k8sgpt — Operator

L’operator esegue k8sgpt periodicamente dentro il cluster e salva i risultati come custom resource Result. Nessun terminale aperto, risultati consultabili con kubectl o GitOps.

Installazione

helm repo add k8sgpt https://charts.k8sgpt.ai/

helm repo update

helm upgrade --install k8sgpt-operator k8sgpt/k8sgpt-operator -n k8sgpt-operator-system --create-namespace --set serviceMonitor.enabled=true

kubectl -n k8sgpt-operator-system get pods

kubectl get crd | grep k8sgpt

CR K8sGPT verso Ollama sull’host

cat > k8sgpt-local.yaml <<'EOF2'
apiVersion: core.k8sgpt.ai/v1alpha1
kind: K8sGPT
metadata:
  name: k8sgpt-local
  namespace: k8sgpt-operator-system
spec:
  ai:
    enabled: true
    backend: localai
    model: qwen2.5:7b-instruct
    baseUrl: http://host.docker.internal:11434/v1
  noCache: false
  repository: ghcr.io/k8sgpt-ai/k8sgpt
  version: v0.4.1
  filters:
    - Pod
    - Service
    - PersistentVolumeClaim
    - Deployment
EOF2

version è il tag dell’immagine runtime che l’operator avvia: allinealo a una release esistente di k8sgpt. Verifica i campi supportati dalla tua versione del CRD:

kubectl explain k8sgpt.spec.ai

kubectl apply -f k8sgpt-local.yaml

kubectl -n k8sgpt-operator-system get k8sgpt,deploy,pods

Risultati

kubectl -n k8sgpt-operator-system get results

kubectl -n k8sgpt-operator-system get results -o custom-columns=KIND:.spec.kind,NAME:.spec.name,PARENT:.spec.parentObject

kubectl -n k8sgpt-operator-system get results -o jsonpath='{range .items[*]}{"== "}{.spec.kind}{"/"}{.spec.name}{"\n"}{.spec.details}{"\n\n"}{end}'

kubectl -n k8sgpt-operator-system get result -o yaml | yq '.items[0].spec'

I Result si aggiornano a ogni ciclo: quando correggi un guasto, il relativo Result sparisce.

Metriche

Con serviceMonitor.enabled=true Prometheus raccoglie le metriche dell’operator. Individua i nomi esposti:

curl -s 'localhost:9090/api/v1/label/__name__/values' | jq -r '.data[]' | grep -i k8sgpt

Sink (notifiche)

Il CR supporta sink come Slack o Mattermost tramite spec.sink. In ambienti offline punta a un’istanza interna:

kubectl explain k8sgpt.spec.sink

Log e troubleshooting

kubectl -n k8sgpt-operator-system logs deploy/k8sgpt-operator-controller-manager -f

kubectl -n k8sgpt-operator-system logs deploy/k8sgpt-local -f

kubectl -n k8sgpt-operator-system describe k8sgpt k8sgpt-local

Rimozione

kubectl delete -f k8sgpt-local.yaml

helm uninstall k8sgpt-operator -n k8sgpt-operator-system