I’m looking to understand how Cloud Run autoscales services with multiple containers.

Does it use the run.googleapis.com/container/cpu/utilizations metrics which seems to be an aggregate of both containers or the run.googleapis.com/container/cpu/usage which is defined for multi containers?

My setup has a small metric collector which uses very little cpu vs my application. I want the autoscaling to kick in when my application is struggling

I have had a read through the docs but there is no mention of which metrics is used.