Hi, thanks for give it a go. I've now got some good Ethereal dumps for anyone to look at to try and identify the problem. They follow in a seperate mail after this one due to the size restrictions.
Firstly to address your questions: I'm on a Win98 (/shudder) Java 1.3.1 based workstation, not my choice of machine... it BSOD'd me while I was writing this mail first time around! The server on the other hand is a new Solaris box with J2EE 1.3.1 running Apache2 with Tomcat and OpenSSL. I've attached 3 dumps. One is from a navigation using Mozilla Firebird, the other 2 are from an identical navigation using JMeter. One of the Jmeter ones is with Follow Redirects enabled, the other without. It's note-worthy that with following redirects enabled, JMeter reported times of over 1000 billion milliseconds, which is just gobbledegook as it still took 15 seconds to my count. If we look at the dumps there something very obvious thats wrong. When doing a GET on the 'Plan Page' (/planrecord/servlets/PlanServlet), Mozilla Firebird and all other broswers (and Raw communication over telnet) receives a '200 OK' response. JMeter on the other hand, despite doing an identical navigation, gets a 302 redirect with the location the same as the request (/planrecord/servlets/PlanServlet). To see this, look at packets 71-72 on the FirebirdNav.dump, and then compare it to 29-30 and 30-31 in JMeter-NoFollowRedirects.dump and JMeter-WithFollowRedirects.dump respectively. Why does JMeter get a 302 when all other browsers get a 200 OK? This is the heart of the problem because upon receiving the 302, with redirect following enabled or not, JMeter just sits and waits for the Timeout to arrive (15 seconds as it says in the header) without GETting any more data. Why is this? Am I missing a very obvious client/server setting that will stop JMeter from doing this? Any help much appreciated guys, Cheers, Duncan Frostick Jordi Salvat i Alabart wrote: > Hi Duncan. > > I tried to reproduce this locally, but the script performs flawlessly. > Since I don't have access to your server, I just created "fake" > responses (mimicking yours where I know the details) on my server. > Didn't help: I still get very low measurements (2 ms for the "Plan Page" > 302 response). > > Questions: > - Does the 302 response for the "Plan Page" request have a large body? > - Which O.S. & JDK version are you using? > - Could you get a tcpdump or ethereal log of the JMeter/server interaction? > > -- > Salut, > > Jordi. > > Duncan Frostick wrote: > > Just tried with 1.9, same problem. > > > > Seems to have something to do with redirects (HTTP code 302)... but I'm not sure. > > > > > > [EMAIL PROTECTED] wrote: > > > > > >>Never seen anything like that - what does your test look like, exactly? Which > >>version of JMeter are you using? Did you try recording using JMeter's proxy, > >>and if so, did you get the same delay there? > >> > >>-Mike > >> > >>On 7 Aug 2003 at 12:20, Duncan Frostick wrote: > >> > >> > >>>Hi, > >>> > >>>I need to stress test a server running Java Servlets, JMeter should be > >>>ideal. With these Servlets, there is a very standard login/out system in > >>>place for users of the system. For example, to get to any of your > >>>records on the server, you must Log in though LoginServlet from which > >>>you can then go to the other servlets. Very standard stuff. > >>> > >>>However, with JMeter, with even the most simple test plan, a login > >>>consistently takes over 15 seconds. The login is done exactly as the > >>>browser does it (verified using BadBoy to capture everythng and then > >>>exported to JMeter), but JMeter sits and waits after posting the login > >>>information for over 15 seconds - and it just does not make any sense. I > >>>haev also tried without the BadBoy generated file, it threw up exactly > >>>the same problem. > >>> > >>>It's not the servers at fault. It's running Apache 2 on Solaris. This 15 > >>>second delay isn't replicable in any browser. Additionally, I captured > >>>the packets JMeter was sending and resent them raw over telnet exactly, > >>>but there was no 15 second delay. It seems to be something totally > >>>internal to JMeter. > >>> > >>>How can I address this? JMeter is my only option for stress testing, but > >>>this 15 second delay on *every* login is rendering it useless. Any > >>>ideas? > >>> > >>>Thanks for you time, > >>> > >>>Duncan Frostick > >>> > >>> > >>>--------------------------------------------------------------------- > >>>To unsubscribe, e-mail: [EMAIL PROTECTED] > >>>For additional commands, e-mail: [EMAIL PROTECTED] > >>> > >> > >>-- > >>Michael Stover > >>[EMAIL PROTECTED] > >>Yahoo IM: mstover_ya > >>ICQ: 152975688 > >>AIM: mstover777 > >> > >>--------------------------------------------------------------------- > >>To unsubscribe, e-mail: [EMAIL PROTECTED] > >>For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
