Yes, they both hit the same server. Whichever one is first will fail maybe 20% of the time while the second one will never fail. If I reverse the order, the first will continue to have issues and the second will work fine. From: Philippe Mouawad <[email protected]> To: JMeter Users List <[email protected]> Sent: Monday, April 27, 2015 2:46 PM Subject: Re: HTTP sampler implementation methods and "target server failed to respond" errors Hi, JAva implementation retries many times by default hiding some failures. See: https://wiki.apache.org/jmeter/JMeterSocketClosed
But what you report seems strange as I understand that if you swap the order then it will not be the same sampler that will fail. But do they both hit the same server ? Could you extract a simple test plan containing those 2 samples with potential timers and open a bigzilla just so that we check ? Thanks On Monday, April 27, 2015, Glenn Caccia <[email protected]> wrote: > I have a JMeter script that includes some HTTP samplers going to > different servers. Two of the HTTP samplers are going to the same server > and represent a small amount of the total traffic getting generated by the > script. The two samplers are only generating around 2 HPS for their target > server. Excluding these two samplers, the script is not sending any other > traffic to this server. Despite the low load, the first of the two > samplers will start generating "target server failed to respond" errors > after a minute or so of running and gradually increase in frequency. If I > swap the order of the samplers, the first one will generate the errors and > the second one will never generate errors. After doing some research, I > changed the implementation method to "Java" and now I don't get any > errors. Is this expected? I've never toyed with the implementation > setting before. Any thoughts on why HttpClient4 would be problematic but > Java would work fine? I'm using JMeter 2.13. > Thanks, > Glenn -- Cordialement. Philippe Mouawad.
