🔵 Compute Engine & GKE

Virtual Machines, Instance Groups, Kubernetes Engine

Google Cloud ACE · ACE · ⏱️ 28 min · ❓ 6 savol

💻 Compute Engine (GCE)

VM instancelar. Machine types: General Purpose (N2, E2), Compute Optimized (C2), Memory Optimized (M2). Persistent Disk: Standard (HDD), SSD, Balanced. Preemptible VM — 80% arzon, 24 soatda to'xtatiladi. Spot VM — Preemptiblega o'xshash, lekin cheksiz davom etadi.

📦 Instance Groups

Managed Instance Group (MIG) — avtomatik auto-healing, auto-scaling, rolling update. Unmanaged IG — manual boshqaruv. MIG + HTTP(S) Load Balancer = production-ready setup.

⎈ Google Kubernetes Engine (GKE)

Managed Kubernetes. Cluster modes: Standard (node boshqarish sizda) va Autopilot (to'liq managed). Node Pool — bir xil konfiguratsiyali nodelar to'plami. Workload Identity — GKE workloads uchun GCP IAM integration.

🔄 GKE Deployment Strategiyalari

Rolling Update — asta-sekin yangilash, zero downtime. Blue/Green — yangi version parallel run, switch. Canary — oz qism trafikni yangi versiyaga yo'naltirish. kubectl: get, describe, apply, delete asosiy buyruqlar.

💡 Asosiy nuqtalar

Preemptible/Spot VM: arzon, lekin interrupt bo'lishi mumkin
MIG: auto-scaling + auto-healing
GKE Autopilot: node pool boshqarishsiz
Workload Identity: Service Account o'rniga GKE-native auth
PersistentVolume GKE da disk kabi ishlaydi

📋 Kod misoli

# GKE cluster yaratish
gcloud container clusters create my-cluster \
  --zone us-central1-a \
  --num-nodes 3 \
  --machine-type e2-medium

# kubectl asosiy buyruqlar
kubectl get pods -n default
kubectl apply -f deployment.yaml
kubectl describe pod my-pod
kubectl logs my-pod --tail=50
kubectl scale deployment my-app --replicas=5

🎯 Imtihon maslahatlari

Preemptible VM max 24 soat, Spot VM cheksiz — lekin ikkisi ham interrupt bo'lishi mumkin
MIG (Managed) vs UIG (Unmanaged): imtihonda "auto-healing" yoki "auto-scaling" kerak bo'lsa → MIG
GKE Standard: node pool boshqarasiz, Autopilot: Google boshqaradi — "fully managed Kubernetes" = Autopilot
Persistent Disk Regional — ikki zonada replikatsiya (high availability uchun)
`kubectl rollout undo` — deployment ni oldingi versiyaga qaytarish (exam da tez-tez chiqadi)

⚠️ Ko'p adashadigan

Spot VM = Preemptible VM deb o'ylash — Spot VM da 24 soat cheklov yo'q, lekin interrupt bo'lishi mumkin
GKE Autopilot da node pool manually boshqarish mumkin deb o'ylash — Autopilot da node pool Google boshqaradi
MIG "stateful" va "stateless" farqini bilmaslik — stateful MIG disk va IP saqlaydi, stateless saqlmaydi

🧠 Eslab qolish: "MIG = Magic Instance Group": Auto-Heal + Auto-Scale + Load Balance. "SNAP" = Standard (nodes you manage), gkE Autopilot (nodes Google manages)