How is traffic spread across the processes?

Hello, I have 9 web processes running, but only 4 seem to receive traffic? is soft_limit used (perhaps I need to lower this value).

Soft limit is used, yes. Soft limit is roughly “optimal number of requests to send to a process”. Lowering that limit should help! Soft limit is just a hint, though, we’ll send requests up to hard limit.

Which app is this, by the way?


Does this matter when we have processes e.g. both web and a background process per app, Im wondering if that is impacting things, and we should just split it out.

If we split the background process out, will autoscaling then start working, or will we need to recreate a new app?

If you revert to a single process group app, we can re-enabled autoscaling. It’s kind of a pain to run an app like this though, so I would recommend against it.

I noticed these are 2 CPU instances with 4GB RAM each. Can you get away with running more instances with 2GB RAM each? If you run shared-cpu-1x VMs with 2GB RAM, it’ll be quite a bit cheaper than what you’re doing now. You can keep the workers big if that helps.

1 Like

Hi Kurt,

Thanks yes I dropped the workers in half (so we can reduce the RAM use) and will change to the 1CPU.

1 Like