Adding worker nodes with custom configurations

This guide will walk you through how to deploy multiple k8s-worker applications with different configurations, to create node groups with specific capabilities or requirements.

What you’ll need

This guide assumes the following:

  • A working Kubernetes cluster deployed with the k8s charm

Example worker configuration

In this example, we will create two k8s-worker applications with different configuration options.

Note

The configurations shown below are examples to demonstrate the deployment pattern. You should adjust the node configurations, labels, and other parameters according to your specific infrastructure requirements, workload needs, and organizational policies. Review the charm configuration options documentation to understand all available parameters that can be customized for your worker nodes.

  1. Workers for memory-intensive workloads (worker-memory-config.yaml):

memory-workers:
  bootstrap-node-taints: "workload=memory:NoSchedule"
  kubelet-extra-args: "system-reserved=memory=2Gi"

  1. Workers for GPU workloads (worker-gpu-config.yaml):

gpu-workers:
  bootstrap-node-taints: "accelerator=nvidia:NoSchedule"
  node-labels: "gpu=true"

Deploy the worker applications with the custom configurations and integrate them with the k8s application:

juju deploy k8s-worker memory-workers --config ./worker-memory-config.yaml
juju integrate k8s memory-workers:cluster
juju integrate k8s memory-workers:containerd
juju integrate k8s memory-workers:cos-tokens

juju deploy k8s-worker gpu-workers --config ./worker-gpu-config.yaml
juju integrate k8s gpu-workers:cluster
juju integrate k8s gpu-workers:containerd
juju integrate k8s gpu-workers:cos-tokens

Monitor the installation progress by running the following command:

juju status --watch 1s