Create a DigitalOcean Managed Kubernetes Cluster with Node Autoscaling Enabled

Desired Outcome

Create a fully functional DigitalOcean Managed Kubernetes (DOKS) cluster that automatically adjusts the number of worker nodes based on workload demand, providing cost‑efficient scaling and high availability.

Prerequisites

• A DigitalOcean account with sufficient quota for at least one worker node.
• A Personal Access Token (PAT) with read/write scope generated in the account settings.
• The doctl CLI installed and configured with the PAT (doctl auth init).
• kubectl installed locally (version compatible with the cluster, e.g., v1.27+).
• Basic familiarity with Kubernetes concepts such as pods, deployments, and node pools.

Procedure

1. Create the cluster with autoscaling enabled

   Run the following command, replacing placeholders with your preferred values:

   doctl k8s cluster create  \
     --region  \
     --node-pool name=,size=,min-nodes=,max-nodes=
   
   Example:
   doctl k8s cluster create my-doks-cluster \
     --region nyc1 \
     --node-pool name=worker-pool,size=s-2vcpu-4gb,min-nodes=1,max-nodes=5
   
   Wait for the cluster to reach the running state
   Check status periodically:
   doctl k8s cluster list
   When the STATUS column shows running, proceed.
   Retrieve and set the kubeconfig
   doctl k8s kubeconfig save 
   export KUBECONFIG=${HOME}/.kube/config
   
   Verify the node pool autoscaling parameters
   Inspect the node pool labels to confirm min/max settings:
   kubectl get nodes -l doctl.digitalocean.com/node-pool= -o wide
   
   The output should list nodes whose count is currently between  and .
   

Verification

1. Confirm API accessibility

   kubectl get nodes
   

   You should see the control plane endpoint reachable and a list of worker nodes.

2. Check that the cluster autoscaler is running

   kubectl -n kube-system get pods -l app=cluster-autoscaler
   

   Ensure the pod status is Running.

3. Trigger a scale‑up event

   Deploy a workload that requests more resources than currently available. For example, a busybox loop that requests 500m CPU:

   cat <<EOF | kubectl apply -f -
   apiVersion: v1
   kind: Pod
   metadata:
     name: cpu-loader
   spec:
     containers:
     - name: loader
       image: busybox
       command: ["sh", "-c", "while true; do echo hello; sleep 10; done"]
       resources:
         requests:
           cpu: "500m"
           memory: "256Mi"
     restartPolicy: Never
   EOF
   

   After a few moments, run kubectl get nodes again. If the current node count is at the minimum and the pod cannot be scheduled, the autoscaler should add a node (up to the max-nodes limit).

4. Inspect autoscaler logs for confirmation

   kubectl -n kube-system logs deployment/cluster-autoscaler --tail=20
   

   Look for lines similar to pod triggered scale-up: ... or scale-down: ... indicating the autoscaler is active.

Recovery Options

If autoscaling does not behave as expected:

• Manually adjust the node pool size

  doctl k8s node-pool update  \
    --cluster  \
    --node-count 
  

  This changes the desired count immediately; the autoscaler will still respect the min-nodes and max-nodes bounds.

• Recycle the cluster

  As a last resort, delete and recreate the cluster with the same specification. Ensure any persistent volumes have snapshots or backups enabled before deletion.

  doctl k8s cluster delete 
  # after confirmation, recreate using the same command from step 1

Limitations and Practical Checks

• Autoscaling may cause additional charges if the node count reaches the max-nodes limit. Monitor usage via DigitalOcean billing alerts or the doctl account command.
• Changing autoscaling parameters (min-nodes or max-nodes) requires a cluster upgrade window and may cause brief disruption while nodes are provisioned or terminated.
• To verify that the autoscaler respects the bounds, repeatedly run kubectl get nodes while applying and removing load‑generating pods, ensuring the node count never falls below min-nodes or exceeds max-nodes.