Websocket出现的错误

前端使用sockjs，后台使用spring的websocket框架

结果在一个网络较慢的地方，发现tomcat报错信息：

Oct 28, 2015 10:10:43 AM org.apache.catalina.core.StandardWrapperValve invoke
SEVERE: Servlet.service() for servlet [mvc-dispatcher] in context with path [/rscc] threw exception [Request processing failed; nested exception is org.springframework.web.socket.sockjs.SockJsException: Uncaught failure in SockJS request, uri=http://xxx/user/854/qckzogtf/xhr_streaming; nested exception is org.springframework.web.socket.sockjs.SockJsException: Uncaught failure for request http://xxx/user/854/qckzogtf/xhr_streaming; nested exception is java.lang.IllegalArgumentException: Async support must be enabled on a servlet and for all filters involved in async request processing. This is done in Java code using the Servlet API or by adding "<async-supported>true</async-supported>" to servlet and filter declarations in web.xml. Also you must use a Servlet 3.0+ container] with root cause
java.lang.IllegalArgumentException: Async support must be enabled on a servlet and for all filters involved in async request processing. This is done in Java code using the Servlet API or by adding "<async-supported>true</async-supported>" to servlet and filter declarations in web.xml. Also you must use a Servlet 3.0+ container
    at org.springframework.util.Assert.isTrue(Assert.java:65)
    at org.springframework.http.server.ServletServerHttpAsyncRequestControl.<init>(ServletServerHttpAsyncRequestControl.java:59)
    at org.springframework.http.server.ServletServerHttpRequest.getAsyncRequestControl(ServletServerHttpRequest.java:202)
    at org.springframework.web.socket.sockjs.transport.session.AbstractHttpSockJsSession.initRequest(AbstractHttpSockJsSession.java:238)
    at org.springframework.web.socket.sockjs.transport.session.AbstractHttpSockJsSession.handleInitialRequest(AbstractHttpSockJsSession.java:203)
    at org.springframework.web.socket.sockjs.transport.session.StreamingSockJsSession.handleInitialRequest(StreamingSockJsSession.java:54)
    at org.springframework.web.socket.sockjs.transport.handler.AbstractHttpSendingTransportHandler.handleRequestInternal(AbstractHttpSendingTransportHandler.java:66)
    at org.springframework.web.socket.sockjs.transport.handler.AbstractHttpSendingTransportHandler.handleRequest(AbstractHttpSendingTransportHandler.java:58)
    at org.springframework.web.socket.sockjs.transport.TransportHandlingSockJsService.handleTransportRequest(TransportHandlingSockJsService.java:254)
    at org.springframework.web.socket.sockjs.support.AbstractSockJsService.handleRequest(AbstractSockJsService.java:317)
    at org.springframework.web.socket.sockjs.support.SockJsHttpRequestHandler.handleRequest(SockJsHttpRequestHandler.java:88)
    at org.springframework.web.servlet.mvc.HttpRequestHandlerAdapter.handle(HttpRequestHandlerAdapter.java:51)
    at org.springframework.web.servlet.DispatcherServlet.doDispatch(DispatcherServlet.java:938)
    at org.springframework.web.servlet.DispatcherServlet.doService(DispatcherServlet.java:870)
    at org.springframework.web.servlet.FrameworkServlet.processRequest(FrameworkServlet.java:961)
    at org.springframework.web.servlet.FrameworkServlet.doPost(FrameworkServlet.java:863)
    at javax.servlet.http.HttpServlet.service(HttpServlet.java:646)
    at org.springframework.web.servlet.FrameworkServlet.service(FrameworkServlet.java:837)
    at javax.servlet.http.HttpServlet.service(HttpServlet.java:727)
    at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:303)
    at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:208)
    at org.apache.tomcat.websocket.server.WsFilter.doFilter(WsFilter.java:52)
    at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:241)
    at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:208)
    at org.springframework.web.filter.CharacterEncodingFilter.doFilterInternal(CharacterEncodingFilter.java:88)
    at org.springframework.web.filter.OncePerRequestFilter.doFilter(OncePerRequestFilter.java:108)
    at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:241)
    at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:208)
    at org.apache.catalina.core.StandardWrapperValve.invoke(StandardWrapperValve.java:220)
    at org.apache.catalina.core.StandardContextValve.invoke(StandardContextValve.java:122)
    at org.apache.catalina.authenticator.AuthenticatorBase.invoke(AuthenticatorBase.java:501)
    at org.apache.catalina.core.StandardHostValve.invoke(StandardHostValve.java:171)
    at org.apache.catalina.valves.ErrorReportValve.invoke(ErrorReportValve.java:102)
    at org.apache.catalina.valves.AccessLogValve.invoke(AccessLogValve.java:950)
    at org.apache.catalina.core.StandardEngineValve.invoke(StandardEngineValve.java:116)
    at org.apache.catalina.connector.CoyoteAdapter.service(CoyoteAdapter.java:408)
    at org.apache.coyote.http11.AbstractHttp11Processor.process(AbstractHttp11Processor.java:1040)
    at org.apache.coyote.AbstractProtocol$AbstractConnectionHandler.process(AbstractProtocol.java:607)
    at org.apache.tomcat.util.net.JIoEndpoint$SocketProcessor.run(JIoEndpoint.java:316)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
    at org.apache.tomcat.util.threads.TaskThread$WrappingRunnable.run(TaskThread.java:61)
    at java.lang.Thread.run(Thread.java:745)

根据报错信息来看，应该是缺少了<async-supported>true</async-supported>这个配置，这个是3.0开始支持的，async的请求需要开启async-supported。

但是该项目在我们本地从来没有出现过这个问题，情况说明及解决分析过程如下：

一、前端连接情况：ws = new SockJS( 'url', undefined, {});

　　不加参数，sockjs默认会选择最优方式来连接，情况如下：

　　

　　关于协议选择：在第一种连接方式超时时，sockjs会选择次优方式进行连接：

var _all_protocols = ['websocket',
                      'xdr-streaming',
                      'xhr-streaming',
                      'iframe-eventsource',
                      'iframe-htmlfile',
                      'xdr-polling',
                      'xhr-polling',
                      'iframe-xhr-polling',
                      'jsonp-polling'];
所有协议（连接方式）

　　检测所有可用协议并按优先级排序

utils.probeProtocols = function() {
    var probed = {};
    for(var i=0; i<_all_protocols.length; i++) {
        var protocol = _all_protocols[i];
        // User can have a typo in protocol name.
        probed[protocol] = SockJS[protocol] &&
                           SockJS[protocol].enabled();
    }
    return probed;
};

utils.detectProtocols = function(probed, protocols_whitelist, info) {
    var pe = {},
        protocols = [];
    if (!protocols_whitelist) protocols_whitelist = _all_protocols;
    for(var i=0; i<protocols_whitelist.length; i++) {
        var protocol = protocols_whitelist[i];
        pe[protocol] = probed[protocol];
    }
    var maybe_push = function(protos) {
        var proto = protos.shift();
        if (pe[proto]) {
            protocols.push(proto);
        } else {
            if (protos.length > 0) {
                maybe_push(protos);
            }
        }
    }

    // 1. Websocket
    if (info.websocket !== false) {
        maybe_push(['websocket']);
    }

    // 2. Streaming
    if (pe['xhr-streaming'] && !info.null_origin) {
        protocols.push('xhr-streaming');
    } else {
        if (pe['xdr-streaming'] && !info.cookie_needed && !info.null_origin) {
            protocols.push('xdr-streaming');
        } else {
            maybe_push(['iframe-eventsource',
                        'iframe-htmlfile']);
        }
    }

    // 3. Polling
    if (pe['xhr-polling'] && !info.null_origin) {
        protocols.push('xhr-polling');
    } else {
        if (pe['xdr-polling'] && !info.cookie_needed && !info.null_origin) {
            protocols.push('xdr-polling');
        } else {
            maybe_push(['iframe-xhr-polling',
                        'jsonp-polling']);
        }
    }
    return protocols;
}

　　连接时若超时切换协议的代码

while(1) {
        var protocol = that.protocol = that._protocols.shift();
        if (!protocol) {
            return false;
        }
        // Some protocols require access to `body`, what if were in
        // the `head`?
        if (SockJS[protocol] &&
            SockJS[protocol].need_body === true &&
            (!_document.body ||
             (typeof _document.readyState !== 'undefined'
              && _document.readyState !== 'complete'))) {
            that._protocols.unshift(protocol);
            that.protocol = 'waiting-for-load';
            utils.attachEvent('load', function(){
                that._try_next_protocol();
            });
            return true;
        }
//下面的to = 就是计算连接超时时间的，调用delay，在连接超时时关闭这个协议的连接。
        if (!SockJS[protocol] ||
              !SockJS[protocol].enabled(that._options)) {
            that._debug('Skipping transport:', protocol);
        } else {
            var roundTrips = SockJS[protocol].roundTrips || 1;
            var to = ((that._options.rto || 0) * roundTrips) || 5000;
            that._transport_tref = utils.delay(to, function() {
                if (that.readyState === SockJS.CONNECTING) {
                    // I can't understand how it is possible to run
                    // this timer, when the state is CLOSED, but
                    // apparently in IE everythin is possible.
                    that._didClose(2007, "Transport timeouted");
                }
            });

            var connid = utils.random_string(8);
            var trans_url = that._base_url + '/' + that._server + '/' + connid;
            that._debug('Opening transport:', protocol, ' url:'+trans_url,
                        ' RTO:'+that._options.rto);
            that._transport = new SockJS[protocol](that, trans_url,
                                                   that._base_url);
            return true;
        }
    }

　　 观察上面报错信息中的uri：uri=http://xxx/user/854/qckzogtf/xhr_streaming; 因为正常情况下连接为websocket连接，这个uri应该为ws=http://xxx/user/854/qckzogtf/websocket，由此可见，这里应该是切换为xhr_streaming协议了

　　前面也说过，他们网络环境较慢，所以才想，应该是连接超时导致的，有上面可知，to为超时时间，计算公式为

　　 var to = ((that._options.rto || 0) * roundTrips) || 5000;

　　关键在于that._options.rto，这个是new的时候的设置项，所以可以设置rto超时长一点。

　　结果发现设置中只有

　　that._options = {devel: false, debug: false, protocols_whitelist: [],info: undefined, rtt: undefined};

　　这几项，并没有rto，于是继续看

var SockJS = function(url, dep_protocols_whitelist, options) {
    if (this === _window) {
        // makes `new` optional
        return new SockJS(url, dep_protocols_whitelist, options);
    }

    var that = this, protocols_whitelist;
    that._options = {devel: false, debug: false, protocols_whitelist: [],
                     info: undefined, rtt: undefined};
    if (options) {
        utils.objectExtend(that._options, options);
    }
    that._base_url = utils.amendUrl(url);
    that._server = that._options.server || utils.random_number_string(1000);
    if (that._options.protocols_whitelist &&
        that._options.protocols_whitelist.length) {
        protocols_whitelist = that._options.protocols_whitelist;
    } else {
        // Deprecated API
        if (typeof dep_protocols_whitelist === 'string' &&
            dep_protocols_whitelist.length > 0) {
            protocols_whitelist = [dep_protocols_whitelist];
        } else if (utils.isArray(dep_protocols_whitelist)) {
            protocols_whitelist = dep_protocols_whitelist
        } else {
            protocols_whitelist = null;
        }
        if (protocols_whitelist) {
            that._debug('Deprecated API: Use "protocols_whitelist" option ' +
                        'instead of supplying protocol list as a second ' +
                        'parameter to SockJS constructor.');
        }
    }
    that._protocols = [];
    that.protocol = null;
    that.readyState = SockJS.CONNECTING;
    that._ir = createInfoReceiver(that._base_url);
    that._ir.onfinish = function(info, rtt) {
        that._ir = null;
        if (info) {
            if (that._options.info) {
                // Override if user supplies the option
                info = utils.objectExtend(info, that._options.info);
            }
            if (that._options.rtt) {
                rtt = that._options.rtt;
            }
            that._applyInfo(info, rtt, protocols_whitelist);
            that._didClose();
        } else {
            that._didClose(1002, 'Can\'t connect to server', true);
        }
    };
};

utils.objectExtend(that._options, options);这一句比较可疑是转移属性的，

that._applyInfo(info, rtt, protocols_whitelist);比较可疑，看一下

SockJS.prototype._applyInfo = function(info, rtt, protocols_whitelist) {
    var that = this;
    that._options.info = info;
    that._options.rtt = rtt;
    that._options.rto = utils.countRTO(rtt);
    that._options.info.null_origin = !_document.domain;
    var probed = utils.probeProtocols();
    that._protocols = utils.detectProtocols(probed, protocols_whitelist, info);
};

rto设置找到了！that._options.rto = utils.countRTO(rtt);

utils.countRTO = function (rtt) {
    var rto;
    if (rtt > 100) {
        rto = 3 * rtt; // rto > 300msec
    } else {
        rto = rtt + 200; // 200msec < rto <= 300msec
    }
    return rto;
}

好了，只用设置rtt既可以设置rto了。。。就这样，连接超时的问题算是解决了。

二、但是还没完，就算websocket连接超时导致协议切换为xhr_streaming，也不会导致后台报错的情况出现，下面解决这个问题：

报错处的代码：在AbstractHttpSockJsSession类中：

private void initRequest(ServerHttpRequest request, ServerHttpResponse response,
            SockJsFrameFormat frameFormat) {

        Assert.notNull(request, "Request must not be null");
        Assert.notNull(response, "Response must not be null");
        Assert.notNull(frameFormat, "SockJsFrameFormat must not be null");

        this.response = response;
        this.frameFormat = frameFormat;
        this.asyncRequestControl = request.getAsyncRequestControl(response);
    }

最后一句报错咯，看代码：

public ServerHttpAsyncRequestControl getAsyncRequestControl(ServerHttpResponse response) {
        if (this.asyncRequestControl == null) {
            Assert.isInstanceOf(ServletServerHttpResponse.class, response);
            ServletServerHttpResponse servletServerResponse = (ServletServerHttpResponse) response;
            this.asyncRequestControl = new ServletServerHttpAsyncRequestControl(this, servletServerResponse);//这一句报错了
        }
        return this.asyncRequestControl;
    }

new ServletServerHttpAsyncRequestControl时报错

public ServletServerHttpAsyncRequestControl(ServletServerHttpRequest request, ServletServerHttpResponse response) {

        Assert.notNull(request, "request is required");
        Assert.notNull(response, "response is required");

        Assert.isTrue(request.getServletRequest().isAsyncSupported(),
                "Async support must be enabled on a servlet and for all filters involved " +
                "in async request processing. This is done in Java code using the Servlet API " +
                "or by adding \"<async-supported>true</async-supported>\" to servlet and " +
                "filter declarations in web.xml. Also you must use a Servlet 3.0+ container");

        this.request = request;
        this.response = response;
    }

　　最长那一句断言报的错，说的是web.xml的servlet和filter中要加入<async-supported>true</async-supported>

　　看下web.xml中，filter中确实没有这一句，之所以一定要在filter中加入这一句，是因为websocket的切换协议请求，是通过filter拦截的，如果不在filter中配置async，则切换协议的请求将不是async的，所以上面就报错了。

　　filter中加上之后，就好了。

　　若为了让controller也支持async则需要在dispatcher中这样配置

<mvc:annotation-driven>
<!--  可不设置，使用默认的超时时间 -->
    <mvc:async-support default-timeout="3000"/>
</mvc:annotation-driven>

　　再返回上面initRequest方法，只有该类AbstractHttpSockJsSession的public void handleInitialRequest(ServerHttpRequest request, ServerHttpResponse response,SockJsFrameFormat frameFormat)方法与

public void handleSuccessiveRequest(ServerHttpRequest request,ServerHttpResponse response, SockJsFrameFormat frameFormat)中有调用，而调用这两个方法的只有AbstractHttpSockJsSession的实现类：StreamingSockJsSession才有，就是spring-websocket为sockjs支持的xhr-streaming方式的实现类，而平时使用websocket的则是另外一个实现类：

这就导致了之前一直没有报错，上了一个网络较差的地方就报错了，原因分析完毕。

上面用的sockjs 0.3.4，最新的是1.0.3有所改变，但是大致相同

补充：相同个毛线，1.0.3的rtt不再是通过参数传进去的，而是计算出来的

function InfoAjax(url, AjaxObject) {
  EventEmitter.call(this);

  var self = this;
  var t0 = +new Date();
  this.xo = new AjaxObject('GET', url);

  this.xo.once('finish', function(status, text) {
    var info, rtt;
    if (status === 200) {
      rtt = (+new Date()) - t0;
      if (text) {
        try {
          info = JSON3.parse(text);
        } catch (e) {
          debug('bad json', text);
        }
      }

      if (!objectUtils.isObject(info)) {
        info = {};
      }
    }
    self.emit('finish', info, rtt);
    self.removeAllListeners();
  });
}

也就是说，超时时间控制不了的...什么鬼

this._transportsWhitelist = options.transports;这个是设置白名单，优先使用这些方式。