🧠 Concept 13: Resource Requests & Limits (Scheduling + Performance 💯)
🚀 1. Core Idea (1-line)
👉 Requests = guaranteed resources, Limits = maximum allowed resources
🧠 2. Why This Concept Exists (VERY IMPORTANT ⚠️)
Without limits:
-
One pod can consume ALL CPU/RAM ❌
-
Other apps starve 😵
👉 Cluster becomes unstable
💡 3. Two Key Terms
🟢 1. Requests (Minimum guarantee)
👉 Scheduler uses this to decide:
- Where to place the pod
Example:
requests:
cpu: "500m"
memory: "256Mi"👉 Means:
-
Needs at least 0.5 CPU
-
Needs at least 256MB RAM
🔴 2. Limits (Maximum cap)
limits:
cpu: "1"
memory: "512Mi"👉 Means:
-
Cannot exceed 1 CPU
-
Cannot exceed 512MB RAM
⚙️ 4. How Scheduling Works (VERY IMPORTANT 🔥)
👉 Kubernetes scheduler checks:
- Node has enough requests capacity?
✅ Yes → Pod scheduled
❌ No → Pod pending
💥 5. Runtime Behavior
CPU:
-
Can burst above request (until limit)
-
If exceeds limit → throttled ⚠️
Memory:
- If exceeds limit → OOMKilled 💀
📦 6. Example YAML
resources:
requests:
cpu: "200m"
memory: "128Mi"
limits:
cpu: "500m"
memory: "256Mi"🧠 7. VERY IMPORTANT Concepts
👉 CPU Units:
-
1000m = 1 CPU
-
500m = 0.5 CPU
👉 Memory Units:
-
Mi = Mebibyte
-
Gi = Gibibyte
🔥 8. Real-world DevOps Insight
For your ML workloads 👀:
-
Inference → moderate CPU + memory
-
Training → high CPU/GPU
👉 Set proper limits to avoid:
-
Node crash
-
Resource starvation
⚠️ 9. Common Mistakes (INTERVIEW TRAPS)
❌ Not setting requests → bad scheduling
❌ Setting limits too low → OOMKill
❌ Setting limits too high → waste resources
💼 10. Interview Answer
👉 “Resource requests define the minimum resources required for scheduling, while limits define the maximum resources a container can consume, ensuring fair resource usage and cluster stability.”
⚡ 11. CKA Commands
kubectl describe pod <name>👉 Shows:
-
Requests
-
Limits
-
OOMKilled events
🧠 12. Memory Trick
👉 Request = reservation 🪑
👉 Limit = boundary 🚧
🔥 13. Pro Insight (Real-world)
-
Always set requests + limits
-
Use:
-
HPA (autoscaling)
-
Metrics (Prometheus)
👉 Optimize based on usage
🚀 Next Step
Bol:
👉 “next”
Then we go to:
🔥 Concept 14: HPA (Horizontal Pod Autoscaling 💯 — VERY IMPORTANT FOR REAL WORLD)