Kubernetes Expert - Google Expertise

Expert: Alex Kim (Google SRE, 11 years) Level: 10/10 - Google created Kubernetes (K8s)

Overview

Kubernetes at Google scale - running billions of containers across thousands of clusters. Google invented K8s based on their internal Borg system. Production K8s knowledge from the source.

When to Use K8s Resources

Workloads:

• Deployments - Stateless apps - StatefulSets - Stateful apps, databases
• DaemonSets - Per-node agents
• Jobs/CronJobs - Batch processing

Networking:

• Services - Load balancing - Ingress - HTTP routing
• NetworkPolicies - Traffic control

Configuration:

• ConfigMaps - Configuration data - Secrets - Sensitive data
• PersistentVolumeClaims - Storage

Advanced:

• Custom Resource Definitions (CRDs) - Extend K8s
• Operators - Custom controllers
• Service Mesh - Istio

Core Workflow

1. Define manifests - YAML for resources
2. Deploy - kubectl apply or GitOps (ArgoCD)
3. Monitor - Prometheus, metrics
4. Scale - HPA, VPA for auto-scaling
5. Update - Rolling updates, zero downtime
6. Troubleshoot - kubectl logs, describe, events

Google GKE Best Practices

High Availability:

• Multi-zone clusters
• Pod anti-affinity
• PodDisruptionBudgets

Zero Downtime:

• Rolling updates: maxUnavailable: 0
• Readiness probes
• Graceful shutdown

Security:

• Run as non-root
• Read-only filesystems
• NetworkPolicies
• Workload Identity (GKE)

See references/deployments.md for production examples.

Google's K8s Scale

Google runs:

• Billions of containers per week
• Thousands of clusters
• Borg (K8s predecessor) since 2003
• 99.99% uptime for critical services

GitOps with ArgoCD

Declarative deployments from Git:

# Application definition
apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: api-server
spec:
  source:
    repoURL: https://github.com/org/k8s-manifests
    targetRevision: HEAD
    path: apps/api-server
  destination:
    server: https://kubernetes.default.svc
    namespace: production
  syncPolicy:
    automated:
      prune: true
      selfHeal: true

Best Practices

1. Declarative config - YAML in Git, not kubectl run
2. Health checks - Liveness + readiness probes
3. Resource limits - CPU + memory requests/limits
4. Auto-scaling - HPA for replicas, VPA for resources
5. Monitoring - Prometheus + Grafana dashboards

Related Skills

• sre-practices - SRE on K8s
• gcp-expert - GKE platform
• observability - K8s monitoring

---

Last Updated: 2026-02-03 Expert: Alex Kim (Google SRE, 11 years) - Runs billion-container systems