不相互影响。只要还有workers存活，则master将继续接收连接。Node不会自动维护workers的数目。我们可以建立自己的连接池。

4. cluster的API

官网地址：http://nodejs.org/api/cluster.html#cluster_cluster

cluster对象
cluster的各种属性和函数

cluster.setttings:配置集群参数对象
cluster.isMaster:判断是不是master节点
cluster.isWorker:判断是不是worker节点
Event: 'fork': 监听创建worker进程事件
Event: 'online': 监听worker创建成功事件
Event: 'listening': 监听worker向master状态事件
Event: 'disconnect': 监听worker断线事件
Event: 'exit': 监听worker退出事件
Event: 'setup': 监听setupMaster事件
cluster.setupMaster([settings]): 设置集群参数
cluster.fork([env]): 创建worker进程
cluster.disconnect([callback]): 关闭worket进程
cluster.worker: 获得当前的worker对象
cluster.workers: 获得集群中所有存活的worker对象

worker对象
worker的各种属性和函数：可以通过cluster.workers, cluster.worket获得。

worker.id: 进程ID号
worker.process: ChildProcess对象
worker.suicide: 在disconnect()后，判断worker是否自杀
worker.send(message, [sendHandle]): master给worker发送消息。注：worker给发master发送消息要用process.send(message)
worker.kill([signal='SIGTERM']): 杀死指定的worker，别名destory()
worker.disconnect(): 断开worker连接，让worker自杀
Event: 'message': 监听master和worker的message事件
Event: 'online': 监听指定的worker创建成功事件
Event: 'listening': 监听master向worker状态事件
Event: 'disconnect': 监听worker断线事件
Event: 'exit': 监听worker退出事件

5. master和worker的通信

实现cluster的API，让master和worker相互通信。

新建文件: cluster.js



~ vi cluster.js

var cluster = require('cluster');

var http = require('http');

var numCPUs = require('os').cpus().length;

if (cluster.isMaster) {

    console.log('[master] ' + "start master...");

    for (var i = 0; i < numCPUs; i++) {

        var wk = cluster.fork();

        wk.send('[master] ' + 'hi worker' + wk.id);

    }

    cluster.on('fork', function (worker) {

        console.log('[master] ' + 'fork: worker' + worker.id);

    });

    cluster.on('online', function (worker) {

        console.log('[master] ' + 'online: worker' + worker.id);

    });

    cluster.on('listening', function (worker, address) {

        console.log('[master] ' + 'listening: worker' + worker.id + ',pid:' + worker.process.pid + ', Address:' + address.address + ":" + address.port);

    });

    cluster.on('disconnect', function (worker) {

        console.log('[master] ' + 'disconnect: worker' + worker.id);

    });

    cluster.on('exit', function (worker, code, signal) {

        console.log('[master] ' + 'exit worker' + worker.id + ' died');

    });

    function eachWorker(callback) {

        for (var id in cluster.workers) {

            callback(cluster.workers[id]);

        }

    }

    setTimeout(function () {

        eachWorker(function (worker) {

            worker.send('[master] ' + 'send message to worker' + worker.id);

        });

    }, 3000);

    Object.keys(cluster.workers).forEach(function(id) {

        cluster.workers[id].on('message', function(msg){

            console.log('[master] ' + 'message ' + msg);

        });

    });

} else if (cluster.isWorker) {

    console.log('[worker] ' + "start worker ..." + cluster.worker.id);

    process.on('message', function(msg) {

        console.log('[worker] '+msg);

        process.send('[worker] worker'+cluster.worker.id+' received!');

    });

    http.createServer(function (req, res) {

            res.writeHead(200, {"content-type": "text/html"});

            res.end('worker'+cluster.worker.id+',PID:'+process.pid);

    }).listen(3000);

}

控制台日志：



~ D:\workspace\javascript\nodejs-cluster>node cluster.js

[master] start master...

[worker] start worker ...1

[worker] [master] hi worker1

[worker] start worker ...2

[worker] [master] hi worker2

[master] fork: worker1

[master] fork: worker2

[master] fork: worker3

[master] fork: worker4

[master] online: worker1

[master] online: worker2

[master] message [worker] worker1 received!

[master] message [worker] worker2 received!

[master] listening: worker1,pid:6068, Address:0.0.0.0:3000

[master] listening: worker2,pid:1408, Address:0.0.0.0:3000

[master] online: worker3

[worker] start worker ...3

[worker] [master] hi worker3

[master] message [worker] worker3 received!

[master] listening: worker3,pid:3428, Address:0.0.0.0:3000

[master] online: worker4

[worker] start worker ...4

[worker] [master] hi worker4

[master] message [worker] worker4 received!

[master] listening: worker4,pid:6872, Address:0.0.0.0:3000

[worker] [master] send message to worker1

[worker] [master] send message to worker2

[worker] [master] send message to worker3

[worker] [master] send message to worker4

[master] message [worker] worker1 received!

[master] message [worker] worker2 received!

[master] message [worker] worker3 received!

[master] message [worker] worker4 received!

6. 用cluster实现负载均衡(Load Balance) -- win7失败

新建文件: server.js



~ vi server.js

var cluster = require('cluster');

var http = require('http');

var numCPUs = require('os').cpus().length;

if (cluster.isMaster) {

    console.log('[master] ' + "start master...");

    for (var i = 0; i < numCPUs; i++) {

         cluster.fork();

    }

    cluster.on('listening', function (worker, address) {

        console.log('[master] ' + 'listening: worker' + worker.id + ',pid:' + worker.process.pid + ', Address:' + address.address + ":" + address.port);

    });

} else if (cluster.isWorker) {

    console.log('[worker] ' + "start worker ..." + cluster.worker.id);

    http.createServer(function (req, res) {

        console.log('worker'+cluster.worker.id);

        res.end('worker'+cluster.worker.id+',PID:'+process.pid);

    }).listen(3000);

}

启动服务器：



~ D:\workspace\javascript\nodejs-cluster>node server.js

[master] start master...

[worker] start worker ...1

[worker] start worker ...2

[master] listening: worker1,pid:1536, Address:0.0.0.0:3000

[master] listening: worker2,pid:5920, Address:0.0.0.0:3000

[worker] start worker ...3

[master] listening: worker3,pid:7156, Address:0.0.0.0:3000

[worker] start worker ...4

[master] listening: worker4,pid:2868, Address:0.0.0.0:3000

worker4

worker4

worker4

worker4

worker4

worker4

worker4

worker4

用curl工具访问



C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

C:\Users\Administrator>curl localhost:3000

worker4,PID:2868

我们发现了cluster在win中的bug，只用到worker4。果断切换到Linux测试。

7. 用cluster实现负载均衡(Load Balance) -- ubuntu成功

Linux的系统环境

Linux: Ubuntu 12.04.2 64bit Server
Node: v0.11.2
Npm: 1.2.21

构建项目:不多解释



~ cd :/home/conan/nodejs/

~ mkdir nodejs-cluster && cd nodejs-cluster

~ vi server.js

var cluster = require('cluster');

var http = require('http');

var numCPUs = require('os').cpus().length;

if (cluster.isMaster) {

    console.log('[master] ' + "start master...");

    for (var i = 0; i < numCPUs; i++) {

         cluster.fork();

    }

    cluster.on('listening', function (worker, address) {

        console.log('[master] ' + 'listening: worker' + worker.id + ',pid:' + worker.process.pid + ', Address:' + address.address + ":" + address.port);

    });

} else if (cluster.isWorker) {

    console.log('[worker] ' + "start worker ..." + cluster.worker.id);

    http.createServer(function (req, res) {

        console.log('worker'+cluster.worker.id);

        res.end('worker'+cluster.worker.id+',PID:'+process.pid);

    }).listen(3000);

}

启动服务器



conan@conan-deskop:~/nodejs/nodejs-cluster$ node server.js

[master] start master...

[worker] start worker ...1

[master] listening: worker1,pid:2925, Address:0.0.0.0:3000

[worker] start worker ...3

[master] listening: worker3,pid:2931, Address:0.0.0.0:3000

[worker] start worker ...4

[master] listening: worker4,pid:2932, Address:0.0.0.0:3000

[worker] start worker ...2

[master] listening: worker2,pid:2930, Address:0.0.0.0:3000

worker4

worker2

worker1

worker3

worker4

worker2

worker1

用curl工具访问



C:\Users\Administrator>curl 192.168.1.20:3000

worker4,PID:2932

C:\Users\Administrator>curl 192.168.1.20:3000

worker2,PID:2930

C:\Users\Administrator>curl 192.168.1.20:3000

worker1,PID:2925

C:\Users\Administrator>curl 192.168.1.20:3000

worker3,PID:2931

C:\Users\Administrator>curl 192.168.1.20:3000

worker4,PID:2932

C:\Users\Administrator>curl 192.168.1.20:3000

worker2,PID:2930

C:\Users\Administrator>curl 192.168.1.20:3000

worker1,PID:2925

在Linux环境中，cluster是运行正确的！！！

8. cluster负载均衡策略的测试

我们在Linux下面，完成测试，用过测试软件： siege

安装siege

~ sudo apt-get install siege

启动node cluster

~ node server.js > server.log

运行siege启动命令,每秒50个并发请求。



~ sudo siege -c 50 http://localhost:3000

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.01 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.01 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.02 secs:      16 bytes ==> /

HTTP/1.1 200   0.00 secs:      16 bytes ==> /

HTTP/1.1 200   0.02 secs:      16 bytes ==> /

HTTP/1.1 200   0.01 secs:      16 bytes ==> /

HTTP/1.1 200   0.01 secs:      16 bytes ==> /

.....

^C

Lifting the server siege...      done.                                                                Transactions:                    3760 hits

Availability:                 100.00 %

Elapsed time:                  39.66 secs

Data transferred:               0.06 MB

Response time:                  0.01 secs

Transaction rate:              94.81 trans/sec

Throughput:                     0.00 MB/sec

Concurrency:                    1.24

Successful transactions:        3760

Failed transactions:               0

Longest transaction:            0.20

Shortest transaction:           0.00

FILE: /var/siege.log

You can disable this annoying message by editing

the .siegerc file in your home directory; change

the directive 'show-logfile' to false.

我们统计结果，执行3760次请求，消耗39.66秒，每秒处理94.81次请求。

查看server.log文件,



~  ls -l

total 64

-rw-rw-r-- 1 conan conan   756  9月 28 15:48 server.js

-rw-rw-r-- 1 conan conan 50313  9月 28 16:26 server.log

~ tail server.log

worker4

worker1

worker2

worker4

worker1

worker2

worker4

worker3

worker2

worker1

最后，用R语言分析一下:server.log



~ R 

> df<-read.table(file="server.log",skip=9,header=FALSE)

> summary(df)

       V1

 worker1:1559

 worker2:1579

 worker3:1570

 worker4:1535

我们看到，请求被分配到worker数据量相当。所以，cluster的负载均衡的策略，应该是随机分配的。

好了，我们又学了一个很有用的技能！利用cluster可以构建出多核应用，充分的利用多CPU带业的性能吧！！

转载请注明出处：
http://blog.fens.me/nodejs-core-cluster/