0

In short, the exact same problem posted here 3 years ago.

Running with 20 workers or 4 workers, the results are peculiar in the UI. They seem to reflect per-server metrics, and not the overall metrics of all workers. I request 500 users, the number of users displayed is only 267 (roughly half). The RPS is around 600.

What's odd, is if I generate a report, the RPS shows the "real" value of 1100.

Unfortunately in my case, the master server is standalone and doesn't run a worker at all.

Any recommendations for troubleshooting this kind of issue?

Jacob Steinberger
  • 1
  • Possibly a red herring; I did find that the master process gets confused if workers are added/removed between tests. Even if a worker thread no longer is running, the core process still lists it, and still expects it to "spawn" and accept its load. It seems like the master process needs to be restarted if there are changes in the worker states. – Jacob Steinberger Jul 16 '23 at 15:28
  • If you can reproduce this (maybe using a script, and a simple locustfile), please file a bug. – Cyberwiz Jul 16 '23 at 16:24

0 Answers0