Sebb, I think there is a good chance that you are half write. I have investigated the issue of running out of sockets on a windows server. The machine can run out of sockets, but it does this in a controlled way. However, I suspect that under very heavy concurrent load windows fails to send socket closed messages properly. I have no real proof of this (ie I have not traced the socket chatter) as - frankly - I don't have a machine powerful enough to monitor that much traffic. However, Windows not closing sockets properly fits all the hanging results I have seen.
Cheers AJ On Fri, 2007-03-02 at 09:50 +0000, sebb wrote: > Just realised that there is another possibility that might be > affecting testing with lots of threads: > > The host might be running out of sockets or having problems releasing them. > > When the hang occurs, try starting a browser on the same host and see > if you can connect to the server under test and/or some other server. > > S/// > On 02/03/07, Marlon Santos <[EMAIL PROTECTED]> wrote: > > > > Ok thanks... I will be trying this and will be keeping you posted... > > > > > > > > > > sebb-2 wrote: > > > > > > If it is JMeter, then we will either need a repeatable test-case to > > > investigate further, or someone who is experiencing the problem may be > > > able to gather some further data about the state of JMeter: e.g. are > > > there any deadlocks? > > > > > > In answer to Marlon, I suggest you try: > > > - removing all unnecessary Listeners. You should only need one. > > > - running in non-GUI mode > > > > > > Both of these reduce the load on JMeter. > > > > > > If that does not solve the problem, then try using multiple JMeter > > > instances. If you then still get hangs, it is unlikely to be JMeter > > > that is at fault. > > > > > > S/// > > > On 01/03/07, Alexander J Turner <[EMAIL PROTECTED]> wrote: > > >> Nino, > > >> > > >> This is an issue that has cropped up in the past. In my experiance, > > >> JMeter (and I am not convinced it is JMeter's fault - it could be to do > > >> with Java) hangs when there are a lot of threads. I have seen this when > > >> I know it is not the web server that is hanging, because I was > > >> load-testing TAG, a web server that has a radically different approach > > >> to the likes of Apache and IIS, and for which I could monitor all the > > >> internals. I have also seen this on Linux and Window XP. > > >> > > >> I think the real issue is that no software works well with very large > > >> numbers of threads. If you really want to run vast numbers of > > >> concurrent actions, you need a non-blocking approach. Note, Apache does > > >> not uses non-blocking, but it does thread pooling, which results in it > > >> actually not handling large numbers of concurrent activities, it just > > >> lets the OS queue them. > > >> > > >> The only real way to destruction test a web server is to have multiple > > >> machines hitting it at once, you cannot expect a single machine running > > >> multiple virtual clients to do the job. > > >> > > Unless the server does not handle multiple connections very well ...! > > > >> Best wishes > > >> > > >> AJ > > >> > > >> On Thu, 2007-03-01 at 09:42 +0100, Nino Wael wrote: > > >> > This is pretty interesting..! > > >> > > > >> > We also experience something like that. My best guess rigth now are > > >> that it's because the http server arent able to handle all those requests > > >> at the same time, and some of them does not get a response, hence the > > >> hanging thread. > > >> > > > >> > What happens in your scenario if you abort the test (ctrl+.) and run it > > >> again. Is it the exact same number of threads that are hanging? > > >> > > > >> > regards Nino > > >> > > > >> > ________________________________ > > >> > > > >> > Fra: Marlon Santos [mailto:[EMAIL PROTECTED] > > >> > Sendt: to 01-03-2007 02:30 > > >> > Til: [email protected] > > >> > Emne: Re: [Newbie] Thread won't die issue > > >> > > > >> > > > >> > > > >> > > > >> > Hi, > > >> > > > >> > We are also experiencing the same thing. Actually we are in an > > >> environment > > >> > where we need to load test a website but when I was running the test on > > >> > small number of virtual users it works fine. But later on, when we > > >> increased > > >> > the number of virtual users at some point in time the threads suddenly > > >> > stopped responding, or may I say the threads hang. We are in a tight > > >> > schedule so the faster this is resolved, the better it will be > > >> > appreciated... > > >> > > > >> > Regards, > > >> > Marlon > > >> > -- > > >> > View this message in context: > > >> http://www.nabble.com/Thread-won%27t-die-issue-tf3059035.html#a9241968 > > >> > Sent from the JMeter - User mailing list archive at Nabble.com. > > >> > > > >> > > > >> > --------------------------------------------------------------------- > > >> > 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] > > >> -- > > >> > > >> ========================================== > > >> Hard Core How-To's, Coding Techniques ETC. > > >> --- http://nerds-central.blogspot.com! --- > > >> Nerds Central welcomes nerds :-) > > >> ========================================== > > >> > > >> > > >> --------------------------------------------------------------------- > > >> 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] > > > > > > > > > > > > > -- > > View this message in context: > > http://www.nabble.com/Thread-won%27t-die-issue-tf3059035.html#a9263341 > > Sent from the JMeter - User mailing list archive at Nabble.com. > > > > > > --------------------------------------------------------------------- > > 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] > > > -- ========================================== Hard Core How-To's, Coding Techniques ETC. --- http://nerds-central.blogspot.com! --- Nerds Central welcomes nerds :-) ==========================================
