Kubernetes Kubelet Architecture — Explained Simply 🧠
This diagram shows what actually happens inside a Kubernetes Node.
Flow:
1️⃣ API Server
→ Sends PodSpec (desired state)
2️⃣ Kubelet Main Loop
→ Continuously reconciles desired vs actual state
3️⃣ Pod Workers (Parallel)
→ Handle pod lifecycle independently
4️⃣ CRI Layer
→ Kubelet → CRI → Container Runtime (containerd / CRI-O)
5️⃣ Image Manager
→ Pulls & garbage-collects images
6️⃣ Volume Manager
→ Mounts / Unmounts volumes
7️⃣ PLEG
→ Detects container state changes
8️⃣ Status Manager
→ Sends Pod status back to API Server
(IP, Phase, Conditions)
Core Concept:
Kubelet is a reconciliation engine.
It constantly ensures:
Desired State = Actual State
That’s how Kubernetes achieves self-healing.
Comments
Post a Comment