tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ivan Larionov <>
Subject Re: Tomcat 7, JAX-WS 2.2.5
Date Fri, 18 Nov 2011 09:41:19 GMT
Hi Konstantin,

Thank you for your reply.

Please see inline below.

On 17.11.2011 11:57, Konstantin Kolinko wrote:
> 2011/11/14 Ivan<>:
>> Dear All,
>> We are running JAX-WS web-services using JAX-WS RI under tomcat. First we
>> upgraded from Tomcat 5.5 to Tomcat 7 and everything was fine. Eventually, we
>> decided to upgrade to the latest JAX-WS RI and went for 2.2.5 version.
>> We faced a problem. Upon requests we started seeing exceptions from JAX-WS
>> indicating that the body of the request was broken. From the point of view of
>> TCPMON everything was fine. We enabled request tracing and saw that the JAX-WS
>> RI was gettting empty requests' input streams which were causing mentioned
>> exceptions (org.jvnet.mimepull.MIMEParsingException: Missing start boundary;
>> com.ctc.wstx.exc.WstxEOFException: Unexpected EOF in prolog).
>> Also, we tried tomcat 5.5 and JAX-WS RI 2.2.5 and they worked together fine.
>> After investigation we found out that all versions of JAX-WS RI since 2.2.2 were
>> causing the same problem. Presumably, the cause of the issue is servlet 3.0
>> async request processing which has been added in JAX-WS RI 2.2.2. Partly, it
>> explains why tomcat 5.5 works fine with JAX-WS RI 2.2.5 as the RI falls back to
>> non-async request processing.
>> We changed connector's protocol to 'org.apache.coyote.http11.Http11NioProtocol'
>> and everything started working.
>> As I understand, both BIO and NIO connectors should be working well with servet
>> 3.0 async. Could you comment on this?
> 1. What exact version of Tomcat 7.0.x ?
Tomcat 7.0.14
> 2. Yes, both should be working, though implementations are
> substantially different.
> I would expect Nio to be better with async handling iт terms of performance.
> 3. Bugs might be there, but without a specific test case there is
> nowhere to start.
Currently, I cannot provide a test case. The problem manifests itself 
somewhere in the JAX-WS RI execution path, not in our code. I believe 
our management will settle for just disabling async.
> 4. Consider running with
> org.apache.catalina.connector.RECYCLE_FACADES=true
> That option will make your configuration more secure and can reveal
> severe bugs with request/response management in web applications (and
> frameworks).
Thanks, I'll try it.
> Best regards,
> Konstantin Kolinko

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message