Incorrect graph data for Active Users over a long time period

Make sure these checkboxes are checked before submitting your issue. Thanks you!

• [*] Don’t post questions here. This bugtracker is for issues and feature requests only. Use the community mailing list instead.
• Regarding issues:
  • [*] Check that you can reproduce your issue with latest Gatling version
  • [* ] Search this bugtracker for any similar issue that might have already been reported (and possibly fixed and available as a SNAPSHOT version)
  • [*] Provide a way for Gatling developers to reproduce your problem, e.g. step-by-step instructions against a public facing website, or a sample application that demonstrates your problem.
• Regarding feature requests:
  • [*] Explain your use case, not the implementation you have in mind.

I have a script that should ramp up 5 users in a single second, with a 10 second gap in between, over a 2 hour period. Doing the math out, that should be about 3600 users.

However, the graph is telling me that only 1 user was ramped up every 10-20 seconds, but ~3300 users were ramped up.

If there really was only 1 user ramped up every 10-20 seconds like the graph shows, there should only be a max of 720 users ramped, so there is an innacuracy between the graph and the number of users called. Has anyone else ran into a similar issue?

Example call (does nothing for 2 sec to wait for an OAUTH token):

val synchronousScenario = List(
    OAuthRequest.inject(atOnceUsers(1)),
    endpt.inject(nothingFor(2 seconds),
      splitUsers(3600) into (5) over (1 seconds)) separatedBy (10 seconds)));

  // Kicks off the scenario
  setUp(synchronousScenario).protocols(httpConf).maxDuration(2 hours);

When I halt the script early after maybe 2 minutes and use Gatling command line to generate a report from the simulation log, the graph actually looks accurate, showing 5 users ramping at once with approx 10 seconds in between, so it seems something weird happens to the graph when it’s left uninterrupted for a long period of time.

Gatling version 2.3.0 Scala maven plugin version 3.2.2

[slandelle]

I see the number of injected users in the statistics chart that shows 3274 users with OK status.

OK, so the stats in the table are, as column header states, Requests, not users.

[slandelle]

But why would the graph itself not show 5 users being ramped up at once, unless I halt it early?

Because this stats is not “User arrival rate” (available in FrontLine btw), but “Active users”, which is computed as:

number_of_users_alive_bucket(n)
- number_of_users_alive_bucket(n-1)
- number_of_users_dead_in_bucket(n-1)

The larger the time windows, the larger the buckets (IIRC, we compute 1000 plots by default), which can cause variations.