Instruction file imported from luysantanadev/k8s-aspnet (.github/instructions/kubernetes-manifests.instructions.md). Copyright stays with the author.

Kubernetes Manifests Instructions

Your Mission

Create production-ready Kubernetes manifests that prioritize security, reliability, and operational excellence with consistent labeling, proper resource management, and comprehensive health checks.

Labeling Conventions

Required Labels (Kubernetes recommended):

• app.kubernetes.io/name: Application name
• app.kubernetes.io/instance: Instance identifier
• app.kubernetes.io/version: Version
• app.kubernetes.io/component: Component role
• app.kubernetes.io/part-of: Application group
• app.kubernetes.io/managed-by: Management tool

Additional Labels:

• environment: Environment name
• team: Owning team
• cost-center: For billing

Useful Annotations:

• Documentation and ownership
• Monitoring: prometheus.io/scrape, prometheus.io/port, prometheus.io/path
• Change tracking: git commit, deployment date

SecurityContext Defaults

Pod-level:

• runAsNonRoot: true
• runAsUser and runAsGroup: Specific IDs
• fsGroup: File system group
• seccompProfile.type: RuntimeDefault

Container-level:

• allowPrivilegeEscalation: false
• readOnlyRootFilesystem: true (with tmpfs mounts for writable dirs)
• capabilities.drop: [ALL] (add only what's needed)

Pod Security Standards

Use Pod Security Admission:

• Restricted (recommended for production): Enforces security hardening
• Baseline: Minimal security requirements
• Apply at namespace level

Resource Requests and Limits

Always define:

• Requests: Guaranteed minimum (scheduling)
• Limits: Maximum allowed (prevents exhaustion)

QoS Classes:

• Guaranteed: requests == limits (best for critical apps)
• Burstable: requests < limits (flexible resource use)
• BestEffort: No resources defined (avoid in production)

Health Probes

Liveness: Restart unhealthy containers Readiness: Control traffic routing Startup: Protect slow-starting applications

Configure appropriate delays, periods, timeouts, and thresholds for each.

Rollout Strategies

Deployment Strategy:

• RollingUpdate with maxSurge and maxUnavailable
• Set maxUnavailable: 0 for zero-downtime

High Availability:

• Minimum 2-3 replicas
• Pod Disruption Budget (PDB)
• Anti-affinity rules (spread across nodes/zones)
• Horizontal Pod Autoscaler (HPA) for variable load

Validation Commands

Pre-deployment:

• kubectl apply --dry-run=client -f manifest.yaml
• kubectl apply --dry-run=server -f manifest.yaml
• kubeconform -strict manifest.yaml (schema validation)
• helm template ./chart | kubeconform -strict (for Helm)

Policy Validation:

• OPA Conftest, Kyverno, or Datree

Rollout & Rollback

Deploy:

• kubectl apply -f manifest.yaml
• kubectl rollout status deployment/NAME

Rollback:

• kubectl rollout undo deployment/NAME
• kubectl rollout undo deployment/NAME --to-revision=N
• kubectl rollout history deployment/NAME

Restart:

• kubectl rollout restart deployment/NAME

Manifest Checklist

• Labels: Standard labels applied
• Annotations: Documentation and monitoring
• Security: runAsNonRoot, readOnlyRootFilesystem, dropped capabilities
• Resources: Requests and limits defined
• Probes: Liveness, readiness, startup configured
• Images: Specific tags (never :latest)
• Replicas: Minimum 2-3 for production
• Strategy: RollingUpdate with appropriate surge/unavailable
• PDB: Defined for production
• Anti-affinity: Configured for HA
• Graceful shutdown: terminationGracePeriodSeconds set
• Validation: Dry-run and kubeconform passed
• Secrets: In Secrets resource, not ConfigMaps
• NetworkPolicy: Least-privilege access (if applicable)

Best Practices Summary

1. Use standard labels and annotations
2. Always run as non-root with dropped capabilities
3. Define resource requests and limits
4. Implement all three probe types
5. Pin image tags to specific versions
6. Configure anti-affinity for HA
7. Set Pod Disruption Budgets
8. Use rolling updates with zero unavailability
9. Validate manifests before applying
10. Enable read-only root filesystem when possible