We use Kubernetes cluster build infrastructure. Right now the lite-engine reserves CPU based on the step limit (step limit + add-on), so each build pod reserves ~3 CPU while it actually uses under 1. That fills our build pool on reservations and pods fail to schedule at peak. We want to set step resource requests lower than the limits — pack pods denser while keeping the burst limit for the steps.

Per your Resource Allocation docs, defining per-step resource requests needs this flag enabled by Support.

Thanks.