Hi, this may be a silly question, but I haven't been able to really find an answer: does the K8s operator autoscaler also spin up additional task managers so there's enough task slots or does it somehow expect that a job definition already has enough task slots for optimal parallelism? Maybe it's a fundamental misunderstanding on my part?

For a session cluster it’ll auto scale up from what I’ve seen

it'll automatically apply for more tms if needed for both application/session mode with active resourcement (aka yarn or native kubernetes). For standalone clusters, you have to manually add up tms

Thanks!