brokerua / devops101-w5-m5-t2

kubectl ai with openai

Write Kubernetes manifests with LLama AI

Run LLama AI

docker run -d --rm -p 8080:8080 ghcr.io/sozercan/llama3.1:8b
export OPENAI_ENDPOINT="http://localhost:8080/v1"
export OPENAI_DEPLOYMENT_NAME="llama-3.1-8b-instruct"
export OPENAI_API_KEY="n/a"

Install kubectl plugin

wget https://github.com/sozercan/kubectl-ai/releases/download/v0.0.13/kubectl-ai_linux_amd64.tar.gz
tar zxf kubectl-ai_linux_amd64.tar.gz
sudo mv kubectl-ai /usr/local/bin/
alias k=kubectl

Example

Write prompt and save result to file

k ai "create deploy redis" --raw > redis-deploy.yaml

NAME	PROMPT	DESCRIPTION	EXAMPLE
app	Create a Kubernetes Pod manifest where should have labels and contain k8s image with expose http port	A basic Pod definition with labels and an exposed HTTP port.	link
app-livenessProbe	Create a Kubernetes Pod named "app-livenessprob" in the "demo" namespace using the image "gcr.io/k8s-k3s/demo:v1.0.0" with a liveness probe checking "/" on port 8000 and exposing port 8080 for HTTP	A Pod with a liveness probe and an exposed port in the "demo" namespace.	link
app-readinessProbe	Create a Kubernetes Pod named "app-readinessprob" using the image "gcr.io/k8s-k3s/demo:v2.0.0" with liveness and readiness probes checking "/" and "/ready" on port 8000, and exposing port 8000 for HTTP.	A Pod with both liveness and readiness probes, and an exposed HTTP port.	link
app-volumeMounts	Create a Kubernetes Pod named "app-volume" using the image "gcr.io/kuar-demo/kuard-amd64:1" with liveness and readiness probes checking "/healthy" and "/ready" on port 8080, exposing port 8080 for HTTP, and mounting the host path "/var/lib/app" to "/data".	A Pod with volume mounts and probes, sharing a volume between containers.	link
app-cronjob	Create a Kubernetes CronJob named "app-cronjob" that runs every 5 minutes, using the "bash" image to execute the command echo "Hello world" with a restart policy of "OnFailure".	A CronJob that prints "Hello world" every 5 minutes with an "OnFailure" restart policy.	link
app-job	Create a Kubernetes Job named "app-job-rsync" using the "google/cloud-sdk:275.0.0-alpine" image to run sutil, mounting a GCE Persistent Disk to /data/input, and set the restart policy to "Never" with a backoff limit of 0.	A Job for syncing data with gsutil, using a GCE Persistent Disk, and a "Never" restart policy.	link
app-multicontainer	Create a Kubernetes Pod named "app-multi-containers" with two containers: the first using the "nginx" image and mounting an empty directory at nginx/html, and the second using the "debian" image, which writes the current date to index.html every second, with both containers sharing the same empty directory volume.	A Pod with multiple containers sharing a volume, where one container serves content and the other updates it regularly.	link
app-resources	Create a Kubernetes Pod named "app-resource" using the image "gcr.io/kuar-demo/kuard-amd64:1" with liveness and readiness probes on port 8080, exposing the same port, and specify resource requests of 100m CPU and 128Mi memory, with limits of 100m CPU and 256Mi memory.	A Pod with specified resource requests and limits, including liveness and readiness probes.	link
app-secret-env	Create a Kubernetes Pod named "app-secret-env" using the "redis" image, with environment variables SECRET_USERNAME and SECRET_PASSWORD populated from the username and password keys in the mysecret1 secret, and set the restart policy to "Never".	A Pod that uses secret values for environment variables and has a "Never" restart policy.	link

Conclusion

LLama is smart model but need manual review and fixes. Maybe try another AI or more clearly prompts!?