0

I am generating a 1000 user load on an application hosted on AWS t2.large server from my 8gb windows machine. If I downsize my load to 500 i get no failed transactions but if I go with 800-1000 user load, I see the following exceptions:- 1. java.net.bindexception 2. java.net.socketexception 3. java.net.connectexception

Does these signify bottleneck at the server side or my system is not capable to generate that much load. Please also provide resolution steps for the same.

Thanks in advance for your help. Yash

yash
  • 23
  • 2
  • 6
  • is a connection pool not what you are looking for? https://stackoverflow.com/questions/2835090/how-to-establish-a-connection-pool-in-jdbc – Tschallacka Jul 06 '17 at 09:36

1 Answers1

0

Most likely you found the bottleneck in your application, I would recommend the following steps:

  1. Check your web or application server connection settings, HTTP servers use to have maximum allowed connections setting and it needs to be more than 1000.
  2. If point 1 is fine you could monitor your AWS instance health via Amazon CloudWatch or JMeter PerfMon Plugin as your server may simply lack CPU or RAM or ran out of available file descriptors/network sockets (it is around 1024 for "normal" Linux user account)
  3. If points 1 and 2 are fine and you are totally sure your application behaves as expected you could also try workarounds listed in Connection Reset since JMeter 2.10 ? JMeter wiki page
Dmitri T
  • 159,985
  • 5
  • 83
  • 133
  • Thanks Dmitri for providing the resolution steps. I was monitoring the server for CPU and Memory by sar and mpstat commands which was under the acceptable limit throughput the test duration. So I feel resolution step 2 could not be the issue...I did a work around with step 3 with which the error rate was significantly less when compared with the earlier result. Will update which one worked best for me as I am yet to configure the test with modification in max and min thread loops.... Thanks again..:) – yash Jul 06 '17 at 13:56