Memory consumption of runners in Kubernetes mode

[DenisPalnitsky]

I'm load testing Kubernetes mode 240-360 jobs running simultaneously with a target to support ~1000 jobs.
Currently two pods are being created for one job "runner" and "runner-worflow". Workflow pod does all the job and runner is only responsible for "managing" the workflow pod. I was surprised to learn that runner pod could reach up to 1Gb which seems a bit excessive considering it's role.

I can speculate that it's due to Node that is used for running hooks so I wonder if rewriting hooks to go would make it more efficient.

[mikemarquez]

Is this still relevant? I am seeing similar resource usage in our system as well.