skypilot-users
Is there a development branch for doing multinode batch jobs on k8s?
An
Andrew Aikawa (asaiacai)
Asked on Sep 23, 2023
Yes, there is a development branch for doing multinode batch jobs on k8s. Some changes that need to be made include creating a headless service for the head node pod, using the service IP on the workers to connect to the head node pod, and potentially modifying the scheduling code to use service IPs instead of pod IPs.
Sep 23, 2023Edited by