Gatling overload

Hi ,

I am running a scenario which consist of sending a simple json object to the server and wait for success message. The load is increased every 1 minute, each time increasing 30 users per second as follows (the test takes 35 minutes):

constantUsersPerSec(1) over (60),
constantUsersPerSec(30) over (60),

constantUsersPerSec(60) over (60),

constantUsersPerSec(90) over (60)
.
.
.
constantUsersPerSec(990) over (60)
constantUsersPerSec(1020) over (60)

The first test run was done with the web application hosted on a server with two cores and 3.5GB of RAM. Up to 600 users per second everything worked fine but then the errors rather coming from the server (example “unsuccessful register”) were coming from gatling as follows:

I have already followed the OS tuning guidelines (inc. sudo sysctl -w net.ipv4.ip_local_port_range=“1025 65535”) but I still have this error. To make sure that the problem was not the application I repeated the test on a web server with 7 GB of RAM and 4 cores and it broke exactly at the same point. Do you think that I should generate load using more than one concurrent server? Or else do you have any idea of a way how to solve this problem?

I really appreciate the help. Thank you.

Jamie Tabone