Ability to scale more workers with remote execute

[FerrisYu]

Hi,

I would like to know if it's possible to run more workers than the CPU # on a single machine. I was think remote Socket Servers is a way to go, but it doesn't seem to work the way I needed( Workers are run locally?).

What I am trying to do:
pytest -n auto --dist=loadscope --tx socket=<second test server ip>:8888 --rsyncdir tests

To get tests loadbalance between two test servers

Any suggestions how I could scale up the tests run in parallel?

[nicoddemus]

-n can receive any number, so you can pass any number of CPUs you like there.

[RonnyPfannschmidt]

@nicoddemus i believe the intent is to have this work for a ssh tx, to automatically use all cpus on the other side, currently we dont support hat

[FerrisYu]

Thank you guys for your response.
@RonnyPfannschmidt Just to clarify, my understanding is that it won't work even if I have 8 CPUs each for 2 servers, and I set -n 16, with -tx set.

[RonnyPfannschmidt]

@FerrisYu -n 16 is roughly like --tx 16*popen

the auto arg is not yet looking at how many cpus remotes have

you can use the * syntax to work it out on remtoes where you know the number of cpus
https://github.com/pytest-dev/pytest-xdist/blob/master/src/xdist/workermanage.py#L14
aka --tx 8*ssh=... --tx 8*ssh=