I havent seen any errors, so i'm not sure that i'm running out. I just did a daytime testrun with keepalives enabled and the number of time_waits was much improved so i'll let that run tonight. Fortunately the max test that i run is just within the MaxThreads parameter in apache.
I also discovered I have a ulimit soft descriptor limit of 256, not sure if that is causing issues, so i've upped it to see if it makes any difference. I also re-profiled JMeter and there is no thread blocking, memory is well within the limits, peaking at about 200mb out of an initial size fof 500mb, max 1gb and cpu all looks ok too. even garbage collection seems to be having little impact. I'll re-analyse the results tomorrow and see if these changes have made any difference. here's hoping! --------- Original Message -------- From: sebb <[EMAIL PROTECTED]> To: JMeter Users List <[email protected]>, Lists <[EMAIL PROTECTED]> Subject: Re: Time spent where? Date: 30/01/08 11:32 > On 30/01/2008, Lists <[EMAIL PROTECTED]> wrote: > > Ok, we have the network stats directly from the switch. > > > > The port for the script server peaks at about 4% utilisation :) > > > > So i dont think it's network contention! > > > > I have noticed we do get up to 20,000 entries for TIME_WAIT in the netstat > > list, not sure if thats significant. > > > > http://wiki.apache.org/HttpComponents/FrequentlyAskedConnectionManagementQuestions > > I'm not sure what happens if the JMeter host runs out of ports - does > it report an error immediately, or does it time-out? If so, how long > is the timeout? > > Might be worth experimenting with Keep-Alive - if enabled, the > HttpClient sampler should re-use the connection (though of course the > server might still close it). > > If it is possible, I guess it would be useful for JMeter to measure > connection time, and perhaps try to keep track of the number of > connections used in a thread. > > > --------- Original Message -------- > > From: JMeter Users List <[email protected]> > > To: JMeter Users List <[email protected]>, Lists > > <[EMAIL PROTECTED]> > > Subject: Re: Time spent where? > > Date: 28/01/08 16:55 > > > > > On 28/01/2008, Lists &lt;[EMAIL PROTECTED]&gt; wrote: > > > &gt; &gt; > > > &gt; &gt; How much longer? A few percent or much more? > > > &gt; &gt; > > > &gt; > > > &gt; significantly. an average of 89ms at the apache logs, and yet 140 at > > > > > > Does not seem too bad considering that the data has to be transferred > > > across the network. > > > > > > &gt; jmeter. The key point is though that this latency is pratically zero > > at > > > &gt; lower loads. The reason it causes us a problem is we define failure > > by > > > &gt; saying 95% of responses must be less than a given response time - so > > this > > > &gt; extra gap is seemingly causing us to fail the tests, but > > unnecessarily. > > > > > > Depends on the target of course. > > > > > > &gt; &gt; Are all samples affected? > > > &gt; > > > &gt; Err, good question, i've been looking at averages, so i dont know at > > this > > > &gt; point. I guess i should look at standard deviation perhaps..? > > > &gt; > > > &gt; &gt; Does the additional time depend on the sample size? > > > &gt; > > > &gt; Err, sample size? or throughput? yes it increases with throughput. > > > &gt; > > > > > > I meant the number of bytes downloaded. > > > > > > Bigger samples generally take longer to download. > > > Some conditions (e.g. network contention) may affect larger samples > > > more than smaller ones as the transfer time is greater as a proportion > > > of total time. > > > > > > &gt; &gt; Note that the elapsed time will necessarily be longer than the > > apache > > > &gt; &gt; log time; the JMeter latency should be closer to the apache log > > > &gt; &gt; timings. > > > &gt; &gt; > > > &gt; &gt; Are you using any functions in the samplers? Function execution > > is > > > &gt; &gt; single-threaded. > > > &gt; > > > &gt; I'm not sure, i'll double check. However i have profiled jmeter > > recently > > > &gt; and it didnt seem to be blocking - i do need to double check this, > > hopefully > > > &gt; i'll do that tomorrow. > > > &gt; > > > &gt; > > > &gt; &gt; &amp;gt; Thanks, > > > &gt; &gt; &amp;gt; Dan > > > &gt; &gt; &amp;gt; > > > &gt; &gt; &amp;gt; ________________________________________________ > > > &gt; &gt; &amp;gt; Message sent using UebiMiau 2.7.10 > > > &gt; &gt; &amp;gt; > > > &gt; &gt; &amp;gt; > > > &gt; &gt; &amp;gt; > > > &gt; &gt; &amp;gt; > > --------------------------------------------------------------------- > > > &gt; &gt; &amp;gt; To unsubscribe, e-mail: > > [EMAIL PROTECTED] > > > &gt; &gt; &amp;gt; For additional commands, e-mail: > > [EMAIL PROTECTED] > > > &gt; &gt; &amp;gt; > > > &gt; &gt; &amp;gt; > > > &gt; &gt; > > > &gt; &gt; > > --------------------------------------------------------------------- > > > &gt; &gt; To unsubscribe, e-mail: > > [EMAIL PROTECTED] > > > &gt; &gt; For additional commands, e-mail: > > [EMAIL PROTECTED] > > > &gt; &gt; > > > &gt; &gt; > > > &gt; &gt; > > > &gt; > > > &gt; ________________________________________________ > > > &gt; Message sent using UebiMiau 2.7.10 > > > &gt; > > > &gt; > > > &gt; > > > &gt; --------------------------------------------------------------------- > > > &gt; To unsubscribe, e-mail: [EMAIL PROTECTED] > > > &gt; For additional commands, e-mail: [EMAIL PROTECTED] > > > &gt; > > > &gt; > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > > > > > > > > ________________________________________________ > > Message sent using UebiMiau 2.7.10 > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > > ________________________________________________ Message sent using UebiMiau 2.7.10 --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
