jakarta-cactus-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pavan Aripirala Venkata" <pvenk...@h5technologies.com>
Subject FW: Cactus with load balancing in a clustered environment - Exception.
Date Thu, 20 Feb 2003 18:25:43 GMT
>>>The one with the unique id?

Yes we return an object containing a unique Id (sourceName). In textXXX
method we do this:
	Result result = new Result(message, execTime, sourceName);
      OutputStream out = response.getOutputStream();
      ObjectOutputStream oos = new ObjectOutputStream(out);

      oos.writeObject(result);
      oos.flush();
      oos.close();
We are getting back the results as expected - I hope we are doing it
correct.
Thank you

Pavan.

-----Original Message-----
From: Vincent Massol [mailto:vmassol@octo.com]
Sent: Thursday, February 20, 2003 10:13 AM
To: 'Cactus Users List'
Cc: Pavan Aripirala Venkata
Subject: RE: Cactus with load balancing in a clustered environment -
Exception.




> -----Original Message-----
> From: Pavan Aripirala Venkata [mailto:pvenkata@h5technologies.com]
> Sent: 20 February 2003 18:48
> To: Cactus Users List
> Subject: RE: Cactus with load balancing in a clustered environment -
> Exception.
> 
> 	Thanks Vincent, I did not know that there were two HTTP calls
> during a cactus test run.
> We did take care of the first issue which you mentioned.

The one with the unique id?

> 	Also this "Missing service name parameter" appears during the
> first HTTP call right?
> 	I wonder why the "Missing service name parameter" is reported as
> a warning and not as an error.

Not, it's not normal and it is an error (not sure why you get this error
though). I don't know why the container does display it as an error (The
Cactus code throws a ServletException to the container). It has to be a
JBoss thingie.

-Vincent

> Thanks
> 
> Pavan.
> 
> -----Original Message-----
> From: Vincent Massol [mailto:vmassol@octo.com]
> Sent: Wednesday, February 19, 2003 12:07 AM
> To: 'Cactus Users List'
> Cc: Pavan Aripirala Venkata
> Subject: RE: Cactus with load balancing in a clustered environment -
> Exception.
> 
> 
> Hi Pavan,
> 
> There is one issue with Cactus that will prevent your setup from
working
> fine. ATM, it is not possible to run multiple concurrent cactus tests.
> The reason is simple and easy to fix: each cactus test result is put
in
> the Servlet Config (application scope) under a fixed key name. Thus a
> second test might overwrite the result of a first test before the
first
> test returns.
> 
> The solution is to add some unique ID to each test. The ID should
> probably be generated on the Cactus client side, passed to the
> redirector as another cactus parameter. The redirector will save the
> test result using this id and the client side will issue the request
to
> get the test result using this id.
> 
> There is another potential problem: The cactus client side issues 2
> requests to the redirector: once to run the server side tests and
> another one to get the test results. ATM, the second HTTP call does
NOT
> use the same HTTP session as the first HTTP call (actually, it's even
> possible that that there is no HTTP session created on the first
call).
> Thus, with you load balancer, the second call may end up on another
> application server instance and thus the test result that will be
> returned will not be the correct one.
> 
> Thanks
> -Vincent
> 
> 
> > -----Original Message-----
> > From: Pavan Aripirala Venkata [mailto:pvenkata@h5technologies.com]
> > Sent: 18 February 2003 19:43
> > To: cactus-user@jakarta.apache.org
> > Subject: Cactus with load balancing in a clustered environment -
> > Exception.
> >
> > 	Hi all,
> > 	I have a clustered environment wherein I have our application
> > running on three application servers (JBoss 3.0) (on three different
> > machines), and we have an Apache load balancer (with sticky
sessions)
> > for load balancing these machines. With this setup I run some of my
> > cactus tests and sometimes the following exception is occurring -
the
> > occurring of this exception is erratic.
> >
> > 2003-02-13 16:36:44,313 WARN  [org.jboss.jbossweb] (Logger.java:167)
> > WARNING: jva.io.IOException: Premature EOF reading params
> > 2003-02-13 16:36:44,322 WARN  [org.jboss.jbossweb] (Logger.java:167)
> > WARNING: Eception for /cactus/ServletRedirector/
> > javax.servlet.ServletException: Missing service name parameter
> > [ServletTestRediector_Service] in HTTP request.
> >         at
> >
>
org.apache.cactus.server.AbstractTestController.getServiceName(AbstrctTe
> > stController.java:176)
> >         at
> >
>
org.apache.cactus.server.AbstractTestController.handleRequest(AbstratTes
> > tController.java:114)
> >         at
> >
>
org.apache.cactus.server.ServletTestRedirector.doPost(ServletTestRedrect
> > or.java:134)
> >         at
> javax.servlet.http.HttpServlet.service(HttpServlet.java:760)
> > and so on
> >
> > 	I don't get this exception when I run the cactus tests with just
> > one single application server (no Apache in between), i.e., the
> regular
> > way. Also I am able to test my application from the web pages with
> > clustering on, without any exceptions.
> >   	Looks like the request parameters are not reaching the app
> > server. You might have seen this if you had just typed
> > http://<your host>:<deployport>/cactus/ServletRedirector/ in the web
> > browser.
> >
> > 	In cactus.properties I set the servletRedirectorURL (or the
> > contextURL) to the ipaddress of the machine where Apache resides. I
> use
> > Java threading (home built code) to run multiple concurrent cactus
> tests
> > - I am not using the JUnit's ActiveTestSuite. I also make sure that
> the
> > input data is not same for two threads. I am successful with this
> > threaded setup when there is no clustering.
> > 	JBoss internally uses jetty. I don't know if that causes any
> > error. Long ago I saw an posting on this list about jetty and cactus
> > problem - I am sure these two are not related. (FYI the subject of
> that
> > e-mail was Re: [jetty-support] Re: Cactus and Jetty problem - sent
by
> > Scott Sayles [ssayles@fgm.com]).
> > 	Please, let me know if anyone knows how to fix it.
> > Thanks
> >
> > Pavan.
> >
> >
> >
> >
---------------------------------------------------------------------
> > To unsubscribe, e-mail: cactus-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: cactus-user-help@jakarta.apache.org
> 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: cactus-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: cactus-user-help@jakarta.apache.org
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: cactus-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: cactus-user-help@jakarta.apache.org



Mime
View raw message