gpu-manager
gpu-manager copied to clipboard
tkestack
kubelet_internal_checkpoint not exist
gpu-manager daemonset start with reading gpu info from kubelet_internal_checkpoint, if the file is not exist, gpu-manager with stop start and log with an error. in my cluster(kubernetes 1.23), kubelet_internal_checkpoint only exist if the node has allocated at least one pod with third part device, such as nvidia/gpu. and i read the logic of kubelet, kubelet_internal_checkpoint will be created after pod allocated. so how to use gpu-manager without nvidia-gpu-device-plugin and additional initialization?
