How do I ensure all nodes are running at the same time in K8S for jobs with parallelism

Question

I need to run a job with parallelism, but I need all nodes/pods running at the same time. If I have only 4 nodes available, but I need 5 then they need to all remain pending or