☸️ Nodes
For small devevelopment cluster, you can create a single Nodepool. Every node should have the following labels.
# Useful for the Service deployments.
- role.datalayer.io/router=true
- role.datalayer.io/system=true
- role.datalayer.io/api=true
- role.datalayer.io/solr=true
# Useful for the Runtime pods.
- role.datalayer.io/jupyter=true
- node.datalayer.io/variant=medium
- xpu.datalayer.io/cpu=true
Run plane k8s-label-nodes to apply all the above labels to every worker node in one go. This is the fastest way to set up a small development cluster.
Use your favorite tool to create those nodes, or use Clouder, a tool to ease the Kubernetes cluster creation and sharing.
For bigger production cluster, we recommend creating separated Ensure the nodes are labelled based on their Nodepool.
Nodepool router
Ensure the nodes are labelled with the following details.
- role.datalayer.io/router=true
- node.datalayer.io/variant=default
Nodepool api
Ensure the nodes are labelled with the following details.
- role.datalayer.io/api=true
Nodepool system
Ensure the nodes are labelled with the following details.
- role.datalayer.io/system=true
Nodepool solr
Ensure the nodes are labelled with the following details.
- role.datalayer.io/solr=true
Nodepool jupyter-cpu-medium
Ensure the nodes are labelled with the following details.
- role.datalayer.io/jupyter=true
- node.datalayer.io/variant=default
- xpu.datalayer.io/cpu=true
Nodepool jupyter-gpu-cuda-medium
Ensure the nodes are labelled with the following details.
- role.datalayer.io/jupyter=true
- node.datalayer.io/variant=default
- xpu.datalayer.io/gpu-cuda=true