Heim >Backend-Entwicklung >PHP-Tutorial >四:深入Nginx之事件和连接 (之一)

四:深入Nginx之事件和连接 (之一)

WBOY
WBOYOriginal
2016-07-28 08:29:55834Durchsuche

Nginx 本质上是基于事件驱动的Web服务器,事件 处理框架所要解决的问题就是如何收集、管理、分发事件。
事件主要 以

  • 网络事件(TCP网络事件为主 )
  • 定时器事件

Nginx 定义一个 核心模块ngx_event_module,参考博客一深入理解Nginx的 模块化 ,全局观,Nginx在启动时会调用ngx_init_cycle方法 解析配置文件时,一旦 在nginx.conf中找到感兴趣的是”events {}”配置项,ngx_event_module模块开始工作了。在 核心模块ngx_event_module中的ngx_event_core_module模块定义了这个模块会使用哪种事件驱动机制以及如何管理 事件。

事件 模块是 一种新的模块类型,nginx_module_t表示Nginx模块接口,而针对每一种 不同类型的模块,都有一个结构体来描述这一类模块的通用接口,这个接口保存在ngx_module_t结构体的 ctx成员中。核心 模块的 通用接口是ngx_core_module_t,事件通用 接口 是 ngx_event_module_t结构 体:

<code><span>typedef</span><span>struct</span> {
<span>//位于文件 ngx_event.h</span><span>//事件模块 的名字</span>
    ngx_str_t              *name;
    <span>//在解析 配置前,这个回调 方法用于创建存储配置选项参数的结构体</span><span>void</span>                 *(*create_conf)(ngx_cycle_t *cycle);
    <span>// 在解析配置项完成 之后,init_conf方法会被调用,用以综合处理当前事件模块感兴趣的全部配置项</span><span>char</span>                 *(*init_conf)(ngx_cycle_t *cycle, <span>void</span> *conf);
    <span>// 对于事件驱动机制,每个事件模块需要实现10个抽象方法 </span>
    ngx_event_actions_t     actions;
} ngx_event_module_t;</code>

ngx_event_module_t中的actions成员是定义事件驱动模块的核心方法。

<code><span>typedef</span><span>struct</span> {
    <span>/* 添加事件方法,它将负责把1个感兴趣 事件添加到操作 系统提供的事件驱动机制 */</span>
    ngx_int_t  (*add)(ngx_event_t *ev, ngx_int_t event, ngx_uint_t flags);
    <span>/*删除事件方法,它将1个已经存在于事件驱动机制中的事件移除 */</span>
    ngx_int_t  (*del)(ngx_event_t *ev, ngx_int_t event, ngx_uint_t flags);
    <span>/*启用 1个 事件,目前事件框架不会调用 这个 方法,大部分事件驱动模块对于该方法的实现都是与上面的add方法完全一致的*/</span>
    ngx_int_t  (*enable)(ngx_event_t *ev, ngx_int_t event, ngx_uint_t flags);
    <span>/*禁用1个事件,目前事件框架不会 调用这个方法 ,大部分事件驱动器对于这个方法的实现与上面的del方法完全一致*/</span>
    ngx_int_t  (*disable)(ngx_event_t *ev, ngx_int_t event, ngx_uint_t flags);
    <span>/*向事件驱动机制中添加一个新的连接,这意味着连接上的读写事件都添加到事件驱动机制中*/</span>
    ngx_int_t  (*add_conn)(ngx_connection_t *c);
    <span>/*从事件驱动机制中移除一个连接的读写事件*/</span>
    ngx_int_t  (*del_conn)(ngx_connection_t *c, ngx_uint_t flags);
    <span>/*仅在多线程环境下会被调用,目前Nginx在产品环境下还不会以多线程方式运行*/</span>
    ngx_int_t  (*notify)(ngx_event_handler_pt handler);
    <span>/*在正常的工作循环中,将调用process_events方法来 处理事件*/</span>
    ngx_int_t  (*process_events)(ngx_cycle_t *cycle, ngx_msec_t timer,
                                 ngx_uint_t flags);
   <span>/*初始化事件驱动模块的方法*/</span>
    ngx_int_t  (*init)(ngx_cycle_t *cycle, ngx_msec_t timer);
    <span>// 退出事件驱动模块前调用的方法</span><span>void</span>       (*done)(ngx_cycle_t *cycle);
} ngx_event_actions_t;</code>

ngx_event_core_module 和9个事件驱动模块都必须ngx_module_t结构体的ctx成员 实现ngx_event_module_t 接口

<code><span>struct</span> ngx_event_s {
    <span>/*事件相关的对象,通常data指向ngx_connection_t连接对象。开启文件异步I/O 时,它可能会指向*/</span><span>void</span>            *data;
     <span>/* 标志位,标识事件可写,意味着对应的TCP连接可写,也即连接处于发送网络包状态 */</span><span>unsigned</span>         write:<span>1</span>;
     <span>/* 标志位,标识可建立新的连接,一般是在ngx_listening_t对应的读事件中标记 */</span><span>unsigned</span>         accept:<span>1</span>;
     <span>/*检测当前事件是否是过期的,它仅仅是给驱动模块使用的,而事件消费模块可以不用关心 */</span><span>/* used to detect the stale events in kqueue and epoll */</span><span>unsigned</span>         instance:<span>1</span>;

    <span>/*
     * the event was passed or would be passed to a kernel;
     * in aio mode - operation was posted.
     */</span><span>unsigned</span>         active:<span>1</span>;

    <span>unsigned</span>         disabled:<span>1</span>;

    <span>/* the ready event; in aio mode 0 means that no operation can be posted */</span><span>unsigned</span>         ready:<span>1</span>;

    <span>unsigned</span>         oneshot:<span>1</span>;

    <span>/* aio operation is complete */</span><span>unsigned</span>         complete:<span>1</span>;

    <span>unsigned</span>         eof:<span>1</span>;
    <span>unsigned</span>         error:<span>1</span>;

    <span>unsigned</span>         timedout:<span>1</span>;
    <span>unsigned</span>         timer_set:<span>1</span>;

    <span>unsigned</span>         delayed:<span>1</span>;

    <span>unsigned</span>         deferred_accept:<span>1</span>;

    <span>/* the pending eof reported by kqueue, epoll or in aio chain operation */</span><span>unsigned</span>         pending_eof:<span>1</span>;

    <span>unsigned</span>         posted:<span>1</span>;

    <span>unsigned</span>         closed:<span>1</span>;

    <span>/* to test on worker exit */</span><span>unsigned</span>         channel:<span>1</span>;
    <span>unsigned</span>         resolver:<span>1</span>;

    <span>unsigned</span>         cancelable:<span>1</span>;

<span>#if (NGX_WIN32)</span><span>/* setsockopt(SO_UPDATE_ACCEPT_CONTEXT) was successful */</span><span>unsigned</span>         accept_context_updated:<span>1</span>;
<span>#endif</span><span>#if (NGX_HAVE_KQUEUE)</span><span>unsigned</span>         kq_vnode:<span>1</span>;

    <span>/* the pending errno reported by kqueue */</span><span>int</span>              kq_errno;
<span>#endif</span><span>/*
     * kqueue only:
     *   accept:     number of sockets that wait to be accepted
     *   read:       bytes to read when event is ready
     *               or lowat when event is set with NGX_LOWAT_EVENT flag
     *   write:      available space in buffer when event is ready
     *               or lowat when event is set with NGX_LOWAT_EVENT flag
     *
     * epoll with EPOLLRDHUP:
     *   accept:     1 if accept many, 0 otherwise
     *   read:       1 if there can be data to read, 0 otherwise
     *
     * iocp: TODO
     *
     * otherwise:
     *   accept:     1 if accept many, 0 otherwise
     */</span><span>#if (NGX_HAVE_KQUEUE) || (NGX_HAVE_IOCP)</span><span>int</span>              available;
<span>#else</span><span>unsigned</span>         available:<span>1</span>;
<span>#endif</span>    ngx_event_handler_pt  handler;


<span>#if (NGX_HAVE_IOCP)</span>
    ngx_event_ovlp_t ovlp;
<span>#endif</span>    ngx_uint_t       index;

    ngx_log_t       *<span>log</span>;

    ngx_rbtree_node_t   timer;

    <span>/* the posted queue */</span>
    ngx_queue_t      <span>queue</span>;

<span>#if 0</span><span>/* the threads support */</span><span>/*
     * the event thread context, we store it here
     * if $(CC) does not understand __thread declaration
     * and pthread_getspecific() is too costly
     */</span><span>void</span>            *thr_ctx;

<span>#if (NGX_EVENT_T_PADDING)</span><span>/* event should not cross cache line in SMP */</span>    uint32_t         padding[NGX_EVENT_T_PADDING];
<span>#endif</span><span>#endif</span>
};


<span>#if (NGX_HAVE_FILE_AIO)</span><span>struct</span> ngx_event_aio_s {
    <span>void</span>                      *data;
    ngx_event_handler_pt       handler;
    ngx_file_t                *file;

<span>#if (NGX_HAVE_AIO_SENDFILE)</span>
    ssize_t                  (*preload_handler)(ngx_buf_t *file);
<span>#endif</span>    ngx_fd_t                   fd;

<span>#if (NGX_HAVE_EVENTFD)</span>
    int64_t                    res;
<span>#endif</span><span>#if !(NGX_HAVE_EVENTFD) || (NGX_TEST_BUILD_EPOLL)</span>
    ngx_err_t                  err;
    size_t                     nbytes;
<span>#endif</span>    ngx_aiocb_t                aiocb;
    ngx_event_t                event;
};

<span>#endif</span></code>

Nginx的连接

 被动连接

这种连接是指 客户端发起的,服务器被动接受的连接

<code><span>//在 文件ngx_connection.h中</span><span>struct</span> ngx_connection_s {
  <span>/*  连接未使用时,data成员用于充当连接池中空闲链表中的next指针。当连接被使用时,data的意义由使用它的Nginx模块而定。在HTTP模块中,data指向ngx_http_request_t请求*/</span><span>void</span>               *data;
    <span>// 连接对应的读事件 </span>
    ngx_event_t        *read;
    <span>// 连接对应的写事件 </span>
    ngx_event_t        *write;
    <span>// 套接字句柄 </span>
    ngx_socket_t        fd;
    <span>//直接接收网络字符流的方法</span>
    ngx_recv_pt         recv;
    <span>// 直接发送网络字符流的办法</span>
    ngx_send_pt         send;
    <span>// 以ngx_chain_t链表为 参数来 接收 网络 字符流的方法 </span>
    ngx_recv_chain_pt   recv_chain;
    <span>// 以ngx_chain_t链表为 参数来 发送 网络 字符流的方法 </span>
    ngx_send_chain_pt   send_chain;
    <span>/*这个连接对应的ngx_listening_t监听对象,此连接由listening监听端口的事件建立*/</span>
    ngx_listening_t    *listening;
    <span>//这个连接上已经发送出去的字节数</span>
    off_t               sent;
    <span>// 可以记录日志的ngx_log_t对象</span>
    ngx_log_t          *<span>log</span>;
    <span>/* 内存池。一般在accept一个新连接时,会创建一个 内存池,而在这个 连接结束时会销毁内存池。所有的ngx_connectionn_t结构 体都是预分配,这个内存池的大小将由上面的listening 监听对象中的 pool_size成员决定*/</span>
    ngx_pool_t         *pool;

    <span>int</span>                 type;
    <span>// 连接客户端的sockaddr结构体</span><span>struct</span> sockaddr    *sockaddr;
    <span>// 连接 sockaddr结构体的 长度</span>
    socklen_t           socklen;
    <span>// 连接客户端字符串形式的IP地址</span>
    ngx_str_t           addr_text;

    ngx_str_t           proxy_protocol_addr;

    in_port_t           proxy_protocol_port;

<span>#if (NGX_SSL)</span>
    ngx_ssl_connection_t  *ssl;
<span>#endif</span><span>/*本机监听端口 对应 的sockaddr结构 体 ,也就是listening监听对象中的sock
    addr成员*/</span><span>struct</span> sockaddr    *local_sockaddr;
    socklen_t           local_socklen;
     <span>/*用于接收、缓存客户端 发来的字节流,每个事件消费模块可自由决定从连接池中分配多大空间给 buffer这个 缓存字段*/</span>
    ngx_buf_t          *buffer;

    ngx_queue_t         <span>queue</span>;
    <span>// 连接使用次数。ngx_connection_t结构体每次建立一条来自客户端的连接,或者主动向后端服务器发起连接时,number都会加1*/</span>
    ngx_atomic_uint_t   number;
    <span>// 处理 请求次数</span>
    ngx_uint_t          requests;

    <span>unsigned</span>            buffered:<span>8</span>;

    <span>unsigned</span>            log_error:<span>3</span>;     <span>/* ngx_connection_log_error_e */</span><span>unsigned</span>            timedout:<span>1</span>;
    <span>unsigned</span>            error:<span>1</span>;
    <span>unsigned</span>            destroyed:<span>1</span>;

    <span>unsigned</span>            idle:<span>1</span>;
    <span>unsigned</span>            reusable:<span>1</span>;
    <span>unsigned</span>            close:<span>1</span>;
    <span>unsigned</span>            shared:<span>1</span>;

    <span>unsigned</span>            sendfile:<span>1</span>;
    <span>unsigned</span>            sndlowat:<span>1</span>;
    <span>unsigned</span>            tcp_nodelay:<span>2</span>;   <span>/* ngx_connection_tcp_nodelay_e */</span><span>unsigned</span>            tcp_nopush:<span>2</span>;    <span>/* ngx_connection_tcp_nopush_e */</span><span>unsigned</span>            need_last_buf:<span>1</span>;

<span>#if (NGX_HAVE_IOCP)</span><span>unsigned</span>            accept_context_updated:<span>1</span>;
<span>#endif</span><span>#if (NGX_HAVE_AIO_SENDFILE)</span><span>unsigned</span>            busy_count:<span>2</span>;
<span>#endif</span><span>#if (NGX_THREADS)</span>
    ngx_thread_task_t  *sendfile_task;
<span>#endif</span>
};</code>

主动连接

作为Web服务器,Nginx也需要向其他服务器发起连接,使用ngx_peer_connection_t结构体来表示 主动连接,一个待处理连接的许多特性在 被动连接ngx_connection_t中都被定义过,因此ngx_peer_connection_t结构体中引用了ngx_connection_t这个结构体。

<code><span>// 当使用长连接与上游服务器通信时,可通过该方法由连接池获取 一个新的连接</span><span>typedef</span> ngx_int_t (*ngx_event_get_peer_pt)(ngx_peer_connection_t *pc,
    <span>void</span> *data);
<span>// 当 使用长连接与上游服务器通信时,通过该方法 将使用完毕 的连接释放给连接池</span><span>typedef</span><span>void</span> (*ngx_event_free_peer_pt)(ngx_peer_connection_t *pc, <span>void</span> *data,
    ngx_uint_t state);
<span>struct</span> ngx_peer_connection_s {
    <span>/*一个主动连接实际 上也需要ngx_connection_t结构体的大部分成员,并且处于重用的考虑 而定义 了connecion*/</span>
    ngx_connection_t                *connection;
    <span>// 远端服务器的socketaddr</span><span>struct</span> sockaddr                 *sockaddr;
    <span>// sockaddr地址长度</span>
    socklen_t                        socklen;
    <span>// 远端服务器的名称 </span>
    ngx_str_t                       *name;
    <span>// 表示在连接 一个 远端服务器,当前连接出现 异常失败后可以重试的次数,也就是允许的最多失败的次数</span>
    ngx_uint_t                       tries;
    ngx_msec_t                       start_time;

    ngx_event_get_peer_pt            get;
    ngx_event_free_peer_pt           <span>free</span>;
    <span>void</span>                            *data;

<span>#if (NGX_SSL)</span>
    ngx_event_set_peer_session_pt    set_session;
    ngx_event_save_peer_session_pt   save_session;
<span>#endif</span><span>// 本机地址信息 </span>
    ngx_addr_t                      *local;

    <span>int</span>                              type;
    <span>// 套接字的接收缓冲区大小</span><span>int</span>                              rcvbuf;
    <span>// 记录日志的ngx_log_t对象</span>
    ngx_log_t                       *<span>log</span>;

    <span>unsigned</span>                         cached:<span>1</span>;
<span>#if (NGX_HAVE_TRANSPARENT_PROXY)</span><span>unsigned</span>                         transparent:<span>1</span>;
<span>#endif</span><span>/* ngx_connection_log_error_e */</span><span>unsigned</span>                         log_error:<span>2</span>;
};</code>
').addClass('pre-numbering').hide(); $(this).addClass('has-numbering').parent().append($numbering); for (i = 1; i ').text(i)); }; $numbering.fadeIn(1700); }); });

以上就介绍了 四:深入Nginx之事件和连接 (之一),包括了方面的内容,希望对PHP教程有兴趣的朋友有所帮助。

Stellungnahme:
Der Inhalt dieses Artikels wird freiwillig von Internetnutzern beigesteuert und das Urheberrecht liegt beim ursprünglichen Autor. Diese Website übernimmt keine entsprechende rechtliche Verantwortung. Wenn Sie Inhalte finden, bei denen der Verdacht eines Plagiats oder einer Rechtsverletzung besteht, wenden Sie sich bitte an admin@php.cn