Closed ritazh closed 5 years ago
@ritazh
in the original config(10Mi
), the pod crashed a lot and then would consume lots of CPU/memory?
@andyzhangx We only experienced this issue when running on GPU nodes. I think the nvidia-docker runtime is consuming more memory.
Reason for Change:
Users reported issues deploying the daemonset to an AKS cluster. This increases the resource requests and limits to ensure the pods are created properly on each node.
Issue Fixed:
Notes for Reviewers: