Kubernetes resource limits and requests explained

Need to properly set CPU and memory limits for Kubernetes pods to avoid OOMKill, throttling, or wasted resources.

Understanding requests vs limits:

resources:
  requests:     # Guaranteed minimum
    cpu: 100m       # 0.1 CPU core
    memory: 128Mi   # 128 MiB
  limits:       # Maximum allowed
    cpu: 500m       # 0.5 CPU core
    memory: 256Mi   # 256 MiB

Requests: Used for scheduling. K8s places pod on node with enough capacity.
Limits: Enforced at runtime. Memory over limit = OOMKill. CPU over limit = throttled.

Guidelines:

• Set memory request = limit (avoid OOMKill surprises)
• Set CPU request based on average usage
• Set CPU limit 2-5x of request (allow bursting)
• Use kubectl top pods to measure actual usage
• Start generous, then tighten based on metrics

# Check actual resource usage
kubectl top pods -n production

# Check resource requests/limits
kubectl describe pod <name> | grep -A 5 'Limits\|Requests'

# Find pods without limits
kubectl get pods -o json | jq '.items[] | select(.spec.containers[].resources.limits == null) | .metadata.name'

Without requests, pods compete for resources unpredictably. Without limits, a single pod can starve others. Memory request != limit means OOMKill when the node is under pressure even if the pod is within its limit.

Deploying applications to Kubernetes clusters