6 Replies Latest reply on Dec 15, 2010 7:12 PM by joe.marques

    HTTP Response Time setup in JON WANTED

    Benson Fung Newbie



      I would like measure the HTTP Response Time of a web application(war).  I have enabled it under the schedule tab.  However, it indicates that the type is calltime, what is this about?  I tried to hit the response time tab to see if there is anything displayed, it shows nothing in the screen.

      Is it possible to display HTTP response time in Graph?   Please help teach me how to setup.




        • 1. Re: HTTP Response Time setup in JON WANTED
          mazz Master

          Here's info on Response Times (thought the screen snapshots are old):




          I don't think the latest developer release supports showing response times in the GWT user interface. You have to use the JSF user interface to see it. To get to the JSF UI, go to your server at http://localhost:7080/Start.do and navigate to your resource's Monitor tab via the menus.

          • 2. Re: HTTP Response Time setup in JON WANTED

            I believe the response time stuff *is* available in the new UI (currently implemented as an iframe to the old JSF page with that information).


            Benson, you went to the right place to see the response time metrics, but you might not have enabled collection of that data on the remote box you're collecting from.  Have you set up the response time filter inside your web application? -- http://rhq-project.org/display/RHQ/Response+Time+Filter

            • 3. Re: HTTP Response Time setup in JON WANTED
              Benson Fung Newbie

              John : I have enabled the HTTP response time.  However, when I clicked the response time in the monitor tab, the screen is empty.

              Joseph : I didn't setup the Response time filter.  Could you elaborate for more?   Thanks

              • 4. Re: HTTP Response Time setup in JON WANTED

                Benson, in order to start aggregating things like min/avg/max URL timings, we need to capture data about your websites.  We do this by way of a servlet filter, which you "instrument" your web application with, which collects the necessary data for us.  Once you add the necessary pieces to your WAR and restart it, the metrics should start flowing.


                Note: RHQ's portal.war is instrumented out-of-box, which is why you can do to the response time for RHQ-specific resources that support it (WARs and EJBs) and see that data.

                • 5. Re: HTTP Response Time setup in JON WANTED
                  Benson Fung Newbie

                  Thanks Joseph.  But a performance question will be asked once installed the servlet filter into the web server.  I know some customers have concern that we put something into their application server in order to measure the application performance, but does the filter itself give impact the application server performance?  I have been asked for many times.

                  • 6. Re: HTTP Response Time setup in JON WANTED

                    No one has reported the filter having an effect in their environment to date...to my knowledge.  However, if the load on the web application is high enough, it theoretically could have a performance impact on it.  The RtFilter uses a BufferedWriter and only flushes if certain thresholds are met (time since last flush or number of requests...both of which are configurable via init params to the filter), so that could add to the round trip time of the requests that result in rt_log flushes.  Also, the internal bookkeeping information that RtFilter uses is synchronized, so  that could be seen/felt as request delays as the number of simultaneous requests increase.  Looking at the code, it could be refactored to use more intelligent locking which would reduce the amount of logic that needs to be strictly serialized (improving access in highly concurrent environments), but considering no one has seen an effect from including the existing filter in their environments, it would likely get low priority.