Tomcat不会在版本7.0.27之后进行流式传输？ [英] Tomcat doesn't do blazeds streaming after version 7.0.27?

问题描述

在tomcat版本$ subject之后，我发现我的flex应用程序不再工作。请帮我解决这个问题。我尝试了许多事情，没有成功。

这是我的配置。

Blaze - 4.x
Tomcat - Latest - 7.0。 32
Java 7

应用程序只是挂起而没有任何错误或任何东西。它从7.0.32再次运作。我检查了tomcat发行说明（ http://tomcat.apache.org/tomcat-7.0 -doc / changelog.html ）和孤立的问题，直到Coyote jar文件。

当我从版本7.0.27替换Coyote.jar到7.0.32时，它可以正常工作。

以下是Coyote的变化（ http://tomcat.apache.org/tomcat-7.0-doc/changelog.html ）

  52858：正确修复CPU负载过高（fhanik）
 53138：7.0.27（fhanik）中引入的SSL上的Sendfile损坏
 52055：需要额外修复以确保InputFilters在请求之间被回收。 （markt）
 53061：修复NIO连接器中的问题，从而如果轮询器处于低负载但一致负载（> 1请求/每秒且始终小于1秒）之间的超时从未发生过。 （markt）
 53063：在BIO中使用执行程序时，请将执行程序的maxThreads用作maxConnections的默认值。 （markt）
 53119：当使用APR /本地连接器从AJP连接完全写入响应之前，防止客户端断开连接时报告缓冲区溢出错误。 （markt）
 53169：通过设置Connection：close标题，允许开发人员避免对未知长度的响应进行分块编码。基于Philippe Marschall建议的补丁。 （markt）
 53173：正确倒计数maxConnections（fhanik）
更新NIO连接器的pollerThreadCount默认值。不管可用的处理器如何，新的默认值永远不会超过2。 （fhanik）
允许通过来自端点的getter检索当前的connectionCount，并作为ThreadPool mbean的JMX属性。 （rjung）
更正Tomcat连接器停止时Comet END事件未发送给连接客户端的边界情况。 （markt）
 53406：修复使用Comet时连接关闭时可能的堆栈溢出。 （fhanik）
改进InternalNioInputBuffer.parseHeaders（）。 （kkolinko）
在连接器上实现maxHeaderCount属性。它相当于Apache HTTPD的LimitRequestFields指令。默认值是100.（kkolinko）
  

任何猜测/想法都会有所帮助。感谢提前！

解决方案

我认为这可能是问题的原因：

 
 53169：允许开发人员通过设置Connection：close标题来避免对未知长度的响应进行分块编码。基于Philippe Marschall建议的补丁。 （markt）
 

从BaseStreamingHTTPEndPoint.java中的BlazeDS（4.0.0.14931）代码：

  718：res.setHeader（Connection，close）; 
  

所以，我写了一个简单的过滤器：

  @Override 
 public void doFilter（ServletRequest request，ServletResponse response，
 FilterChain chain）throws IOException，ServletException {
 chain.doFilter（request，new HttpServletResponseWrapper（
（HttpServletResponse）response）{$ b $ public void setHeader（String name，String value）{
 if（！（Connection.equalsIgnoreCase（name）&&Close 
 .equalsIgnoreCase（value）））{
 super.setHeader（name，value）; 
} 
} 
}）; 
 
 
 
 $ b $使用web.xml中的配置：
 
 
 < filter> 
< filter-name> MessageBrokerHack< / filter-name> 
< filter-class> 
 com.vivimice.testing.MessageBrokerHack 
< / filter-class> 
< / filter> 
< filter-mapping> 
< filter-name> MessageBrokerHack< / filter-name> 
< url-pattern> / messagebroker / *< / url-pattern> 
< / filter-mapping> 
  
然后问题就解决了。 
 
 
在AIR客户端（Flex 4.5.1）和Tomcat 6.0.36& 7.0.35 
 
 
 
注意：尚未在常规RemoteObject调用中测试。
 
I found my flex application doesn't work anymore after tomcat version $subject. Please help me to get through this. I tried many things without any success. 

Here is my configurations.

Blaze - 4.x
Tomcat - Latest - 7.0.32
Java 7

Application just hangs without any error or anything. It works again from 7.0.32. I checked tomcat release notes (http://tomcat.apache.org/tomcat-7.0-doc/changelog.html) and isolated problem up to Coyote jar file. 

When I replace Coyote.jar from version 7.0.27 to 7.0.32 it works properly. 

following are changes for Coyote (http://tomcat.apache.org/tomcat-7.0-doc/changelog.html)
52858: Correct fix for high CPU load (fhanik)
53138: Broken Sendfile on SSL introduced in 7.0.27 (fhanik)
52055: Additional fix required to ensure that InputFilters are recycled between requests. (markt)
53061: Fix a problem in the NIO connector whereby if the poller was under low but consistent load (>1 request/per second and always less than 1 second between requests) timeouts never took place. (markt)
53063: When using an Executor with BIO, use the executor's maxThreads as the default for maxConnections. (markt)
53119: Prevent buffer overflow errors being reported when a client disconnects before the response has been fully written from an AJP connection using the APR/native connector. (markt)
53169: Allow developers to avoid chunked encoding for a response of unknown length by setting the Connection: close header. Based on a patch suggested by Philippe Marschall. (markt)
53173: Properly count down maxConnections (fhanik)
 Update default value of pollerThreadCount for the NIO connector. The new default value will never go above 2 regardless of available processors. (fhanik)
 Allow to retrieve the current connectionCount via getter from the endpoint and as JMX attribute of the ThreadPool mbean. (rjung)
 Correct an edge case where Comet END events were not send to connected clients when the Tomcat connector was stopped. (markt)
53406: Fix possible stack overflow on connection close when using Comet. (fhanik)
 Improve InternalNioInputBuffer.parseHeaders(). (kkolinko)
 Implement maxHeaderCount attribute on Connector. It is equivalent of LimitRequestFields directive of Apache HTTPD. Default value is 100. (kkolinko)
Any guess / idea would help. Thanks in advance!
解决方案
I think this may be the cause of the problem:
53169: Allow developers to avoid chunked encoding for a response of unknown length by setting the Connection: close header. Based on a patch suggested by Philippe Marschall. (markt)
And from BlazeDS (4.0.0.14931) code in BaseStreamingHTTPEndPoint.java:
718: res.setHeader("Connection", "close");
So, I wrote a simple filter:
@Override
public void doFilter(ServletRequest request, ServletResponse response,
        FilterChain chain) throws IOException, ServletException {
    chain.doFilter(request, new HttpServletResponseWrapper(
            (HttpServletResponse) response) {
        public void setHeader(String name, String value) {
            if (!("Connection".equalsIgnoreCase(name) && "Close"
                    .equalsIgnoreCase(value))) {
                super.setHeader(name, value);
            }
        }
    });
}
With configurations in web.xml:
<filter>
    <filter-name>MessageBrokerHack</filter-name>
    <filter-class>
        com.vivimice.testing.MessageBrokerHack
    </filter-class>
</filter>
<filter-mapping>
    <filter-name>MessageBrokerHack</filter-name>
    <url-pattern>/messagebroker/*</url-pattern>
</filter-mapping>
And then the problem solved.

Tested on AIR client (Flex 4.5.1) and Tomcat 6.0.36 & 7.0.35

Note: Not tested on regular RemoteObject calls yet.

                        
这篇关于Tomcat不会在版本7.0.27之后进行流式传输？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！