Skip to content

Streamline kubernetes operations with kubectl-ai

Lab Outline

  • Duration: 30 minutes
  • Objective: Enable engineering students to quickly install, configure, and use kubectl-ai to manage Kubernetes clusters using natural language queries.

Content

Prerequisites

  • OpenAI compatible LLM server exposing models (Mistral Large as example in this lab).
    • If running this lab as part of a workshop, one is given to you e.g. http://models.apps.devopsp.mop.demo.
  • Access to a terminal with kubectl installed and configured to a K8s cluster.
  • We recommend minikube or kind.

Introduction to kubectl-ai

  • kubectl-ai is an AI-powered Kubernetes assistant that allows you to manage clusters using plain English, translating your requests into valid kubectl commands or YAML manifests.
  • It supports multiple AI models (Google Gemini by default, OpenAI, Azure OpenAI, local LLMs like Ollama).
  • Key benefits: reduces Kubernetes learning curve, boosts productivity, and democratizes cluster access.

Lab Instructions

Installing kubectl-ai (Linux and MacOS)

  1. Install the kubectl-ai plugin: 
    curl -sSL https://raw.githubusercontent.com/GoogleCloudPlatform/kubectl-ai/main/install.sh | bash
  2. Verify installation: 
    kubectl ai --help

Running kubectl-ai with OpenAI compatible provider

export OPENAI_API_KEY=sk-CHANGEME
export OPENAI_ENDPOINT=http://CHANGEME/v1
kubectl ai --llm-provider=openai --model=mistral-large

Note

If running this lab in a workshop setup, you should have a LiteLLM proxy OpenAI compatible server running, you can find the route (endpoint) and API key in the ai-models namespace. The API key is the LITELLM_MASTER_KEY value of the litellm-secret secret.

This has started an interactive session, you can start asking questions with follow-up:

Create an nginx deployment with 1 replicas in the genai-CHANGEME namespace

Get the logs

Running individual queries

Try the following commands and observe the AI-generated kubectl commands. Adapt as required by your context:

kubectl ai "Create an nginx deployment with 1 replicas in the genai-CHANGEME namespace for OpenShift" --llm-provider=openai --model=mistral-large 

kubectl ai "Scale my nginx deployment to 2 replicas" --llm-provider=openai --model=mistral-large 

kubectl ai "Show me all pods that failed in the last hour" --llm-provider=openai --model=mistral-large 

kubectl ai "Generate a HorizontalPodAutoscaler YAML for the web-api deployment" --llm-provider=openai --model=mistral-large 

kubectl ai "Show me the logs for the nginx pod in the genai-CHANGEME namespace" --llm-provider=openai --model=mistral-large

Generating YAML manifests

kubectl ai "Write a deployment with nginx and a service that exposes port 80" --llm-provider=openai --model=mistral-large

The tool will generate YAML and ask if you want to apply it.

Explore additional examples

Explore these extra scenarios to deepen your hands-on experience with kubectl-ai. These examples cover both common and advanced Kubernetes operations, all using natural language.

Create and Update Resources

Create an nginx deployment with 1 replicas

Update the nginx deployment to use 2 replicas and expose port 8080

Create a genai-CHANGEME namespace, then create an nginx pod in that namespace

Service and Networking

Create a service of type LoadBalancer for my nginx deployment

Expose the redis deployment on port 6379 as a ClusterIP service

Set up a NetworkPolicy to only allow traffic to the frontend deployment from the genai-CHANGEME namespace

Pod and Deployment Management

Restart all pods in the genai-CHANGEME namespace

Delete the pod named my-app-123 in the genai-CHANGEME namespace

Show the status of all deployments in the kube-system namespace

Logs and Troubleshooting

Fetch logs for the nginx app in the genai-CHANGEME namespace

Show events for the backend deployment

Explain the error in this log

Note

(pipe a log file: cat error.log | kubectl-ai "explain the error --llm-provider=openai --model=mistral-large )

Scaling and Autoscaling

Scale the production app to 2 replicas

Generate a HorizontalPodAutoscaler YAML for the api-server deployment

Advanced YAML Generation

Write a deployment YAML for a Python app using the image python:3.9 with environment variable DEBUG=true

Create a ConfigMap named app-config with keys LOG_LEVEL=debug and TIMEOUT=30

Generate a Secret manifest with username=admin and password=Pa\$\$w0rd

Batch and Multiple Resources

Create a namespace called genai-CHANGEME, then deploy a busybox pod in it

Create a deployment and a service for a MongoDB database in the genai-CHANGEME namespace

Conclusion

By the end of this lab, you should be able to: - Install and configure kubectl-ai - Use natural language to manage Kubernetes resources - Generate and apply YAML manifests using AI

Feel free to experiment with your own queries and switch to other provided LLMs, then reflect on the productivity gains from using AI-powered Kubernetes tooling, make sure to also challenge the risks of such tools in the context of Agentic AI.