Scaling out

slandelle edited this page Mar 27, 2012 · 2 revisions

Sometimes, generating some very heavy load from a single machine might lead to saturating the OS or the network interface controller.

In this case, you might want to use several Gatling instances hosted on multiple machines.

Gatling doesn't have a cluster mode yet, but you can achieve similar results manually:

  • deploy Gatling on several machines along with the Simulation classes and the associated resources (data, request-bodies, etc..)
  • launch them remotely from a script, with the -nr (no reports) option
  • retrieve all the simulation.log files and merge them:
    • edit the RUN records so that you leave only one in the merged result
    • place the merged result into a folder in the results folder of a Gatling instance
  • generate the reports with Gatling with the -ro name-of-the-simulation-folder(reports only)