Skip to main content

Install Using YAML Files

Install CRDs

From project root directory, run

kubectl apply -f config/crd/bases/

Install KubeDL controller

A single yaml file including everything: deployment, rbac etc.

kubectl apply -f

KubeDL controller is installed under kubedl-system namespace.

Running the command from master branch uses the daily docker image.

Install the KubeDL Dashboard

kubectl apply -f

The dashboard will list nodes. Hence, its service account requires the list node permission. Check the dashboard.

Uninstall KubeDL controller and dashboard

kubectl delete namespace kubedl-system

Delete CRDs

kubectl get crd | grep | cut -d ' ' -f 1 | xargs kubectl delete crd

Delete ClusterRole and ClusterRoleBindings

kubectl delete clusterrole kubedl-leader-election-role
kubectl delete clusterrolebinding kubedl-manager-rolebinding

Enable specific job Kind

KubeDL supports all kinds of jobs(tensorflow, pytorch etc.) in a single Kubernetes operator. You can selectively enable the kind of jobs to support. There are three options:

  1. Default option. Just install the job CRDs required. KubeDL will automatically enable the corresponding job controller.
  2. Set env WORKLOADS_ENABLE in KubeDL container. The value is a list of job types to be enabled. For example, WORKLOADS_ENABLE=TFJob,PytorchJob means only Tensorflow and Pytorch Job are enabled.
  3. Set startup flags --workloads in KubeDL container command args. The value is a list of job types to be enabled like --workloads TFJob,PytorchJob.