[ 
https://issues.apache.org/jira/browse/CXF-865?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12540798
 ] 

Jonathan Gathman commented on CXF-865:
--------------------------------------

I ran through a few scenarios, and several things come to light for me.

JettyHTTPServerEngine.stop() FIRST calls "connector.close()".

It is this closing connector which is causing the exception to be thrown, not 
Jetty's Server.stop().

Second, it's clear that it's a thread timing issue, because when I run in Debug 
mode, it sometimes ends without error (depending on when I have a breakpoint)

Finally, I'm able to alleviate the Exception Symptoms by first calling 
"connector.stop()", which is a method on Jetty's "Lifecycle Interface".

According to Jetty's JavaDoc,

void stop()
          throws Exception

    Stops the component. The component may wait for current activities to 
complete normally, but it can be interrupted.

It appears to me, therefore, that calling "stop()" before close() in the 
"JettyHttpServerEngine.stop()" method will coordinate the thread ends better, 
quelling the exception.


I don't have a use case for restarting Jetty Server, but, in 
JettyHttpServerEngine.stop() , it calls "server.destroy()" after 
"server.stop()".  My guess would be that the server would subsequently be 
unusable because of this call.

Hope this helps.


> CXF http-jetty transport do not call the jetty server engine shutdown when 
> the server stop
> ------------------------------------------------------------------------------------------
>
>                 Key: CXF-865
>                 URL: https://issues.apache.org/jira/browse/CXF-865
>             Project: CXF
>          Issue Type: Bug
>          Components: Transports
>    Affects Versions: 2.0
>            Reporter: willem Jiang
>
> It appears that stopping the server in CXF is not shutting down the Jetty 
> engine, which causes my application to hang because Jetty's threads are still 
> blocking on the server socket accept. After a shutdown, I still see Thread 
> [btpool0-0 - Acceptor0 [EMAIL PROTECTED]:10001] (Running).
> I have a simple test application pasted below that can reproduce the problem. 
> At this point I had to add some calls to get the destination and cast it to a 
> Jetty specific class to get things to exit. Is there a better way to do this? 
> Is there something I'm not configuring correctly? Any help is appreciated.
>  
> This is under Java 1.5 with CXF 2.0. Jetty is being included from the CXF 
> libs directory. 
>  
> package org.mpilone.cxftest;
>  
> import java.io.IOException;
> import org.apache.cxf.endpoint.Server;
> import org.apache.cxf.frontend.ServerFactoryBean;
>  
> public class CxfJettyTest
> {
>   public void doIt()
>   {
>     System.out.println("Running test method");
>   }
>  
>   public static void main(String[] args) throws IOException
>   {
>     // Create an Service and Server
>     ServerFactoryBean serverFactory = new ServerFactoryBean();
>     serverFactory.setServiceClass(CxfJettyTest.class);
>     serverFactory.setServiceBean(new CxfJettyTest());
>     serverFactory.setAddress("http://localhost:10001/RemoteApi";);
>     Server mServer = serverFactory.create();
>  
>     mServer.start();
>     System.in.read();
>     mServer.stop();
>  
>     // Adding these lines allows the application to exit, but
>     // WARNING: EXCEPTION
>     // java.nio.channels.ClosedChannelException
>     // log statements are produced.
>  
> // JettyHTTPDestination jettyDest = (JettyHTTPDestination)
> // mServer.getDestination();
> // JettyHTTPServerEngine jettyEngine = (JettyHTTPServerEngine)
> // jettyDest.getEngine();
> // jettyEngine.shutdown();
>  
>     System.out.println("Exiting");
>   }
> }
> If you looking into the code you will find there is a note in the 
> JettyHTTPServerEngine's removeServant(URL url)
>             /* Bug in Jetty, we cannot do this.  If we restart later, data 
> goes off
>              * someplace unknown
>             if (servantCount == 0) {
>                 try {
>                    ....
>             }*/
> I just added the shutdown engine's code below the comments and tested it 
> against the systest. There were some  tests (such as ws.rm , mtom )failed.
> It looks like if we shutdown the engine, when the engine start again , the 
> engine can not get the request info again , specially the http keep alive 
> message.
> Current I can't reproduce this bug by adding the unit test which simply calls 
> start and shutdown method the JettyHTTPServerEngine.
> It may take some time to resolve this issue. 

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to