Hi all,
Some times ago, we discussed about observed corruptions of SOAP requests under load, see thread:
Corrupted XML (SOAP responses) under load
http://forums.java.net/jive/thread.jspa?threadID=63273
The latest tests I have made allowed me to define the specific case when I have that kind of problems.
To reproduce with a failure ratio around 33%, I just need to force the pipelining of my SOAP requests. That means that I open several connections to a web service deployed on GlassFish then I send several SOAP requests on each of these connections without waiting for any response.
The first SOAP responses received - those corresponding to the first SOAP requests sent on each connection - are OK.
Most of the second SOAP responses are KO.
Almost all of the other SOAP responses are KO.
With enableSnoop=true, I see logs containing:
[#|2009-07-08T13:49:36.583+0200|INFO|sun-appserver2.1|javax.enterprise.system.container.web|_ThreadID=26;_ThreadName=httpSSLWorkerThread-8080-2;|SocketChannel headersjava.nio.channels.SocketChannel[connected local=XXXX/XXXX:8080 remote=/XXXX:55403] are: === MimeHeaders ===
[i]XML content[/i] </soapenv:Body>
</soapenv = Envelope>POST
http://XXXX:8080/service HTTP/1.1
content-type = text/xml;charset=UTF-8
soapaction = ""
user-agent = Jakarta Commons-HttpClient/3.1
host = localhost:8080
content-length = 986
This is observed with GlassFish v2.1 build 60e.
I have tried to reproduce this with GlassFish v2.1.1 build 21 but failed to do so.
What has been changed in this new build (or anyother one) that can explain this difference of behaviour ? Any issue fixed ?
I have not found anything related in the bug tracker.
Thanks for any information about this.
Bye
[Message sent by forum member 'vkoniecz' (vkoniecz)]
http://forums.java.net/jive/thread.jspa?messageID=354643