I am using JMeter to do some mild load testing. I've started with a CentOS installation at Digital Ocean.
I run X virtual users in a span of 100 seconds. When I ramp X up any higher than 200, I get the connections after 200 (or thereabouts) refused. The server is still running. Without doing anything to it, I can connect again in another minute or so.
If I run JMeter on the VPS (droplet in Digital Ocean parlance), it runs fine ... until I ramp up high enough to get Java memory problems in JMeter. If I run JMeter on a VPS beside the target, I get the same errors as I get from my PC.
So it has something to do with some networking ... something. I opened a ticket with Digital Ocean and they say it's not an intermediate device they have that is causing this.
So, what that commonly installs with CentOS would be shutting this off? I've tried running with IPTables stopped and that hasn't changed anything.