代码改变世界

Webdis内部解析

2012-03-13 13:56  轩脉刃  阅读(2758)  评论(0编辑  收藏  举报

Webdis是redis的http代理,源代码在:git://github.com/nicolasff/webdis.git

webdis.json是配置文件

webdis.c是入口程序

 

其中有三个比较重要的结构:

struct server {

	int fd; 
	struct event ev;
	struct event_base *base;   //libevent的event事件

	struct conf *cfg;  //配置文件,设置有多少个进程(http_threads)啥的放在里面

	/* worker threads */
	struct worker **w;  //有多个worker,父进程有多少worker线程
	int next_worker;

	/* log lock */
	struct {
		pid_t self;
		int fd;
	} log;  //日志结构
};

 

 

struct worker {

	/* self */
	pthread_t thread;  
	struct event_base *base;  //libevent的event事件

	/* connection dispatcher */
	struct server *s;  //父server
	int link[2]; //由pipe建立的管道,link[0]是管道读取端,link[1]是管道写入端

	/* Redis connection pool */
	struct pool *pool;  //连接池,与redis连接的连接池
};
struct pool {

struct worker *w; //worker线程

struct conf *cfg; //配置文件

const redisAsyncContext **ac; //redis的同步上下文

int count; //pool大小,即s->cfg->pool_size_per_thread

int cur;

};

 

这三个结构每个结构都有一个指针指向父结构,比如pool的worker*

这样能保证从任意一个结构中都能取得父结构的需要的属性

 

 

webdis的server-worker-pool的关系是这样的:

一个Server包含多个worker,每个woker占一个进程的资源

一个worker包含一个pool

Server的任务是接受HTTP请求,传递HTTP的套接字给worker

Worker才是webdis的实际处理类,一方面接受Server传递过来的HTTP请求,一方面由pool保持和redis的连接

Pool是连接池,保持了与redis的连接,防止重复的连接操作造成过多的资源浪费

 

webdis的入口是Webdis.c文件

主要运行了:

Server_new(conf)

server_start(s)

 

-------------------------------new 的过程开始----------------------------------------

Server_new主要函数:

conf_read

worker_new * n

 

worker_new主要函数:

Pipe(w->link) //建立管道

w->pool = pool_new(w, s->cfg->pool_size_per_thread);

 

pool_new主要函数:

p->ac = calloc(count, sizeof(redisAsyncContext*));

p->cfg = w->s->cfg;

 

pool中有一个redisAsyncContext结构,这个结构是hiredis的范围了:

Hiredis是redis的C客户端库

https://github.com/antirez/hiredis

Hiredis is a minimalistic C client library for the Redis database.

 

Hiredis:

使用方法大是:

redisAsyncContext *c = redisAsyncConnect("127.0.0.1", 6379);

int redisAsyncCommand(

redisAsyncContext *ac, redisCallbackFn *fn, void *privdata,

const char *format, ...);

int redisAsyncCommandArgv(

redisAsyncContext *ac, redisCallbackFn *fn, void *privdata,

int argc, const char **argv, const size_t *argvlen);

void redisAsyncDisconnect(redisAsyncContext *ac);
-------------------------------new 的过程结束----------------------------------------
 
-------------------------------start 的过程开始---------------------------------------

server_start主要函数:

 

s->base = event_base_new(); //注册一个事件

worker_start(s->w[i]); //开启worker

s->fd = socket_setup(s->cfg->http_host, s->cfg->http_port); //建立socket



/* start http server */

event_set(&s->ev, s->fd, EV_READ | EV_PERSIST, server_can_accept, s);

event_base_set(s->base, &s->ev);

event_add(&s->ev, NULL);

event_base_dispatch(s->base);

 

worker_start主要函数:

pthread_create(&w->thread, NULL, worker_main, w);//开了一个线程来运行worker_main函数

worker_main主要函数:

w->base = event_base_new(); //注册event

/* monitor pipe link */
event_set(&ev, w->link[0], EV_READ | EV_PERSIST, worker_on_new_client, w);
event_base_set(w->base, &ev);
event_add(&ev, NULL);

/* connect to Redis */
worker_pool_connect(w); //worker和pool的连接,即worker和redis的连接

/* loop */
event_base_dispatch(w->base);

worker_pool_connect主要函数:

pool_connect(w->pool, 1); //指定w->pool来连接redis

 

pool_connect主要函数: //连接redis

ac = redisAsyncConnect(p->cfg->redis_host, p->cfg->redis_port);

redisLibeventAttach(ac, p->w->base);

redisAsyncSetConnectCallback(ac, pool_on_connect);

redisAsyncSetDisconnectCallback(ac, pool_on_disconnect);

redisAsyncCommand(ac, NULL, NULL, "AUTH %s", p->cfg->redis_auth);
下面就进入到了hiredis的部分了
-------------------------------start 的过程结束---------------------------------------
 

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

作者:yjf512(轩脉刃)

出处:http://www.cnblogs.com/yjf512/

本文版权归yjf512和cnBlog共有,欢迎转载,但未经作者同意必须保留此段声明