3.2 : QoS Class – BestEffort

https://drive.google.com/file/d/1uQarejDCpGhAfyNhvSyw4CmMIgDq2VyW/view?usp=sharing

🔍 YAML Breakdown

apiVersion: v1
kind: Pod
metadata:
  name: nnappone
  namespace: learning
  labels:
    app: nnappone
spec:
  containers:
    - name: networknuts-app
      image: nginx
# ❌ NO `resources` block!

🔴 Key Observation:

There is no resources section → no requests, no limits for CPU or memory.

📌 What Is "BestEffort" QoS?

Kubernetes automatically assigns a Quality of Service (QoS) class to every Pod based on its resource configuration.

QoS Class	Condition	Priority	Eviction Risk
Guaranteed	`requests == limits` for all containers & all resources (CPU + memory)	🔝 Highest	Never evicted due to resource pressure (if node has capacity)
Burstable	Some resources have `requests` ≠ `limits`, or only `requests`/`limits` set for some resources	🟡 Medium	Evicted after BestEffort, before Guaranteed
BestEffort	❌ No `requests` or `limits` for any container	🔻 Lowest	First to be evicted under memory pressure!

💡 BestEffort = "Use whatever is free — but we make no promises."

⚠️ Why Avoid BestEffort in Production?

Unpredictable performance: Your app may get starved if other Pods consume resources.
First to die: During Node Memory Pressure, kubelet kills BestEffort Pods first.
No scheduling guarantees: Scheduler doesn’t reserve resources for it.

✅ Only acceptable for:

Temporary debug Pods

Labs/learning

Truly stateless, non-critical workloads

🧪 Lab: Deploy BestEffort Pod & Verify QoS

🔧 Steps

# 1. Create namespace
kubectl create namespace learning

# 2. Apply the Pod
kubectl apply -f pod-simple-qos-besteffort.yml

# 3. Check Pod status
kubectl get pods -n learning

# 4. Describe Pod → look for "QoS Class"
kubectl describe pod nnappone -n learning | grep "QoS Class"

# ✅ Expected output:
# QoS Class:       BestEffort

# 5. (Optional) View full resource info in JSON
kubectl get pod nnappone -n learning -o jsonpath='{.status.qosClass}{"\\\\n"}'

# 6. Clean up
kubectl delete pod nnappone -n learning
kubectl delete namespace learning