When are actors related to users stopped?


I’m new to Gatling, I have very simple scenario that runs for around 12hs, the test injects around 10 users per second.
At some point the Gatling instance crashes:

Uncaught error from thread [GatlingSystem-akka.actor.default-dispatcher-131]: 
GC overhead limit exceeded, shutting down JVM since 'akka.jvm-exit-on-fatal-error' is enabled for ActorSystem[GatlingSystem]
java.lang.OutOfMemoryError: GC overhead limit exceeded

A heap dump on my local machine trying to reproduce the issue shows:
( it could perfectly happen that this dump has nothing to do with the real issue, i will try to get a dump from pre prod env )

I wonder if could be that actors are never stopped so they are not garbage collected.

The target service works perfectly well during all the run and responds all request under 1s.


I have no idea what you do, but this really doesn’t look like standard Gatling.

You have 2M instances of an akka-http connection pool (see Suspect 4).
But Gatling doesn’t use akka-http, our HTTP module is built on top of Netty!!!