How to spin up locust on multicore to avoid master failures

[sagar694]

I am using 1 locust master with 800 workers to generate load on different user flows and the script is written in a format to distributed test data across the workers. During the load tests, locust master CPU is breaching the memory/CPU thresholds and is bringing down and I suspect the primary reason for that is due polling various stats from this large pool of works with around 60 group names.

is there anyway to spin up locust master process on multicore to make it much more reliable ?

[cyberw]

Hi @sagar694 !

First of all: Do you really need 800 workers? Do you run this distributed on 800 CPU cores? If not, you should reduce the number of workers to match your core count, that will make things a lot easier for the master.

How many Users are you running?

If you are really doing massive tests where you need 800 worker processes, send me some logs and locustfile examples and I can take a look.