Scaling Guide

Application: [Application Name]
Last Updated: [Date]
Owner: [Team/Contact]

Overview

This guide explains how to scale [Application Name] to handle increased load.

Scaling Types

Horizontal Scaling (Recommended)

What it is: Adding more pod replicas

When to use:

• Increased traffic
• Need for high availability
• Stateless applications

How to scale:

# Scale deployment
kubectl scale deployment/[deployment-name] --replicas=[count] -n [namespace-name]

# Or use Helm
helm upgrade [release-name] [chart-path] \
  --set replicaCount=[count] \
  --namespace [namespace-name]

Vertical Scaling

What it is: Increasing pod resource limits

When to use:

• Application needs more CPU/memory
• Horizontal scaling not possible
• Resource-intensive workloads

How to scale:

# Update deployment with new resource limits
kubectl set resources deployment/[deployment-name] \
  --limits=cpu=[cpu],memory=[memory] \
  --requests=cpu=[cpu],memory=[memory] \
  -n [namespace-name]

Cluster Scaling

What it is: Adding more nodes to cluster

When to use:

• Cluster resources exhausted
• Need for more capacity
• High availability requirements

How to scale:

• GKE: Use node pools and autoscaling
• Kind: Not applicable (local only)

Autoscaling

Horizontal Pod Autoscaler (HPA)

Setup:

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: [deployment-name]-hpa
  namespace: [namespace-name]
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: [deployment-name]
  minReplicas: 2
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 70
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 80

Apply:

kubectl apply -f hpa.yaml

Verify:

kubectl get hpa -n [namespace-name]
kubectl describe hpa [deployment-name]-hpa -n [namespace-name]

Vertical Pod Autoscaler (VPA)

Note: VPA requires VPA controller installed

Setup:

apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: [deployment-name]-vpa
  namespace: [namespace-name]
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: [deployment-name]
  updatePolicy:
    updateMode: "Auto"

Scaling Decision Matrix

Scenario	Action	Method
Increased traffic	Scale horizontally	Increase replicas or enable HPA
High CPU usage	Scale horizontally	Add more pods
High memory usage	Scale vertically	Increase memory limits
Slow response times	Scale horizontally	Add more pods
Cluster resource exhaustion	Scale cluster	Add more nodes
Predictable load spikes	Pre-scale	Increase replicas before event

Monitoring Scaling

Key Metrics

Before Scaling:

• Current pod count
• CPU/memory usage
• Request rate
• Response times
• Error rates

After Scaling:

• New pod count
• CPU/memory usage per pod
• Request distribution
• Response times
• Error rates

Grafana Dashboards

Monitor scaling effectiveness:

• Cluster Overview dashboard
• Application Health dashboard
• Pod Metrics dashboard

Scaling Procedures

Manual Scaling

Step 1: Assess Current State

# Check current replicas
kubectl get deployment [deployment-name] -n [namespace-name]

# Check resource usage
kubectl top pods -n [namespace-name]

# Check metrics in Grafana

Step 2: Determine Target Replicas

• Based on current load
• Based on expected load
• Based on resource availability

Step 3: Scale

kubectl scale deployment/[deployment-name] --replicas=[count] -n [namespace-name]

Step 4: Monitor

# Watch rollout
kubectl rollout status deployment/[deployment-name] -n [namespace-name]

# Check new pods
kubectl get pods -n [namespace-name] -w

# Monitor metrics
# Check Grafana dashboards

Step 5: Verify

• [ ] All pods running
• [ ] Load distributed
• [ ] Response times improved
• [ ] No errors

Autoscaling Setup

Step 1: Enable Metrics Server

# Verify metrics server
kubectl get deployment metrics-server -n kube-system

Step 2: Create HPA

kubectl apply -f hpa.yaml

Step 3: Test Autoscaling

# Generate load (example)
kubectl run -it --rm load-generator --image=busybox --restart=Never -- \
  /bin/sh -c "while true; do wget -q -O- http://[service-name].[namespace-name].svc.cluster.local; done"

# Watch HPA
kubectl get hpa -n [namespace-name] -w

# Watch pods
kubectl get pods -n [namespace-name] -w

Scaling Best Practices

1. Start Small: Begin with minimum viable replicas
2. Monitor First: Understand baseline before scaling
3. Use Autoscaling: Enable HPA for dynamic scaling
4. Test Scaling: Verify scaling works before production
5. Document Thresholds: Document when to scale
6. Plan for Scale-Down: Ensure scale-down works too

Troubleshooting

Pods Not Scaling Up

Check:

• HPA configuration
• Resource availability
• Pod resource requests
• Cluster capacity

Pods Scaling Too Aggressively

Adjust:

• HPA thresholds
• Min/max replicas
• Scaling policies

Resource Exhaustion

Solutions:

• Scale cluster (add nodes)
• Optimize application
• Reduce resource requests
• Scale down other workloads

Related Documentation

• Deployment Runbook
• Incident Response Playbook
• Architecture Documentation

---

Remember: Scale proactively based on metrics, not reactively after issues occur.