diff --git a/.gitignore b/.gitignore
index f018e29..964eb88 100644
--- a/.gitignore
+++ b/.gitignore
@@ -72,12 +72,10 @@ Thumbs.db
 # mypy
 .mypy_cache/
 
-# Terraform
-terraform/.terraform/
-terraform/terraform.tfstate
-terraform/terraform.tfstate.backup
-terraform/terraform.tfvars
-terraform/tfplan
+# Monitoring volumes
+docker/monitoring/grafana/data/
+docker/monitoring/prometheus/data/
+docker/monitoring/loki/data/
 
 # Claude Code
 .claude/
diff --git a/Makefile b/Makefile
deleted file mode 100644
index 9ad7c10..0000000
--- a/Makefile
+++ /dev/null
@@ -1,202 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Build & Operations Makefile
-# =============================================================================
-
-# Configuration
-GCP_PROJECT_ID ?= $(shell gcloud config get-value project 2>/dev/null)
-GCP_REGION ?= asia-northeast3
-GCP_ZONE ?= asia-northeast3-a
-REGISTRY ?= $(GCP_REGION)-docker.pkg.dev/$(GCP_PROJECT_ID)/speedcam
-
-# Image Tags
-TAG ?= latest
-MAIN_IMAGE = $(REGISTRY)/main:$(TAG)
-OCR_IMAGE = $(REGISTRY)/ocr:$(TAG)
-ALERT_IMAGE = $(REGISTRY)/alert:$(TAG)
-
-# Colors for output
-GREEN = \033[0;32m
-YELLOW = \033[0;33m
-RED = \033[0;31m
-NC = \033[0m
-
-.PHONY: help build push clean tf-init tf-plan tf-apply tf-destroy tf-output restart-services restart-main restart-ocr restart-alert status health dev-up dev-down dev-logs dev-build
-
-# =============================================================================
-# Help
-# =============================================================================
-
-help: ## Show this help message
-	@echo "Usage: make [target]"
-	@echo ""
-	@echo "Targets:"
-	@grep -E '^[a-zA-Z_-]+:.*?## .*$$' $(MAKEFILE_LIST) | sort | awk 'BEGIN {FS = ":.*?## "}; {printf "  \033[36m%-20s\033[0m %s\n", $$1, $$2}'
-	@echo ""
-	@echo "Examples:"
-	@echo "  make build push restart-services  # Build, push, and restart (image update)"
-	@echo "  make tf-plan                      # Preview infrastructure changes"
-	@echo "  make tf-apply                     # Apply infrastructure changes"
-	@echo "  make TAG=v1.0.0 build push        # Build and push with specific tag"
-
-# =============================================================================
-# Build
-# =============================================================================
-
-build: build-main build-ocr build-alert ## Build all Docker images
-
-build-main: ## Build main service image
-	@echo "$(GREEN)Building main image...$(NC)"
-	docker build --platform linux/amd64 \
-		-t $(MAIN_IMAGE) \
-		-f docker/Dockerfile.main .
-
-build-ocr: ## Build OCR worker image
-	@echo "$(GREEN)Building OCR image...$(NC)"
-	docker build --platform linux/amd64 \
-		-t $(OCR_IMAGE) \
-		-f docker/Dockerfile.ocr .
-
-build-alert: ## Build alert worker image
-	@echo "$(GREEN)Building alert image...$(NC)"
-	docker build --platform linux/amd64 \
-		-t $(ALERT_IMAGE) \
-		-f docker/Dockerfile.alert .
-
-# =============================================================================
-# Push
-# =============================================================================
-
-push: push-main push-ocr push-alert ## Push all Docker images
-
-push-main: ## Push main service image
-	@echo "$(GREEN)Pushing main image...$(NC)"
-	docker push $(MAIN_IMAGE)
-
-push-ocr: ## Push OCR worker image
-	@echo "$(GREEN)Pushing OCR image...$(NC)"
-	docker push $(OCR_IMAGE)
-
-push-alert: ## Push alert worker image
-	@echo "$(GREEN)Pushing alert image...$(NC)"
-	docker push $(ALERT_IMAGE)
-
-# =============================================================================
-# Terraform (Infrastructure Management)
-# =============================================================================
-
-TF_DIR = terraform
-
-tf-init: ## Initialize Terraform
-	cd $(TF_DIR) && terraform init
-
-tf-plan: ## Preview infrastructure changes
-	cd $(TF_DIR) && terraform plan
-
-tf-apply: ## Apply infrastructure changes
-	cd $(TF_DIR) && terraform apply
-
-tf-destroy: ## Destroy all infrastructure
-	cd $(TF_DIR) && terraform destroy
-
-tf-output: ## Show Terraform outputs
-	cd $(TF_DIR) && terraform output
-
-# =============================================================================
-# Operations
-# =============================================================================
-
-restart-services: ## Restart all service instances
-	@echo "$(GREEN)Restarting services...$(NC)"
-	gcloud compute instances reset speedcam-main speedcam-ocr speedcam-alert --zone=$(GCP_ZONE)
-
-restart-main: ## Restart main service
-	gcloud compute instances reset speedcam-main --zone=$(GCP_ZONE)
-
-restart-ocr: ## Restart OCR worker
-	gcloud compute instances reset speedcam-ocr --zone=$(GCP_ZONE)
-
-restart-alert: ## Restart alert worker
-	gcloud compute instances reset speedcam-alert --zone=$(GCP_ZONE)
-
-status: ## Show deployment status
-	@echo "$(GREEN)Instance Status:$(NC)"
-	@gcloud compute instances list --filter="name~speedcam" \
-		--format="table(name,zone,machineType,status,networkInterfaces[0].networkIP,networkInterfaces[0].accessConfigs[0].natIP)"
-	@echo ""
-	@echo "$(GREEN)Service URLs:$(NC)"
-	@MAIN_IP=$$(gcloud compute instances describe speedcam-main --zone=$(GCP_ZONE) --format='get(networkInterfaces[0].accessConfigs[0].natIP)' 2>/dev/null); \
-	RMQ_IP=$$(gcloud compute instances describe speedcam-rabbitmq --zone=$(GCP_ZONE) --format='get(networkInterfaces[0].accessConfigs[0].natIP)' 2>/dev/null); \
-	echo "  API:      http://$$MAIN_IP:8000/"; \
-	echo "  Swagger:  http://$$MAIN_IP:8000/swagger/"; \
-	echo "  Health:   http://$$MAIN_IP:8000/health/"; \
-	echo "  RabbitMQ: http://$$RMQ_IP:15672/"
-
-health: ## Check health of all services
-	@echo "$(GREEN)Checking health...$(NC)"
-	@MAIN_IP=$$(gcloud compute instances describe speedcam-main --zone=$(GCP_ZONE) --format='get(networkInterfaces[0].accessConfigs[0].natIP)' 2>/dev/null); \
-	curl -s http://$$MAIN_IP:8000/health/ && echo ""
-
-logs-main: ## Show main service logs
-	gcloud compute ssh speedcam-main --zone=$(GCP_ZONE) \
-		--command="docker logs \$$(docker ps -q) 2>&1 | tail -50"
-
-logs-ocr: ## Show OCR worker logs
-	gcloud compute ssh speedcam-ocr --zone=$(GCP_ZONE) \
-		--command="docker logs \$$(docker ps -q) 2>&1 | tail -50"
-
-logs-alert: ## Show alert worker logs
-	gcloud compute ssh speedcam-alert --zone=$(GCP_ZONE) \
-		--command="docker logs \$$(docker ps -q) 2>&1 | tail -50"
-
-ssh-main: ## SSH into main instance
-	gcloud compute ssh speedcam-main --zone=$(GCP_ZONE)
-
-ssh-ocr: ## SSH into OCR instance
-	gcloud compute ssh speedcam-ocr --zone=$(GCP_ZONE)
-
-ssh-alert: ## SSH into alert instance
-	gcloud compute ssh speedcam-alert --zone=$(GCP_ZONE)
-
-# =============================================================================
-# Cleanup
-# =============================================================================
-
-clean: ## Destroy infrastructure (use 'make tf-destroy' instead)
-	@echo "$(YELLOW)Infrastructure is now managed by Terraform.$(NC)"
-	@echo "$(YELLOW)Please use 'make tf-destroy' to destroy all resources.$(NC)"
-
-clean-services: ## Note: Services are managed by Terraform
-	@echo "$(YELLOW)Services are now managed by Terraform.$(NC)"
-	@echo "$(YELLOW)Please use 'make tf-destroy' to destroy all resources.$(NC)"
-
-clean-infra: ## Note: Infrastructure is managed by Terraform
-	@echo "$(YELLOW)Infrastructure is now managed by Terraform.$(NC)"
-	@echo "$(YELLOW)Please use 'make tf-destroy' to destroy all resources.$(NC)"
-
-clean-firewall: ## Note: Firewall rules are managed by Terraform
-	@echo "$(YELLOW)Firewall rules are now managed by Terraform.$(NC)"
-	@echo "$(YELLOW)Please use 'make tf-destroy' to destroy all resources.$(NC)"
-
-clean-registry: ## Delete Artifact Registry (not managed by Terraform)
-	@echo "$(RED)Deleting Artifact Registry...$(NC)"
-	-gcloud artifacts repositories delete speedcam --location=$(GCP_REGION) --quiet 2>/dev/null || true
-
-clean-all: clean clean-registry ## Note: Infrastructure is managed by Terraform
-	@echo "$(YELLOW)Infrastructure is now managed by Terraform.$(NC)"
-	@echo "$(YELLOW)Use 'make tf-destroy' to destroy infrastructure, then 'make clean-registry' for registry.$(NC)"
-
-# =============================================================================
-# Local Development
-# =============================================================================
-
-dev-up: ## Start local development environment
-	docker-compose -f docker/docker-compose.yml up -d
-
-dev-down: ## Stop local development environment
-	docker-compose -f docker/docker-compose.yml down
-
-dev-logs: ## Show local development logs
-	docker-compose -f docker/docker-compose.yml logs -f
-
-dev-build: ## Build local development images
-	docker-compose -f docker/docker-compose.yml build
diff --git a/backend.env.example b/backend.env.example
index 171facf..bc51cb1 100644
--- a/backend.env.example
+++ b/backend.env.example
@@ -83,16 +83,11 @@ LOG_LEVEL=info
 CORS_ALLOWED_ORIGINS=http://localhost:5173,http://localhost:3000
 
 # ===========================================
-# DataDog 설정 (Optional)
-# ===========================================
-# DataDog 모니터링을 사용하려면 아래 주석을 해제하고 설정
-# DD_API_KEY=your-datadog-api-key
-# DD_SITE=ap1.datadoghq.com
-# DD_AGENT_HOST=datadog-agent
-# DD_TRACE_AGENT_PORT=8126
-# DD_ENV=dev
-# DD_SERVICE=speedcam
-# DD_LOGS_INJECTION=true
-# DD_TRACE_SAMPLE_RATE=1
-# DD_PROFILING_ENABLED=true
-# _DD_TRACE_WRITER_NATIVE=false  # gunicorn prefork 호환 (ddtrace v4+)
+# OpenTelemetry 설정
+# ===========================================
+OTEL_EXPORTER_OTLP_ENDPOINT=http://otel-collector:4317
+OTEL_EXPORTER_OTLP_PROTOCOL=grpc
+OTEL_RESOURCE_ATTRIBUTES=service.namespace=speedcam,deployment.environment=dev
+# Valid values: always_on, always_off, traceidratio, parentbased_always_on, parentbased_always_off, parentbased_traceidratio
+OTEL_TRACES_SAMPLER=parentbased_always_on
+OTEL_PYTHON_LOG_CORRELATION=true
diff --git a/config/settings/base.py b/config/settings/base.py
index f54208b..7c325de 100644
--- a/config/settings/base.py
+++ b/config/settings/base.py
@@ -29,6 +29,7 @@
     "django_filters",
     "drf_yasg",
     "django_celery_results",
+    "django_prometheus",
     # Apps
     "apps.vehicles",
     "apps.detections",
@@ -36,6 +37,7 @@
 ]
 
 MIDDLEWARE = [
+    "django_prometheus.middleware.PrometheusBeforeMiddleware",
     "corsheaders.middleware.CorsMiddleware",
     "django.middleware.security.SecurityMiddleware",
     "django.contrib.sessions.middleware.SessionMiddleware",
@@ -44,6 +46,7 @@
     "django.contrib.auth.middleware.AuthenticationMiddleware",
     "django.contrib.messages.middleware.MessageMiddleware",
     "django.middleware.clickjacking.XFrameOptionsMiddleware",
+    "django_prometheus.middleware.PrometheusAfterMiddleware",
 ]
 
 ROOT_URLCONF = "config.urls"
@@ -155,8 +158,9 @@
     "disable_existing_loggers": False,
     "formatters": {
         "verbose": {
-            "format": "{levelname} {asctime} {module} {message}",
+            "format": "{levelname} {asctime} {module} [trace_id={otelTraceID} span_id={otelSpanID}] {message}",
             "style": "{",
+            "defaults": {"otelTraceID": "0", "otelSpanID": "0"},
         },
     },
     "handlers": {
diff --git a/config/urls.py b/config/urls.py
index 5117ec7..bb61067 100644
--- a/config/urls.py
+++ b/config/urls.py
@@ -92,6 +92,8 @@ def health(request):
     re_path(
         r"^redoc/$", schema_view.with_ui("redoc", cache_timeout=0), name="schema-redoc"
     ),
+    # Prometheus Metrics
+    path("", include("django_prometheus.urls")),
     # API v1
     path("api/v1/", include("apps.vehicles.urls")),
     path("api/v1/", include("apps.detections.urls")),
diff --git a/datadog.env.example b/datadog.env.example
deleted file mode 100644
index cc663a0..0000000
--- a/datadog.env.example
+++ /dev/null
@@ -1,13 +0,0 @@
-# ===========================================
-# DataDog Agent Environment Variables
-# ===========================================
-# 사용법: 이 파일을 datadog.env로 복사하여 사용
-# cp datadog.env.example datadog.env
-
-DD_API_KEY=your-datadog-api-key
-DD_SITE=ap1.datadoghq.com
-DD_APM_ENABLED=true
-DD_APM_NON_LOCAL_TRAFFIC=true
-DD_DOGSTATSD_NON_LOCAL_TRAFFIC=true
-DD_LOGS_ENABLED=true
-DD_ENV=dev
diff --git a/docker/docker-compose.monitoring.yml b/docker/docker-compose.monitoring.yml
new file mode 100644
index 0000000..03d74d7
--- /dev/null
+++ b/docker/docker-compose.monitoring.yml
@@ -0,0 +1,170 @@
+services:
+  # ===========================================
+  # OpenTelemetry Collector
+  # ===========================================
+  otel-collector:
+    image: otel/opentelemetry-collector-contrib:0.98.0
+    container_name: speedcam-otel-collector
+    command: ["--config", "/etc/otel-collector-config.yml"]
+    volumes:
+      - ./monitoring/otel-collector/otel-collector-config.yml:/etc/otel-collector-config.yml:ro
+    ports:
+      - "4317:4317"   # OTLP gRPC
+      - "4318:4318"   # OTLP HTTP
+      - "8889:8889"   # Prometheus exporter
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # Jaeger (Distributed Tracing)
+  # ===========================================
+  jaeger:
+    image: jaegertracing/all-in-one:1.57
+    container_name: speedcam-jaeger
+    environment:
+      - COLLECTOR_OTLP_ENABLED=true
+    ports:
+      - "16686:16686"  # Jaeger UI
+      - "14250:14250"  # gRPC (collector)
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # Prometheus (Metrics)
+  # ===========================================
+  prometheus:
+    image: prom/prometheus:v2.51.2
+    container_name: speedcam-prometheus
+    volumes:
+      - ./monitoring/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
+      - ./monitoring/prometheus/data:/prometheus
+    ports:
+      - "9090:9090"
+    command:
+      - "--config.file=/etc/prometheus/prometheus.yml"
+      - "--storage.tsdb.retention.time=15d"
+      - "--web.enable-remote-write-receiver"
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # Grafana (Dashboards)
+  # ===========================================
+  grafana:
+    image: grafana/grafana:10.4.2
+    container_name: speedcam-grafana
+    environment:
+      - GF_SECURITY_ADMIN_USER=admin
+      - GF_SECURITY_ADMIN_PASSWORD=admin
+      - GF_USERS_ALLOW_SIGN_UP=false
+    volumes:
+      - ./monitoring/grafana/provisioning:/etc/grafana/provisioning:ro
+      - ./monitoring/grafana/data:/var/lib/grafana
+    ports:
+      - "3000:3000"
+    depends_on:
+      - prometheus
+      - jaeger
+      - loki
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # Loki (Log Aggregation)
+  # ===========================================
+  loki:
+    image: grafana/loki:2.9.6
+    container_name: speedcam-loki
+    volumes:
+      - ./monitoring/loki/loki-config.yml:/etc/loki/local-config.yaml:ro
+      - ./monitoring/loki/data:/loki
+    ports:
+      - "3100:3100"
+    command: -config.file=/etc/loki/local-config.yaml
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # Promtail (Log Shipper -> Loki)
+  # ===========================================
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    volumes:
+      - ./monitoring/promtail/promtail-config.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+    depends_on:
+      - loki
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # MySQL Exporter
+  # ===========================================
+  mysqld-exporter:
+    image: prom/mysqld-exporter:v0.15.1
+    container_name: speedcam-mysqld-exporter
+    volumes:
+      - ./monitoring/mysqld-exporter/.my.cnf:/cfg/.my.cnf:ro
+    command:
+      - "--config.my-cnf=/cfg/.my.cnf"
+    ports:
+      - "9104:9104"
+    restart: unless-stopped
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # Celery Exporter (Queue/Task Metrics)
+  # ===========================================
+  celery-exporter:
+    image: danihodovic/celery-exporter:0.10.3
+    container_name: speedcam-celery-exporter
+    environment:
+      CE_BROKER_URL: "amqp://sa:1234@rabbitmq:5672//"
+    ports:
+      - "9808:9808"
+    restart: unless-stopped
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # cAdvisor (Container Metrics)
+  # ===========================================
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    profiles:
+      - linux
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
+    ports:
+      - "8080:8080"
+    networks:
+      - speedcam-network
+
+  # ===========================================
+  # K6 Load Test Runner (on-demand)
+  # ===========================================
+  k6:
+    image: grafana/k6:latest
+    container_name: speedcam-k6
+    profiles:
+      - loadtest
+    volumes:
+      - ./k6:/scripts
+    environment:
+      K6_PROMETHEUS_RW_SERVER_URL: http://prometheus:9090/api/v1/write
+      K6_PROMETHEUS_RW_TREND_AS_NATIVE_HISTOGRAM: "true"
+      MAIN_SERVICE_URL: http://main:8000
+    networks:
+      - speedcam-network
+
+networks:
+  speedcam-network:
+    external: true
diff --git a/docker/docker-compose.yml b/docker/docker-compose.yml
index 8d9870a..94270e4 100644
--- a/docker/docker-compose.yml
+++ b/docker/docker-compose.yml
@@ -29,10 +29,11 @@ services:
       - "5672:5672"
       - "1883:1883"
       - "15672:15672"
+      - "15692:15692"
     volumes:
       - rabbitmq_data:/var/lib/rabbitmq
     command: >
-      bash -c "rabbitmq-plugins enable --offline rabbitmq_mqtt &&
+      bash -c "rabbitmq-plugins enable --offline rabbitmq_mqtt rabbitmq_prometheus &&
                rabbitmq-server"
     healthcheck:
       test: ["CMD", "rabbitmq-diagnostics", "check_running"]
@@ -109,26 +110,11 @@ services:
     networks:
       - speedcam-network
 
-  # ===========================================
-  # Monitoring (Optional)
-  # ===========================================
-  datadog-agent:
-    image: gcr.io/datadoghq/agent:7
-    container_name: speedcam-datadog
-    profiles:
-      - monitoring
-    env_file:
-      - ../datadog.env
-    ports:
-      - "8125:8125/udp"
-      - "8126:8126"
-    networks:
-      - speedcam-network
-
 volumes:
   mysql_data:
   rabbitmq_data:
 
 networks:
   speedcam-network:
+    name: speedcam-network
     driver: bridge
diff --git a/docker/k6/load-test.js b/docker/k6/load-test.js
new file mode 100644
index 0000000..999909b
--- /dev/null
+++ b/docker/k6/load-test.js
@@ -0,0 +1,115 @@
+import http from 'k6/http';
+import { check, group, sleep } from 'k6';
+import { Rate, Trend } from 'k6/metrics';
+
+// Custom metrics
+const errorRate = new Rate('errors');
+const vehicleCreateDuration = new Trend('vehicle_create_duration', true);
+
+const BASE_URL = __ENV.MAIN_SERVICE_URL || 'http://main:8000';
+
+export const options = {
+    scenarios: {
+        // Scenario 1: Smoke test (basic connectivity)
+        smoke: {
+            executor: 'constant-vus',
+            vus: 1,
+            duration: '10s',
+            startTime: '0s',
+            tags: { scenario: 'smoke' },
+        },
+        // Scenario 2: Average load
+        average_load: {
+            executor: 'ramping-vus',
+            startVUs: 0,
+            stages: [
+                { duration: '30s', target: 10 },   // ramp up
+                { duration: '1m', target: 10 },     // steady
+                { duration: '10s', target: 0 },     // ramp down
+            ],
+            startTime: '15s',
+            tags: { scenario: 'average_load' },
+        },
+        // Scenario 3: Spike test
+        spike: {
+            executor: 'ramping-vus',
+            startVUs: 0,
+            stages: [
+                { duration: '5s', target: 30 },    // spike up
+                { duration: '15s', target: 30 },   // hold spike
+                { duration: '5s', target: 0 },     // recover
+            ],
+            startTime: '2m',
+            tags: { scenario: 'spike' },
+        },
+    },
+    thresholds: {
+        http_req_duration: ['p(95)<500'],   // 95% of requests under 500ms
+        errors: ['rate<0.1'],                // error rate under 10%
+    },
+};
+
+// Helper to generate random Korean plate number
+function randomPlate() {
+    const nums1 = Math.floor(Math.random() * 900) + 100;
+    const chars = '가나다라마바사아자차카타파하';
+    const char = chars.charAt(Math.floor(Math.random() * chars.length));
+    const nums2 = Math.floor(Math.random() * 9000) + 1000;
+    return `${nums1}${char}${nums2}`;
+}
+
+export default function () {
+    group('Health Check', function () {
+        const res = http.get(`${BASE_URL}/health/`);
+        check(res, {
+            'health status 200': (r) => r.status === 200,
+            'health is healthy': (r) => r.json('status') === 'healthy',
+        });
+        errorRate.add(res.status !== 200);
+    });
+
+    group('Vehicle CRUD', function () {
+        // Create
+        const plate = randomPlate();
+        const createPayload = JSON.stringify({
+            plate_number: plate,
+            owner_name: `테스트유저_${__VU}`,
+            owner_phone: `010-${Math.floor(Math.random() * 9000) + 1000}-${Math.floor(Math.random() * 9000) + 1000}`,
+        });
+
+        const createRes = http.post(`${BASE_URL}/api/v1/vehicles/`, createPayload, {
+            headers: { 'Content-Type': 'application/json' },
+        });
+
+        check(createRes, {
+            'vehicle created 201': (r) => r.status === 201,
+        });
+        errorRate.add(createRes.status !== 201);
+        vehicleCreateDuration.add(createRes.timings.duration);
+
+        // List
+        const listRes = http.get(`${BASE_URL}/api/v1/vehicles/`);
+        check(listRes, {
+            'vehicle list 200': (r) => r.status === 200,
+        });
+        errorRate.add(listRes.status !== 200);
+    });
+
+    group('Detections Read', function () {
+        const res = http.get(`${BASE_URL}/api/v1/detections/`);
+        check(res, {
+            'detections list 200': (r) => r.status === 200,
+        });
+        errorRate.add(res.status !== 200);
+    });
+
+    group('Notifications Read', function () {
+        const res = http.get(`${BASE_URL}/api/v1/notifications/`);
+        check(res, {
+            'notifications list 200': (r) => r.status === 200,
+        });
+        errorRate.add(res.status !== 200);
+    });
+
+    sleep(1);
+}
diff --git a/docker/k6/mqtt-load-test.py b/docker/k6/mqtt-load-test.py
new file mode 100644
index 0000000..89b836e
--- /dev/null
+++ b/docker/k6/mqtt-load-test.py
@@ -0,0 +1,184 @@
+#!/usr/bin/env python3
+"""
+MQTT Load Test - IoT Device Simulation
+
+Simulates Raspberry Pi cameras sending detection messages via MQTT.
+Full pipeline: MQTT → Detection (pending) → OCR Worker → Alert Worker
+"""
+
+import argparse
+import json
+import os
+import random
+import threading
+import time
+from datetime import datetime, timedelta, timezone
+
+import paho.mqtt.client as mqtt
+
+# Config from environment
+MQTT_HOST = os.getenv("MQTT_HOST", "rabbitmq")
+MQTT_PORT = int(os.getenv("MQTT_PORT", "1883"))
+MQTT_USER = os.getenv("MQTT_USER", "sa")
+MQTT_PASS = os.getenv("MQTT_PASS", "1234")
+TOPIC = "detections/new"
+
+# Locations for realistic simulation
+LOCATIONS = [
+    "서울시 강남구 테헤란로",
+    "서울시 서초구 반포대로",
+    "서울시 송파구 올림픽로",
+    "경기도 성남시 분당구 판교역로",
+    "인천시 연수구 송도대로",
+    "서울시 마포구 월드컵북로",
+    "서울시 영등포구 여의대방로",
+    "부산시 해운대구 해운대로",
+]
+
+CAMERA_IDS = [f"CAM-{str(i).zfill(3)}" for i in range(1, 21)]
+
+# Stats
+stats = {
+    "published": 0,
+    "failed": 0,
+    "total_latency_ms": 0,
+    "start_time": None,
+}
+stats_lock = threading.Lock()
+
+
+def generate_message():
+    """Generate a realistic detection message."""
+    kst = timezone(timedelta(hours=9))
+    speed_limit = random.choice([60.0, 80.0, 100.0, 110.0])
+    detected_speed = speed_limit + random.uniform(5, 50)
+
+    return json.dumps(
+        {
+            "camera_id": random.choice(CAMERA_IDS),
+            "location": random.choice(LOCATIONS),
+            "detected_speed": round(detected_speed, 1),
+            "speed_limit": speed_limit,
+            "detected_at": datetime.now(kst).isoformat(),
+            "image_gcs_uri": (
+                f"gs://speedcam-bucket/detections/"
+                f"{int(time.time() * 1000)}-{random.randint(1000, 9999)}.jpg"
+            ),
+        }
+    )
+
+
+def publish_worker(worker_id, rate_per_sec, duration_sec):
+    """Single worker thread that publishes MQTT messages."""
+    client = mqtt.Client(
+        callback_api_version=mqtt.CallbackAPIVersion.VERSION2,
+        protocol=mqtt.MQTTv311,
+        client_id=f"loadtest-{worker_id}-{os.getpid()}",
+    )
+    client.username_pw_set(MQTT_USER, MQTT_PASS)
+
+    try:
+        client.connect(MQTT_HOST, MQTT_PORT, keepalive=60)
+        client.loop_start()
+    except Exception as e:
+        print(f"[Worker-{worker_id}] Connection failed: {e}")
+        with stats_lock:
+            stats["failed"] += 1
+        return
+
+    interval = 1.0 / rate_per_sec if rate_per_sec > 0 else 1.0
+    end_time = time.time() + duration_sec
+
+    while time.time() < end_time:
+        msg = generate_message()
+        start = time.time()
+        result = client.publish(TOPIC, msg, qos=1)
+
+        if result.rc == mqtt.MQTT_ERR_SUCCESS:
+            latency_ms = (time.time() - start) * 1000
+            with stats_lock:
+                stats["published"] += 1
+                stats["total_latency_ms"] += latency_ms
+        else:
+            with stats_lock:
+                stats["failed"] += 1
+
+        elapsed = time.time() - start
+        sleep_time = max(0, interval - elapsed)
+        if sleep_time > 0:
+            time.sleep(sleep_time)
+
+    client.loop_stop()
+    client.disconnect()
+
+
+def print_stats():
+    """Print periodic stats."""
+    elapsed = time.time() - stats["start_time"]
+    published = stats["published"]
+    failed = stats["failed"]
+    total = published + failed
+    rate = published / elapsed if elapsed > 0 else 0
+    avg_latency = stats["total_latency_ms"] / published if published > 0 else 0
+
+    print(f"\n{'='*60}")
+    print(f"  Elapsed: {elapsed:.1f}s | Published: {published} | Failed: {failed}")
+    print(f"  Rate: {rate:.1f} msg/s | Avg Latency: {avg_latency:.2f}ms")
+    print(f"  Error Rate: {(failed/total*100) if total > 0 else 0:.2f}%")
+    print(f"{'='*60}")
+
+
+def run_load_test(workers, rate_per_worker, duration):
+    """Run the load test with multiple workers."""
+    print("\n MQTT Load Test Starting")
+    print(f"  Host: {MQTT_HOST}:{MQTT_PORT}")
+    print(f"  Workers: {workers}")
+    print(
+        f"  Rate: {rate_per_worker}/s per worker ({workers * rate_per_worker}/s total)"
+    )
+    print(f"  Duration: {duration}s")
+    print(f"  Topic: {TOPIC}")
+    print()
+
+    stats["start_time"] = time.time()
+    threads = []
+
+    for i in range(workers):
+        t = threading.Thread(
+            target=publish_worker,
+            args=(i, rate_per_worker, duration),
+        )
+        t.start()
+        threads.append(t)
+
+    # Print stats periodically
+    monitor_end = time.time() + duration
+    while time.time() < monitor_end:
+        time.sleep(5)
+        print_stats()
+
+    for t in threads:
+        t.join(timeout=10)
+
+    print("\n FINAL RESULTS")
+    print_stats()
+
+
+def main():
+    parser = argparse.ArgumentParser(description="MQTT Load Test")
+    parser.add_argument(
+        "--workers", type=int, default=5, help="Number of concurrent workers"
+    )
+    parser.add_argument(
+        "--rate", type=int, default=2, help="Messages per second per worker"
+    )
+    parser.add_argument(
+        "--duration", type=int, default=60, help="Test duration in seconds"
+    )
+    args = parser.parse_args()
+
+    run_load_test(args.workers, args.rate, args.duration)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/docker/monitoring/grafana/provisioning/dashboards/dashboards.yml b/docker/monitoring/grafana/provisioning/dashboards/dashboards.yml
new file mode 100644
index 0000000..3063794
--- /dev/null
+++ b/docker/monitoring/grafana/provisioning/dashboards/dashboards.yml
@@ -0,0 +1,12 @@
+apiVersion: 1
+
+providers:
+  - name: "default"
+    orgId: 1
+    folder: ""
+    type: file
+    disableDeletion: false
+    editable: true
+    options:
+      path: /etc/grafana/provisioning/dashboards
+      foldersFromFilesStructure: false
diff --git a/docker/monitoring/grafana/provisioning/datasources/datasources.yml b/docker/monitoring/grafana/provisioning/datasources/datasources.yml
new file mode 100644
index 0000000..19e94ac
--- /dev/null
+++ b/docker/monitoring/grafana/provisioning/datasources/datasources.yml
@@ -0,0 +1,29 @@
+apiVersion: 1
+
+datasources:
+  - name: Prometheus
+    type: prometheus
+    access: proxy
+    url: http://prometheus:9090
+    isDefault: true
+    editable: false
+
+  - name: Jaeger
+    type: jaeger
+    uid: jaeger
+    access: proxy
+    url: http://jaeger:16686
+    editable: false
+
+  - name: Loki
+    type: loki
+    access: proxy
+    url: http://loki:3100
+    editable: false
+    jsonData:
+      derivedFields:
+        - datasourceUid: jaeger
+          matcherRegex: "trace_id=(\\w+)"
+          name: TraceID
+          url: "$${__value.raw}"
+          datasourceName: Jaeger
diff --git a/docker/monitoring/loki/loki-config.yml b/docker/monitoring/loki/loki-config.yml
new file mode 100644
index 0000000..39c9a79
--- /dev/null
+++ b/docker/monitoring/loki/loki-config.yml
@@ -0,0 +1,37 @@
+auth_enabled: false
+
+server:
+  http_listen_port: 3100
+
+common:
+  path_prefix: /loki
+  storage:
+    filesystem:
+      chunks_directory: /loki/chunks
+      rules_directory: /loki/rules
+  replication_factor: 1
+  ring:
+    kvstore:
+      store: inmemory
+
+schema_config:
+  configs:
+    - from: "2024-01-01"
+      store: tsdb
+      object_store: filesystem
+      schema: v13
+      index:
+        prefix: index_
+        period: 24h
+
+limits_config:
+  retention_period: 168h  # 7 days
+  max_global_streams_per_user: 10000
+  ingestion_burst_size_mb: 16
+  ingestion_rate_mb: 8
+
+compactor:
+  working_directory: /loki/compactor
+  compaction_interval: 10m
+  retention_enabled: true
+  retention_delete_delay: 2h
diff --git a/docker/monitoring/mysqld-exporter/.my.cnf b/docker/monitoring/mysqld-exporter/.my.cnf
new file mode 100644
index 0000000..d314180
--- /dev/null
+++ b/docker/monitoring/mysqld-exporter/.my.cnf
@@ -0,0 +1,5 @@
+[client]
+user=sa
+password=1234
+host=mysql
+port=3306
diff --git a/docker/monitoring/otel-collector/otel-collector-config.yml b/docker/monitoring/otel-collector/otel-collector-config.yml
new file mode 100644
index 0000000..56d2256
--- /dev/null
+++ b/docker/monitoring/otel-collector/otel-collector-config.yml
@@ -0,0 +1,39 @@
+receivers:
+  otlp:
+    protocols:
+      grpc:
+        endpoint: 0.0.0.0:4317
+      http:
+        endpoint: 0.0.0.0:4318
+
+processors:
+  batch:
+    timeout: 5s
+    send_batch_size: 1024
+  resource:
+    attributes:
+      - key: service.namespace
+        value: speedcam
+        action: upsert
+
+exporters:
+  otlp/jaeger:
+    endpoint: jaeger:4317
+    tls:
+      insecure: true
+  prometheus:
+    endpoint: 0.0.0.0:8889
+    namespace: speedcam
+    resource_to_telemetry_conversion:
+      enabled: true
+
+service:
+  pipelines:
+    traces:
+      receivers: [otlp]
+      processors: [batch, resource]
+      exporters: [otlp/jaeger]
+    metrics:
+      receivers: [otlp]
+      processors: [batch, resource]
+      exporters: [prometheus]
diff --git a/docker/monitoring/prometheus/prometheus.yml b/docker/monitoring/prometheus/prometheus.yml
new file mode 100644
index 0000000..39a18f9
--- /dev/null
+++ b/docker/monitoring/prometheus/prometheus.yml
@@ -0,0 +1,34 @@
+global:
+  scrape_interval: 15s
+  evaluation_interval: 15s
+
+scrape_configs:
+  # --- Application ---
+  - job_name: "django"
+    metrics_path: /metrics
+    static_configs:
+      - targets: ["main:8000"]
+
+  # --- OpenTelemetry Collector ---
+  - job_name: "otel-collector"
+    static_configs:
+      - targets: ["otel-collector:8889"]
+
+  # --- Infrastructure ---
+  - job_name: "rabbitmq"
+    static_configs:
+      - targets: ["rabbitmq:15692"]
+
+  - job_name: "mysql"
+    static_configs:
+      - targets: ["mysqld-exporter:9104"]
+
+  # --- Workers ---
+  - job_name: "celery"
+    static_configs:
+      - targets: ["celery-exporter:9808"]
+
+  # --- Container Resources ---
+  - job_name: "cadvisor"
+    static_configs:
+      - targets: ["cadvisor:8080"]
diff --git a/docker/monitoring/promtail/promtail-config.yml b/docker/monitoring/promtail/promtail-config.yml
new file mode 100644
index 0000000..4a64917
--- /dev/null
+++ b/docker/monitoring/promtail/promtail-config.yml
@@ -0,0 +1,31 @@
+server:
+  http_listen_port: 9080
+  grpc_listen_port: 0
+
+positions:
+  filename: /tmp/positions.yaml
+
+clients:
+  - url: http://loki:3100/loki/api/v1/push
+
+scrape_configs:
+  - job_name: docker
+    docker_sd_configs:
+      - host: unix:///var/run/docker.sock
+        refresh_interval: 5s
+        filters:
+          - name: name
+            values:
+              - "speedcam-.*"
+    relabel_configs:
+      - source_labels: ["__meta_docker_container_name"]
+        regex: "/(.*)"
+        target_label: "container"
+      - source_labels: ["__meta_docker_container_name"]
+        regex: "/speedcam-(.*)"
+        target_label: "service"
+    pipeline_stages:
+      - regex:
+          expression: ".*trace_id=(?P<trace_id>[a-fA-F0-9]+).*"
+      - labels:
+          trace_id:
diff --git a/docker/rabbitmq/enabled_plugins b/docker/rabbitmq/enabled_plugins
deleted file mode 100644
index 5358cb0..0000000
--- a/docker/rabbitmq/enabled_plugins
+++ /dev/null
@@ -1 +0,0 @@
-[rabbitmq_management,rabbitmq_mqtt].
diff --git a/docker/rabbitmq/rabbitmq.conf b/docker/rabbitmq/rabbitmq.conf
deleted file mode 100644
index 8192f11..0000000
--- a/docker/rabbitmq/rabbitmq.conf
+++ /dev/null
@@ -1,13 +0,0 @@
-# MQTT Plugin 설정
-mqtt.listeners.tcp.default = 1883
-mqtt.allow_anonymous = false
-mqtt.default_user = sa
-mqtt.default_pass = 1234
-mqtt.vhost = /
-mqtt.exchange = amq.topic
-mqtt.subscription_ttl = 86400000
-mqtt.prefetch = 10
-
-# Management Plugin
-management.tcp.port = 15672
-
diff --git a/docs/DEPLOYMENT.md b/docs/DEPLOYMENT.md
index e1af5e8..18cbdc5 100644
--- a/docs/DEPLOYMENT.md
+++ b/docs/DEPLOYMENT.md
@@ -1,12 +1,19 @@
-# GCP 배포 가이드
+# GCP 멀티 인스턴스 배포 가이드
 
 ## 목차
+
 1. [사전 요구사항](#1-사전-요구사항)
 2. [시스템 아키텍처](#2-시스템-아키텍처)
-3. [배포 단계](#3-배포-단계)
-4. [배포 검증](#4-배포-검증)
-5. [운영 가이드](#5-운영-가이드)
-6. [트러블슈팅](#6-트러블슈팅)
+3. [GCP 인프라 설정](#3-gcp-인프라-설정)
+4. [배포 디렉토리 구조](#4-배포-디렉토리-구조)
+5. [Docker Compose 파일](#5-docker-compose-파일)
+6. [설정 파일](#6-설정-파일)
+7. [Docker 이미지 빌드 및 배포](#7-docker-이미지-빌드-및-배포)
+8. [배포 순서](#8-배포-순서)
+9. [배포 검증](#9-배포-검증)
+10. [운영 가이드](#10-운영-가이드)
+11. [트러블슈팅](#11-트러블슈팅)
+12. [리소스 정리](#12-리소스-정리)
 
 ---
 
@@ -18,8 +25,7 @@
 |------|------|----------|
 | Google Cloud SDK | 최신 | https://cloud.google.com/sdk/docs/install |
 | Docker | 20.10+ | https://docs.docker.com/get-docker/ |
-| Terraform | 1.5+ | https://developer.hashicorp.com/terraform/downloads |
-| Make | 3.81+ | 기본 설치됨 (macOS/Linux) |
+| Docker Compose | 2.0+ | Docker Desktop 포함 또는 별도 설치 |
 
 ### 1.2 GCP 프로젝트 설정
 
@@ -39,12 +45,12 @@ gcloud services enable artifactregistry.googleapis.com
 
 ```bash
 # 프로젝트 설정
-export GCP_PROJECT_ID=your-project-id
+export GCP_PROJECT_ID=<your-project-id>
 export GCP_REGION=asia-northeast3
 export GCP_ZONE=asia-northeast3-a
 
-# Docker Registry
-export REGISTRY=${GCP_REGION}-docker.pkg.dev/${GCP_PROJECT_ID}/speedcam
+# Docker Registry (Artifact Registry 사용)
+export ARTIFACT_REGISTRY=${GCP_REGION}-docker.pkg.dev/${GCP_PROJECT_ID}/speedcam
 ```
 
 ---
@@ -61,13 +67,14 @@ graph TB
     end
 
     subgraph GCP["Google Cloud Platform (asia-northeast3)"]
-        subgraph VPC["VPC Network (default)"]
+        subgraph VPC["VPC Network"]
             subgraph Compute["GCE Instances"]
-                RMQ[speedcam-rabbitmq<br/>e2-small<br/>10.178.0.2]
-                MySQL[speedcam-mysql<br/>e2-small<br/>10.178.0.3]
-                Main[speedcam-main<br/>e2-medium<br/>10.178.0.4]
-                OCR[speedcam-ocr<br/>e2-medium<br/>10.178.0.5]
-                Alert[speedcam-alert<br/>e2-small<br/>10.178.0.6]
+                DB[speedcam-db<br/>e2-highmem-2<br/>MySQL + Exporters]
+                MQ[speedcam-mq<br/>e2-medium<br/>RabbitMQ]
+                App[speedcam-app<br/>e2-medium<br/>Django API + Flower]
+                OCR[speedcam-ocr<br/>e2-standard-2<br/>OCR Worker]
+                Alert[speedcam-alert<br/>e2-small<br/>Alert Worker]
+                Mon[speedcam-mon<br/>e2-standard-2<br/>Monitoring Stack]
             end
         end
 
@@ -79,22 +86,33 @@ graph TB
         FCM[Firebase FCM]
     end
 
-    Client -->|HTTP :8000| Main
-    Pi -->|MQTT :1883| RMQ
+    Client -->|HTTP :8000| App
+    Pi -->|MQTT :1883| MQ
     Pi -->|Upload| GCS
 
-    Main -->|AMQP :5672| RMQ
-    Main -->|MySQL :3306| MySQL
+    App -->|AMQP :5672| MQ
+    App -->|MySQL :3306| DB
 
-    OCR -->|AMQP| RMQ
-    OCR -->|MySQL| MySQL
+    OCR -->|AMQP| MQ
+    OCR -->|MySQL| DB
     OCR -->|Download| GCS
 
-    Alert -->|AMQP| RMQ
-    Alert -->|MySQL| MySQL
+    Alert -->|AMQP| MQ
+    Alert -->|MySQL| DB
     Alert -->|Push| FCM
 
-    AR -.->|Pull Image| Main
+    Mon -->|Scrape Metrics| DB
+    Mon -->|Scrape Metrics| MQ
+    Mon -->|Scrape Metrics| App
+    Mon -->|Scrape Metrics| OCR
+    Mon -->|Scrape Metrics| Alert
+    Mon -->|Collect Logs| DB
+    Mon -->|Collect Logs| MQ
+    Mon -->|Collect Logs| App
+    Mon -->|Collect Logs| OCR
+    Mon -->|Collect Logs| Alert
+
+    AR -.->|Pull Image| App
     AR -.->|Pull Image| OCR
     AR -.->|Pull Image| Alert
 ```
@@ -182,351 +200,1262 @@ erDiagram
 
 ### 2.4 인스턴스 사양
 
-| 인스턴스 | 역할 | Machine Type | vCPU | Memory | 포트 |
-|----------|------|--------------|------|--------|------|
-| speedcam-rabbitmq | Message Broker | e2-small | 0.5-2 | 2GB | 5672, 1883, 15672 |
-| speedcam-mysql | Database | e2-small | 0.5-2 | 2GB | 3306 |
-| speedcam-main | Django API + MQTT | e2-medium | 1-2 | 4GB | 8000 |
-| speedcam-ocr | OCR Worker (prefork) | e2-medium | 1-2 | 4GB | - |
-| speedcam-alert | Alert Worker (gevent) | e2-small | 0.5-2 | 2GB | - |
+| 인스턴스 이름 | 역할 | 머신 타입 | vCPU | Memory | 열어야 할 포트 (내부) |
+|--------------|------|----------|------|--------|---------------------|
+| `speedcam-db` | MySQL + Exporters | e2-highmem-2 | 2 | 16GB | 3306, 9104, 8080 |
+| `speedcam-mq` | RabbitMQ | e2-medium | 2 | 4GB | 5672, 1883, 15672, 15692, 8080 |
+| `speedcam-app` | Django API + Flower | e2-medium | 2 | 4GB | 8000, 5555, 8080 |
+| `speedcam-ocr` | OCR Celery Worker | e2-standard-2 | 2 | 8GB | 8080 |
+| `speedcam-alert` | Alert Celery Worker | e2-small | 2 | 2GB | 8080 |
+| `speedcam-mon` | 모니터링 전체 스택 | e2-standard-2 | 2 | 8GB | 3000, 9090, 16686, 3100, 4317, 4318, 8889, 9808, 8080 |
 
 ---
 
-## 3. 배포 단계
+## 3. GCP 인프라 설정
+
+### 3.1 VPC 네트워크 생성
+
+```bash
+# VPC 네트워크 생성 (커스텀 모드)
+gcloud compute networks create speedcam-vpc \
+    --subnet-mode=custom \
+    --bgp-routing-mode=regional
+
+# 서브넷 생성 (asia-northeast3)
+gcloud compute networks subnets create speedcam-subnet \
+    --network=speedcam-vpc \
+    --region=${GCP_REGION} \
+    --range=10.178.0.0/20
+```
+
+### 3.2 방화벽 규칙 설정
 
-### 3.1 Step 1: Artifact Registry 설정
+#### 3.2.1 내부 통신 허용
+
+```bash
+# 내부 통신 허용 (VPC 내부에서만)
+gcloud compute firewall-rules create speedcam-internal \
+    --network=speedcam-vpc \
+    --allow=tcp:3306,tcp:5672,tcp:1883,tcp:15672,tcp:15692,tcp:8000,tcp:5555,tcp:4317,tcp:4318,tcp:8889,tcp:9090,tcp:3000,tcp:16686,tcp:3100,tcp:9104,tcp:9808,tcp:8080,tcp:9080 \
+    --source-ranges=10.178.0.0/20 \
+    --target-tags=speedcam \
+    --description="SpeedCam internal communication"
+```
+
+#### 3.2.2 외부 접근 허용 (필요한 서비스만)
+
+```bash
+# Django API 외부 접근 (프론트엔드)
+gcloud compute firewall-rules create speedcam-api-external \
+    --network=speedcam-vpc \
+    --allow=tcp:8000 \
+    --source-ranges=0.0.0.0/0 \
+    --target-tags=speedcam-app \
+    --description="Django API external access"
+
+# MQTT 외부 접근 (Edge Device)
+gcloud compute firewall-rules create speedcam-mqtt-external \
+    --network=speedcam-vpc \
+    --allow=tcp:1883 \
+    --source-ranges=0.0.0.0/0 \
+    --target-tags=speedcam-mq \
+    --description="MQTT external access for edge devices"
+
+# Grafana UI 외부 접근 (운영자만)
+gcloud compute firewall-rules create speedcam-grafana-external \
+    --network=speedcam-vpc \
+    --allow=tcp:3000 \
+    --source-ranges=<ADMIN_IP>/32 \
+    --target-tags=speedcam-mon \
+    --description="Grafana external access (admin only)"
+```
+
+### 3.3 Artifact Registry 설정
 
 ```bash
 # 저장소 생성
 gcloud artifacts repositories create speedcam \
-  --repository-format=docker \
-  --location=${GCP_REGION} \
-  --description="Speedcam MSA Docker images"
+    --repository-format=docker \
+    --location=${GCP_REGION} \
+    --description="Speedcam MSA Docker images"
 
 # Docker 인증 설정
 gcloud auth configure-docker ${GCP_REGION}-docker.pkg.dev
 ```
 
-### 3.2 Step 2: Docker 이미지 빌드
+### 3.4 GCE 인스턴스 생성
 
 ```bash
-# linux/amd64 플랫폼으로 빌드 (GCE용)
-docker build --platform linux/amd64 \
-  -t ${REGISTRY}/main:latest \
-  -f docker/Dockerfile.main .
+# 1. speedcam-db 인스턴스
+gcloud compute instances create speedcam-db \
+    --zone=${GCP_ZONE} \
+    --machine-type=e2-highmem-2 \
+    --network-interface=subnet=speedcam-subnet,no-address \
+    --tags=speedcam \
+    --metadata=startup-script='#!/bin/bash
+apt-get update
+apt-get install -y docker.io docker-compose
+systemctl start docker
+systemctl enable docker'
+
+# 2. speedcam-mq 인스턴스
+gcloud compute instances create speedcam-mq \
+    --zone=${GCP_ZONE} \
+    --machine-type=e2-medium \
+    --network-interface=subnet=speedcam-subnet,no-address \
+    --tags=speedcam,speedcam-mq \
+    --metadata=startup-script='#!/bin/bash
+apt-get update
+apt-get install -y docker.io docker-compose
+systemctl start docker
+systemctl enable docker'
+
+# 3. speedcam-app 인스턴스
+gcloud compute instances create speedcam-app \
+    --zone=${GCP_ZONE} \
+    --machine-type=e2-medium \
+    --network-interface=subnet=speedcam-subnet,no-address \
+    --tags=speedcam,speedcam-app \
+    --scopes=cloud-platform \
+    --metadata=startup-script='#!/bin/bash
+apt-get update
+apt-get install -y docker.io docker-compose
+systemctl start docker
+systemctl enable docker'
+
+# 4. speedcam-ocr 인스턴스
+gcloud compute instances create speedcam-ocr \
+    --zone=${GCP_ZONE} \
+    --machine-type=e2-standard-2 \
+    --network-interface=subnet=speedcam-subnet,no-address \
+    --tags=speedcam \
+    --scopes=cloud-platform \
+    --metadata=startup-script='#!/bin/bash
+apt-get update
+apt-get install -y docker.io docker-compose
+systemctl start docker
+systemctl enable docker'
+
+# 5. speedcam-alert 인스턴스
+gcloud compute instances create speedcam-alert \
+    --zone=${GCP_ZONE} \
+    --machine-type=e2-small \
+    --network-interface=subnet=speedcam-subnet,no-address \
+    --tags=speedcam \
+    --scopes=cloud-platform \
+    --metadata=startup-script='#!/bin/bash
+apt-get update
+apt-get install -y docker.io docker-compose
+systemctl start docker
+systemctl enable docker'
+
+# 6. speedcam-mon 인스턴스
+gcloud compute instances create speedcam-mon \
+    --zone=${GCP_ZONE} \
+    --machine-type=e2-standard-2 \
+    --network-interface=subnet=speedcam-subnet,no-address \
+    --tags=speedcam,speedcam-mon \
+    --metadata=startup-script='#!/bin/bash
+apt-get update
+apt-get install -y docker.io docker-compose
+systemctl start docker
+systemctl enable docker'
+```
 
-docker build --platform linux/amd64 \
-  -t ${REGISTRY}/ocr:latest \
-  -f docker/Dockerfile.ocr .
+---
 
-docker build --platform linux/amd64 \
-  -t ${REGISTRY}/alert:latest \
-  -f docker/Dockerfile.alert .
+## 4. 배포 디렉토리 구조
+
+각 인스턴스에 배포할 파일 구조:
+
+```
+deploy/
+├── env/
+│   ├── backend.env              # Django/Celery 공통 환경변수
+│   ├── mysql.env                # MySQL 전용 (DB 인스턴스만)
+│   └── rabbitmq.env             # RabbitMQ 전용 (MQ 인스턴스만)
+├── compose/
+│   ├── docker-compose.db.yml
+│   ├── docker-compose.mq.yml
+│   ├── docker-compose.app.yml
+│   ├── docker-compose.ocr.yml
+│   ├── docker-compose.alert.yml
+│   └── docker-compose.mon.yml
+├── config/
+│   ├── mysql/
+│   │   └── init.sql
+│   ├── monitoring/
+│   │   ├── otel-collector/
+│   │   │   └── otel-collector-config.yml
+│   │   ├── prometheus/
+│   │   │   └── prometheus.yml
+│   │   ├── loki/
+│   │   │   └── loki-config.yml
+│   │   ├── promtail/
+│   │   │   ├── promtail-config.yml       # 각 인스턴스용
+│   │   │   └── promtail-config.mon.yml   # 모니터링 인스턴스용
+│   │   ├── grafana/
+│   │   │   └── provisioning/
+│   │   │       ├── datasources/
+│   │   │       │   └── datasources.yml
+│   │   │       └── dashboards/
+│   │   │           └── dashboards.yml
+│   │   └── mysqld-exporter/
+│   │       └── .my.cnf
+│   └── credentials/
+│       └── (GCP, Firebase 인증 파일)
+└── images/
+    └── (빌드된 이미지는 Artifact Registry 사용)
+```
+
+---
+
+## 5. Docker Compose 파일
+
+### 5.1 로컬 → 멀티 인스턴스 주요 변경점
+
+| 항목 | 로컬 (현재) | 멀티 인스턴스 |
+|------|------------|--------------|
+| 네트워크 | `networks: speedcam-network` (bridge) | `network_mode: host` |
+| 서비스 디스커버리 | 컨테이너명 (`mysql`, `rabbitmq`) | GCP 내부 IP |
+| 포트 매핑 | `ports: "3306:3306"` | 불필요 (host 모드에서 직접 바인딩) |
+| depends_on | 서비스 간 의존성 | 제거 (다른 인스턴스에 있으므로) |
+| 이미지 | `build: context` | Artifact Registry 이미지 |
+| cAdvisor | 모니터링에 1개 | 모든 인스턴스에 1개씩 |
+| Promtail | 모니터링에 1개 | 모든 인스턴스에 1개씩 |
+
+### 5.2 docker-compose.db.yml
+
+speedcam-db 인스턴스에 배포
+
+```yaml
+services:
+  mysql:
+    image: mysql:8.0
+    container_name: speedcam-mysql
+    restart: always
+    network_mode: host
+    env_file:
+      - ../env/mysql.env
+    volumes:
+      - mysql_data:/var/lib/mysql
+      - ../config/mysql/init.sql:/docker-entrypoint-initdb.d/init.sql:ro
+    healthcheck:
+      test: ["CMD", "mysqladmin", "ping", "-h", "localhost"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
+  mysqld-exporter:
+    image: prom/mysqld-exporter:v0.15.1
+    container_name: speedcam-mysqld-exporter
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/mysqld-exporter/.my.cnf:/cfg/.my.cnf:ro
+    command:
+      - "--config.my-cnf=/cfg/.my.cnf"
+    depends_on:
+      mysql:
+        condition: service_healthy
+
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/promtail/promtail-config.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    restart: always
+    network_mode: host
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
+
+volumes:
+  mysql_data:
 ```
 
-### 3.3 Step 3: Docker 이미지 푸시
+### 5.3 docker-compose.mq.yml
+
+speedcam-mq 인스턴스에 배포
+
+```yaml
+services:
+  rabbitmq:
+    image: rabbitmq:3.13-management
+    container_name: speedcam-rabbitmq
+    restart: always
+    network_mode: host
+    env_file:
+      - ../env/rabbitmq.env
+    volumes:
+      - rabbitmq_data:/var/lib/rabbitmq
+    command: >
+      bash -c "rabbitmq-plugins enable --offline rabbitmq_mqtt rabbitmq_prometheus &&
+               rabbitmq-server"
+    healthcheck:
+      test: ["CMD", "rabbitmq-diagnostics", "check_running"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/promtail/promtail-config.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    restart: always
+    network_mode: host
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
+
+volumes:
+  rabbitmq_data:
+```
 
-```bash
-docker push ${REGISTRY}/main:latest
-docker push ${REGISTRY}/ocr:latest
-docker push ${REGISTRY}/alert:latest
+### 5.4 docker-compose.app.yml
+
+speedcam-app 인스턴스에 배포 (이미지는 Artifact Registry에서 pull)
+
+```yaml
+services:
+  main:
+    image: ${ARTIFACT_REGISTRY}/speedcam-main:latest
+    container_name: speedcam-main
+    restart: always
+    network_mode: host
+    env_file:
+      - ../env/backend.env
+    volumes:
+      - ../config/credentials:/app/credentials:ro
+
+  flower:
+    image: ${ARTIFACT_REGISTRY}/speedcam-main:latest
+    container_name: speedcam-flower
+    restart: always
+    network_mode: host
+    env_file:
+      - ../env/backend.env
+    command: celery -A config flower --port=5555
+
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/promtail/promtail-config.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    restart: always
+    network_mode: host
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
 ```
 
-### 3.4 Step 4: 방화벽 규칙 생성
+### 5.5 docker-compose.ocr.yml
+
+speedcam-ocr 인스턴스에 배포
+
+```yaml
+services:
+  ocr-worker:
+    image: ${ARTIFACT_REGISTRY}/speedcam-ocr:latest
+    container_name: speedcam-ocr
+    restart: always
+    network_mode: host
+    env_file:
+      - ../env/backend.env
+    volumes:
+      - ../config/credentials:/app/credentials:ro
+
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/promtail/promtail-config.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    restart: always
+    network_mode: host
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
+```
 
-```bash
-# 내부 통신용
-gcloud compute firewall-rules create speedcam-internal \
-  --network=default \
-  --allow=tcp:3306,tcp:5672,tcp:1883,tcp:15672,tcp:8000 \
-  --source-ranges=10.0.0.0/8 \
-  --target-tags=speedcam
+### 5.6 docker-compose.alert.yml
+
+speedcam-alert 인스턴스에 배포
+
+```yaml
+services:
+  alert-worker:
+    image: ${ARTIFACT_REGISTRY}/speedcam-alert:latest
+    container_name: speedcam-alert
+    restart: always
+    network_mode: host
+    env_file:
+      - ../env/backend.env
+    volumes:
+      - ../config/credentials:/app/credentials:ro
+
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/promtail/promtail-config.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    restart: always
+    network_mode: host
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
+```
 
-# 외부 접근용
-gcloud compute firewall-rules create speedcam-external \
-  --network=default \
-  --allow=tcp:8000,tcp:15672 \
-  --source-ranges=0.0.0.0/0 \
-  --target-tags=speedcam-web
+### 5.7 docker-compose.mon.yml
+
+speedcam-mon 인스턴스에 배포
+
+```yaml
+services:
+  otel-collector:
+    image: otel/opentelemetry-collector-contrib:0.98.0
+    container_name: speedcam-otel-collector
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/otel-collector/otel-collector-config.yml:/etc/otel-collector-config.yml:ro
+    command: ["--config", "/etc/otel-collector-config.yml"]
+
+  jaeger:
+    image: jaegertracing/all-in-one:1.57
+    container_name: speedcam-jaeger
+    restart: always
+    network_mode: host
+    environment:
+      - COLLECTOR_OTLP_ENABLED=true
+
+  prometheus:
+    image: prom/prometheus:v2.51.2
+    container_name: speedcam-prometheus
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
+      - prometheus_data:/prometheus
+    command:
+      - "--config.file=/etc/prometheus/prometheus.yml"
+      - "--storage.tsdb.retention.time=15d"
+      - "--web.enable-remote-write-receiver"
+
+  grafana:
+    image: grafana/grafana:10.4.2
+    container_name: speedcam-grafana
+    restart: always
+    network_mode: host
+    environment:
+      - GF_SECURITY_ADMIN_USER=admin
+      - GF_SECURITY_ADMIN_PASSWORD=${GRAFANA_PASSWORD:-admin}
+      - GF_USERS_ALLOW_SIGN_UP=false
+    volumes:
+      - ../config/monitoring/grafana/provisioning:/etc/grafana/provisioning:ro
+      - grafana_data:/var/lib/grafana
+
+  loki:
+    image: grafana/loki:2.9.6
+    container_name: speedcam-loki
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/loki/loki-config.yml:/etc/loki/local-config.yaml:ro
+      - loki_data:/loki
+    command: -config.file=/etc/loki/local-config.yaml
+
+  promtail:
+    image: grafana/promtail:2.9.6
+    container_name: speedcam-promtail
+    restart: always
+    network_mode: host
+    volumes:
+      - ../config/monitoring/promtail/promtail-config.mon.yml:/etc/promtail/config.yml:ro
+      - /var/run/docker.sock:/var/run/docker.sock:ro
+      - /var/lib/docker/containers:/var/lib/docker/containers:ro
+    command: -config.file=/etc/promtail/config.yml
+
+  celery-exporter:
+    image: danihodovic/celery-exporter:0.10.3
+    container_name: speedcam-celery-exporter
+    restart: always
+    network_mode: host
+    environment:
+      CE_BROKER_URL: "amqp://sa:<password>@${MQ_HOST}:5672//"
+
+  cadvisor:
+    image: gcr.io/cadvisor/cadvisor:v0.49.1
+    container_name: speedcam-cadvisor
+    restart: always
+    network_mode: host
+    volumes:
+      - /:/rootfs:ro
+      - /var/run:/var/run:ro
+      - /sys:/sys:ro
+      - /var/lib/docker/:/var/lib/docker:ro
+
+volumes:
+  prometheus_data:
+  grafana_data:
+  loki_data:
 ```
 
-### 3.5 Step 5: 인프라 인스턴스 생성
+---
+
+## 6. 설정 파일
+
+### 6.1 환경 변수 파일
 
-#### RabbitMQ
+#### 6.1.1 backend.env
+
+모든 앱/워커 인스턴스에 배포. 컨테이너명을 IP로 교체.
 
 ```bash
-gcloud compute instances create-with-container speedcam-rabbitmq \
-  --zone=${GCP_ZONE} \
-  --machine-type=e2-small \
-  --tags=speedcam,speedcam-web \
-  --container-image=rabbitmq:3.13-management \
-  --container-env="RABBITMQ_DEFAULT_USER=sa,RABBITMQ_DEFAULT_PASS=1234"
+# Django
+SECRET_KEY=<production-secret-key>
+DJANGO_SETTINGS_MODULE=config.settings.prod
+DEBUG=False
+
+# Database — ${DB_HOST}를 실제 IP로 교체
+DB_HOST=${DB_HOST}
+DB_PORT=3306
+DB_USER=sa
+DB_PASSWORD=<production-password>
+DB_NAME=speedcam
+DB_NAME_VEHICLES=speedcam_vehicles
+DB_NAME_DETECTIONS=speedcam_detections
+DB_NAME_NOTIFICATIONS=speedcam_notifications
+
+# RabbitMQ — ${MQ_HOST}를 실제 IP로 교체
+CELERY_BROKER_URL=amqp://sa:<password>@${MQ_HOST}:5672//
+RABBITMQ_HOST=${MQ_HOST}
+MQTT_PORT=1883
+MQTT_USER=sa
+MQTT_PASS=<password>
+
+# GCS / Firebase
+GOOGLE_APPLICATION_CREDENTIALS=/app/credentials/gcp-cloud-storage.json
+FIREBASE_CREDENTIALS=/app/credentials/firebase-service-account.json
+
+# Workers
+OCR_CONCURRENCY=4
+ALERT_CONCURRENCY=100
+OCR_MOCK=false
+FCM_MOCK=false
+
+# Gunicorn
+GUNICORN_WORKERS=4
+GUNICORN_THREADS=2
+
+# Logging
+LOG_LEVEL=info
+
+# CORS — 프론트엔드 도메인으로 교체
+CORS_ALLOWED_ORIGINS=https://your-frontend-domain.com
+
+# OpenTelemetry — ${MON_HOST}를 실제 IP로 교체
+OTEL_EXPORTER_OTLP_ENDPOINT=http://${MON_HOST}:4317
+OTEL_EXPORTER_OTLP_PROTOCOL=grpc
+OTEL_RESOURCE_ATTRIBUTES=service.namespace=speedcam,deployment.environment=prod
+OTEL_TRACES_SAMPLER=parentbased_always_on
+OTEL_PYTHON_LOG_CORRELATION=true
 ```
 
-#### MySQL
+#### 6.1.2 mysql.env
+
+DB 인스턴스에 배포.
 
 ```bash
-gcloud compute instances create-with-container speedcam-mysql \
-  --zone=${GCP_ZONE} \
-  --machine-type=e2-small \
-  --tags=speedcam \
-  --container-image=mysql:8.0 \
-  --container-env="MYSQL_ROOT_PASSWORD=root,MYSQL_USER=sa,MYSQL_PASSWORD=1234,MYSQL_DATABASE=speedcam"
+MYSQL_ROOT_PASSWORD=<production-root-password>
+MYSQL_USER=sa
+MYSQL_PASSWORD=<production-password>
+MYSQL_DATABASE=speedcam
 ```
 
-### 3.6 Step 6: 인프라 초기화
+#### 6.1.3 rabbitmq.env
+
+MQ 인스턴스에 배포.
 
 ```bash
-# RabbitMQ MQTT 플러그인 활성화
-gcloud compute ssh speedcam-rabbitmq --zone=${GCP_ZONE} \
-  --command="docker exec \$(docker ps -q) rabbitmq-plugins enable rabbitmq_mqtt"
+RABBITMQ_DEFAULT_USER=sa
+RABBITMQ_DEFAULT_PASS=<production-password>
+```
 
-# MySQL 추가 데이터베이스 생성
-gcloud compute ssh speedcam-mysql --zone=${GCP_ZONE} \
-  --command="docker exec \$(docker ps -q) mysql -u root -proot -e \"
-    CREATE DATABASE IF NOT EXISTS speedcam_vehicles;
-    CREATE DATABASE IF NOT EXISTS speedcam_detections;
-    CREATE DATABASE IF NOT EXISTS speedcam_notifications;
-    GRANT ALL PRIVILEGES ON speedcam_vehicles.* TO 'sa'@'%';
-    GRANT ALL PRIVILEGES ON speedcam_detections.* TO 'sa'@'%';
-    GRANT ALL PRIVILEGES ON speedcam_notifications.* TO 'sa'@'%';
-    FLUSH PRIVILEGES;\""
+### 6.2 모니터링 설정
+
+#### 6.2.1 prometheus.yml
+
+모니터링 인스턴스에 배포. 모든 타겟을 실제 IP로 지정.
+
+```yaml
+global:
+  scrape_interval: 15s
+  evaluation_interval: 15s
+
+scrape_configs:
+  # --- Application ---
+  - job_name: "django"
+    metrics_path: /metrics
+    static_configs:
+      - targets: ["${APP_HOST}:8000"]
+
+  # --- OpenTelemetry Collector ---
+  - job_name: "otel-collector"
+    static_configs:
+      - targets: ["localhost:8889"]
+
+  # --- Infrastructure ---
+  - job_name: "rabbitmq"
+    static_configs:
+      - targets: ["${MQ_HOST}:15692"]
+
+  - job_name: "mysql"
+    static_configs:
+      - targets: ["${DB_HOST}:9104"]
+
+  # --- Workers ---
+  - job_name: "celery"
+    static_configs:
+      - targets: ["localhost:9808"]
+
+  # --- Container Resources (모든 인스턴스) ---
+  - job_name: "cadvisor"
+    static_configs:
+      - targets:
+          - "${DB_HOST}:8080"
+          - "${MQ_HOST}:8080"
+          - "${APP_HOST}:8080"
+          - "${OCR_HOST}:8080"
+          - "${ALERT_HOST}:8080"
+          - "localhost:8080"
+    relabel_configs:
+      - source_labels: [__address__]
+        target_label: instance
 ```
 
-### 3.7 Step 7: Internal IP 확인
+#### 6.2.2 otel-collector-config.yml
+
+모니터링 인스턴스에 배포.
+
+```yaml
+receivers:
+  otlp:
+    protocols:
+      grpc:
+        endpoint: 0.0.0.0:4317
+      http:
+        endpoint: 0.0.0.0:4318
+
+processors:
+  batch:
+    timeout: 5s
+    send_batch_size: 1024
+  resource:
+    attributes:
+      - key: service.namespace
+        value: speedcam
+        action: upsert
+
+exporters:
+  otlp/jaeger:
+    endpoint: localhost:4317
+    tls:
+      insecure: true
+  prometheus:
+    endpoint: 0.0.0.0:8889
+    namespace: speedcam
+    resource_to_telemetry_conversion:
+      enabled: true
+
+service:
+  pipelines:
+    traces:
+      receivers: [otlp]
+      processors: [batch, resource]
+      exporters: [otlp/jaeger]
+    metrics:
+      receivers: [otlp]
+      processors: [batch, resource]
+      exporters: [prometheus]
+```
 
-```bash
-# 인스턴스 IP 확인
-gcloud compute instances list --filter="name~speedcam" \
-  --format="table(name,networkInterfaces[0].networkIP)"
+#### 6.2.3 promtail-config.yml (앱/워커/DB/MQ 인스턴스 공통)
+
+각 인스턴스에 배포. Loki 주소를 모니터링 인스턴스 IP로 지정.
+
+```yaml
+server:
+  http_listen_port: 9080
+  grpc_listen_port: 0
+
+positions:
+  filename: /tmp/positions.yaml
+
+clients:
+  - url: http://${MON_HOST}:3100/loki/api/v1/push
+
+scrape_configs:
+  - job_name: docker
+    docker_sd_configs:
+      - host: unix:///var/run/docker.sock
+        refresh_interval: 5s
+        filters:
+          - name: name
+            values:
+              - "speedcam-.*"
+    relabel_configs:
+      - source_labels: ["__meta_docker_container_name"]
+        regex: "/(.*)"
+        target_label: "container"
+      - source_labels: ["__meta_docker_container_name"]
+        regex: "/speedcam-(.*)"
+        target_label: "service"
+    pipeline_stages:
+      - regex:
+          expression: ".*trace_id=(?P<trace_id>[a-f0-9]+).*"
+      - labels:
+          trace_id:
+```
 
-# 예시 출력:
-# speedcam-rabbitmq  10.178.0.2
-# speedcam-mysql     10.178.0.3
+#### 6.2.4 promtail-config.mon.yml (모니터링 인스턴스 전용)
+
+Loki가 같은 인스턴스이므로 localhost.
+
+```yaml
+server:
+  http_listen_port: 9080
+  grpc_listen_port: 0
+
+positions:
+  filename: /tmp/positions.yaml
+
+clients:
+  - url: http://localhost:3100/loki/api/v1/push
+
+scrape_configs:
+  - job_name: docker
+    docker_sd_configs:
+      - host: unix:///var/run/docker.sock
+        refresh_interval: 5s
+        filters:
+          - name: name
+            values:
+              - "speedcam-.*"
+    relabel_configs:
+      - source_labels: ["__meta_docker_container_name"]
+        regex: "/(.*)"
+        target_label: "container"
+      - source_labels: ["__meta_docker_container_name"]
+        regex: "/speedcam-(.*)"
+        target_label: "service"
+    pipeline_stages:
+      - regex:
+          expression: ".*trace_id=(?P<trace_id>[a-f0-9]+).*"
+      - labels:
+          trace_id:
 ```
 
-### 3.8 Step 8: 서비스 인스턴스 생성
+#### 6.2.5 grafana/provisioning/datasources/datasources.yml
+
+모니터링 인스턴스에 배포.
+
+```yaml
+apiVersion: 1
+
+datasources:
+  - name: Prometheus
+    type: prometheus
+    access: proxy
+    url: http://localhost:9090
+    isDefault: true
+    editable: false
+
+  - name: Jaeger
+    type: jaeger
+    access: proxy
+    url: http://localhost:16686
+    editable: false
+
+  - name: Loki
+    type: loki
+    access: proxy
+    url: http://localhost:3100
+    editable: false
+    jsonData:
+      derivedFields:
+        - datasourceUid: jaeger
+          matcherRegex: "trace_id=(\\w+)"
+          name: TraceID
+          url: "$${__value.raw}"
+          datasourceName: Jaeger
+```
+
+#### 6.2.6 mysqld-exporter/.my.cnf
+
+DB 인스턴스에 배포.
+
+```ini
+[client]
+user=sa
+password=<production-password>
+host=localhost
+port=3306
+```
+
+### 6.3 envsubst를 활용한 IP 자동 주입
+
+수동으로 IP를 교체하는 대신 환경변수 파일 하나로 관리할 수 있다.
 
 ```bash
-# 환경 변수 (Internal IP로 대체)
-RABBITMQ_IP=10.178.0.2
-MYSQL_IP=10.178.0.3
+# deploy/env/hosts.env — IP 정의 (이것만 수정)
+export DB_HOST=10.178.0.11
+export MQ_HOST=10.178.0.12
+export APP_HOST=10.178.0.13
+export OCR_HOST=10.178.0.14
+export ALERT_HOST=10.178.0.15
+export MON_HOST=10.178.0.20
+```
 
-# Main Service
-gcloud compute instances create-with-container speedcam-main \
-  --zone=${GCP_ZONE} \
-  --machine-type=e2-medium \
-  --tags=speedcam,speedcam-web \
-  --scopes=cloud-platform \
-  --container-image=${REGISTRY}/main:latest \
-  --container-env="DJANGO_SETTINGS_MODULE=config.settings.dev,\
-DB_HOST=${MYSQL_IP},DB_PORT=3306,\
-DB_NAME=speedcam,DB_NAME_VEHICLES=speedcam_vehicles,\
-DB_NAME_DETECTIONS=speedcam_detections,DB_NAME_NOTIFICATIONS=speedcam_notifications,\
-DB_USER=sa,DB_PASSWORD=1234,\
-CELERY_BROKER_URL=amqp://sa:1234@${RABBITMQ_IP}:5672//,\
-RABBITMQ_HOST=${RABBITMQ_IP},MQTT_PORT=1883,MQTT_USER=sa,MQTT_PASS=1234,\
-OCR_MOCK=true,FCM_MOCK=true"
+```bash
+# 배포 스크립트 예시
+source env/hosts.env
 
-# OCR Worker
-gcloud compute instances create-with-container speedcam-ocr \
-  --zone=${GCP_ZONE} \
-  --machine-type=e2-medium \
-  --tags=speedcam \
-  --scopes=cloud-platform \
-  --container-image=${REGISTRY}/ocr:latest \
-  --container-env="DJANGO_SETTINGS_MODULE=config.settings.dev,\
-DB_HOST=${MYSQL_IP},DB_PORT=3306,\
-DB_NAME=speedcam,DB_NAME_VEHICLES=speedcam_vehicles,\
-DB_NAME_DETECTIONS=speedcam_detections,DB_NAME_NOTIFICATIONS=speedcam_notifications,\
-DB_USER=sa,DB_PASSWORD=1234,\
-CELERY_BROKER_URL=amqp://sa:1234@${RABBITMQ_IP}:5672//,\
-OCR_CONCURRENCY=2,OCR_MOCK=true"
+# 템플릿에서 실제 설정 파일 생성
+envsubst < config/monitoring/prometheus/prometheus.yml.template \
+         > config/monitoring/prometheus/prometheus.yml
 
-# Alert Worker
-gcloud compute instances create-with-container speedcam-alert \
-  --zone=${GCP_ZONE} \
-  --machine-type=e2-small \
-  --tags=speedcam \
-  --scopes=cloud-platform \
-  --container-image=${REGISTRY}/alert:latest \
-  --container-env="DJANGO_SETTINGS_MODULE=config.settings.dev,\
-DB_HOST=${MYSQL_IP},DB_PORT=3306,\
-DB_NAME=speedcam,DB_NAME_VEHICLES=speedcam_vehicles,\
-DB_NAME_DETECTIONS=speedcam_detections,DB_NAME_NOTIFICATIONS=speedcam_notifications,\
-DB_USER=sa,DB_PASSWORD=1234,\
-CELERY_BROKER_URL=amqp://sa:1234@${RABBITMQ_IP}:5672//,\
-ALERT_CONCURRENCY=50,FCM_MOCK=true"
+envsubst < env/backend.env.template \
+         > env/backend.env
+
+envsubst < config/monitoring/promtail/promtail-config.yml.template \
+         > config/monitoring/promtail/promtail-config.yml
+
+envsubst < compose/docker-compose.mon.yml.template \
+         > compose/docker-compose.mon.yml
 ```
 
-### 3.9 Step 9: Django 마이그레이션
+이렇게 하면 GCP 계정이 바뀌어도 `hosts.env`만 수정하고 envsubst를 다시 실행하면 된다.
+
+---
+
+## 7. Docker 이미지 빌드 및 배포
+
+### 7.1 Docker 이미지 빌드
 
 ```bash
-gcloud compute ssh speedcam-main --zone=${GCP_ZONE} --command="\
-docker exec \$(docker ps -q) python manage.py makemigrations vehicles detections notifications && \
-docker exec \$(docker ps -q) python manage.py migrate --database=default --noinput && \
-docker exec \$(docker ps -q) python manage.py migrate vehicles --database=vehicles_db --noinput && \
-docker exec \$(docker ps -q) python manage.py migrate detections --database=detections_db --noinput && \
-docker exec \$(docker ps -q) python manage.py migrate notifications --database=notifications_db --noinput"
+# linux/amd64 플랫폼으로 빌드 (GCE용)
+docker build --platform linux/amd64 \
+  -t ${ARTIFACT_REGISTRY}/speedcam-main:latest \
+  -f docker/Dockerfile.main .
+
+docker build --platform linux/amd64 \
+  -t ${ARTIFACT_REGISTRY}/speedcam-ocr:latest \
+  -f docker/Dockerfile.ocr .
+
+docker build --platform linux/amd64 \
+  -t ${ARTIFACT_REGISTRY}/speedcam-alert:latest \
+  -f docker/Dockerfile.alert .
+```
+
+### 7.2 Docker 이미지 푸시
+
+```bash
+docker push ${ARTIFACT_REGISTRY}/speedcam-main:latest
+docker push ${ARTIFACT_REGISTRY}/speedcam-ocr:latest
+docker push ${ARTIFACT_REGISTRY}/speedcam-alert:latest
 ```
 
 ---
 
-## 4. 배포 검증
+## 8. 배포 순서
 
-### 4.1 인스턴스 상태 확인
+인프라 → 앱 → 모니터링 순서로 배포한다.
+
+### Step 1: IP 확인
 
 ```bash
-gcloud compute instances list --filter="name~speedcam"
+# GCP 콘솔 또는 CLI에서 각 인스턴스 내부 IP 확인
+gcloud compute instances list --filter="name~speedcam" \
+    --format="table(name, networkInterfaces[0].networkIP)"
+
+# 예시 출력:
+# NAME             INTERNAL_IP
+# speedcam-db      10.178.0.11
+# speedcam-mq      10.178.0.12
+# speedcam-app     10.178.0.13
+# speedcam-ocr     10.178.0.14
+# speedcam-alert   10.178.0.15
+# speedcam-mon     10.178.0.20
 ```
 
-### 4.2 서비스 헬스체크
+### Step 2: 설정 파일에 IP 주입
 
 ```bash
-# Main API External IP 확인
-MAIN_IP=$(gcloud compute instances describe speedcam-main \
-  --zone=${GCP_ZONE} \
-  --format='get(networkInterfaces[0].accessConfigs[0].natIP)')
+# backend.env, prometheus.yml, promtail-config.yml 등에서
+# ${DB_HOST}, ${MQ_HOST} 등을 실제 IP로 교체
+# (sed 또는 envsubst 사용 가능)
+
+# envsubst 예시
+source deploy/env/hosts.env
+envsubst < deploy/env/backend.env.template > deploy/env/backend.env
+envsubst < deploy/config/monitoring/prometheus/prometheus.yml.template > deploy/config/monitoring/prometheus/prometheus.yml
+```
 
-# Health Check
-curl http://${MAIN_IP}:8000/health/
-# Expected: {"status": "healthy"}
+### Step 3: DB 인스턴스 배포 (먼저)
 
-# Swagger UI
-echo "Swagger: http://${MAIN_IP}:8000/swagger/"
+```bash
+gcloud compute ssh speedcam-db --zone=${GCP_ZONE}
+
+cd deploy/compose
+docker compose -f docker-compose.db.yml up -d
+
+# MySQL healthy 확인
+docker compose -f docker-compose.db.yml ps
+docker logs speedcam-mysql
 ```
 
-### 4.3 RabbitMQ 확인
+### Step 4: MQ 인스턴스 배포
 
 ```bash
-RABBITMQ_IP=$(gcloud compute instances describe speedcam-rabbitmq \
-  --zone=${GCP_ZONE} \
-  --format='get(networkInterfaces[0].accessConfigs[0].natIP)')
+gcloud compute ssh speedcam-mq --zone=${GCP_ZONE}
+
+cd deploy/compose
+docker compose -f docker-compose.mq.yml up -d
+
+# RabbitMQ healthy 확인
+docker compose -f docker-compose.mq.yml ps
+docker logs speedcam-rabbitmq
+```
+
+### Step 5: App 인스턴스 배포
+
+```bash
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE}
+
+cd deploy/compose
+docker compose -f docker-compose.app.yml up -d
 
-echo "RabbitMQ Management: http://${RABBITMQ_IP}:15672/"
-# Credentials: sa / 1234
+# curl localhost:8000/health/ 로 확인
+curl localhost:8000/health/
 ```
 
-### 4.4 Worker 로그 확인
+### Step 6: Django 마이그레이션
+
+```bash
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE}
+
+docker exec speedcam-main python manage.py makemigrations vehicles detections notifications
+docker exec speedcam-main python manage.py migrate --database=default --noinput
+docker exec speedcam-main python manage.py migrate vehicles --database=vehicles_db --noinput
+docker exec speedcam-main python manage.py migrate detections --database=detections_db --noinput
+docker exec speedcam-main python manage.py migrate notifications --database=notifications_db --noinput
+```
+
+### Step 7: Worker 인스턴스 배포
 
 ```bash
 # OCR Worker
-gcloud compute ssh speedcam-ocr --zone=${GCP_ZONE} \
-  --command="docker logs \$(docker ps -q) 2>&1 | tail -20"
+gcloud compute ssh speedcam-ocr --zone=${GCP_ZONE}
+cd deploy/compose
+docker compose -f docker-compose.ocr.yml up -d
 
 # Alert Worker
-gcloud compute ssh speedcam-alert --zone=${GCP_ZONE} \
-  --command="docker logs \$(docker ps -q) 2>&1 | tail -20"
+gcloud compute ssh speedcam-alert --zone=${GCP_ZONE}
+cd deploy/compose
+docker compose -f docker-compose.alert.yml up -d
 ```
 
+### Step 8: 모니터링 인스턴스 배포 (마지막)
+
+```bash
+gcloud compute ssh speedcam-mon --zone=${GCP_ZONE}
+
+cd deploy/compose
+docker compose -f docker-compose.mon.yml up -d
+
+# Prometheus targets 확인
+curl -s localhost:9090/api/v1/targets | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+for t in data['data']['activeTargets']:
+    print(f\"{t['labels']['job']:20s} {t['labels']['instance']:30s} {t['health']}\")
+"
+```
+
+---
+
+## 9. 배포 검증
+
+배포 후 아래 항목을 순서대로 확인한다.
+
+### 9.1 인프라
+
+- [ ] MySQL 접속: `mysql -h ${DB_HOST} -u sa -p`
+- [ ] RabbitMQ Management UI: `http://${MQ_HOST}:15672` (sa/<password>)
+- [ ] RabbitMQ Prometheus metrics: `curl http://${MQ_HOST}:15692/metrics | head`
+
+### 9.2 애플리케이션
+
+- [ ] Django health: `curl http://${APP_HOST}:8000/health/`
+- [ ] Django metrics: `curl http://${APP_HOST}:8000/metrics | head`
+- [ ] Swagger UI: `http://${APP_HOST}:8000/swagger/`
+- [ ] Flower: `http://${APP_HOST}:5555`
+
+### 9.3 워커
+
+- [ ] OCR Worker 로그: `gcloud compute ssh speedcam-ocr --zone=${GCP_ZONE} --command="docker logs speedcam-ocr"`
+- [ ] Alert Worker 로그: `gcloud compute ssh speedcam-alert --zone=${GCP_ZONE} --command="docker logs speedcam-alert"`
+- [ ] RabbitMQ에서 consumer 확인: Management UI → Queues → ocr_queue, fcm_queue
+
+### 9.4 모니터링
+
+- [ ] Prometheus targets 전부 UP: `http://${MON_HOST}:9090/targets`
+- [ ] Grafana 접속: `http://${MON_HOST}:3000` (admin/admin)
+- [ ] Grafana 데이터소스 3개 연결: Prometheus, Jaeger, Loki
+- [ ] Jaeger에서 서비스 트레이스 확인: `http://${MON_HOST}:16686`
+- [ ] Loki에서 로그 확인: Grafana → Explore → Loki → `{service="main"}`
+- [ ] cAdvisor: 6개 인스턴스 모두 cadvisor:8080 → Prometheus에서 수집 확인
+
 ---
 
-## 5. 운영 가이드
+## 10. 운영 가이드
 
-### 5.1 인스턴스 재시작
+### 10.1 서비스 재시작
 
 ```bash
-# 개별 재시작
-gcloud compute instances reset speedcam-main --zone=${GCP_ZONE}
+# 개별 인스턴스 재시작
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE}
+cd deploy/compose
+docker compose -f docker-compose.app.yml restart
 
 # 전체 서비스 재시작
-gcloud compute instances reset speedcam-main speedcam-ocr speedcam-alert --zone=${GCP_ZONE}
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} --command="cd deploy/compose && docker compose -f docker-compose.app.yml restart"
+gcloud compute ssh speedcam-ocr --zone=${GCP_ZONE} --command="cd deploy/compose && docker compose -f docker-compose.ocr.yml restart"
+gcloud compute ssh speedcam-alert --zone=${GCP_ZONE} --command="cd deploy/compose && docker compose -f docker-compose.alert.yml restart"
 ```
 
-### 5.2 이미지 업데이트 배포
+### 10.2 이미지 업데이트 배포
 
 ```bash
 # 1. 새 이미지 빌드 & 푸시
-docker build --platform linux/amd64 -t ${REGISTRY}/main:latest -f docker/Dockerfile.main .
-docker push ${REGISTRY}/main:latest
-
-# 2. 인스턴스 재시작 (새 이미지 pull)
-gcloud compute instances reset speedcam-main --zone=${GCP_ZONE}
+docker build --platform linux/amd64 -t ${ARTIFACT_REGISTRY}/speedcam-main:latest -f docker/Dockerfile.main .
+docker push ${ARTIFACT_REGISTRY}/speedcam-main:latest
+
+# 2. 인스턴스에서 새 이미지 pull 및 재시작
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE}
+cd deploy/compose
+docker compose -f docker-compose.app.yml pull
+docker compose -f docker-compose.app.yml up -d
 ```
 
-### 5.3 스케일링
+### 10.3 스케일링
 
 ```bash
 # OCR Worker 추가 인스턴스
-gcloud compute instances create-with-container speedcam-ocr-2 \
+gcloud compute instances create speedcam-ocr-2 \
   --zone=${GCP_ZONE} \
-  --machine-type=e2-medium \
+  --machine-type=e2-standard-2 \
+  --network-interface=subnet=speedcam-subnet,no-address \
   --tags=speedcam \
-  --scopes=cloud-platform \
-  --container-image=${REGISTRY}/ocr:latest \
-  --container-env="..." # 동일한 환경변수
+  --scopes=cloud-platform
+
+# 동일한 설정 파일로 배포
+gcloud compute scp --recurse deploy/ speedcam-ocr-2:~ --zone=${GCP_ZONE}
+gcloud compute ssh speedcam-ocr-2 --zone=${GCP_ZONE}
+cd deploy/compose
+docker compose -f docker-compose.ocr.yml up -d
 ```
 
-### 5.4 로그 모니터링
+### 10.4 로그 모니터링
 
 ```bash
 # 실시간 로그
-gcloud compute ssh speedcam-main --zone=${GCP_ZONE} \
-  --command="docker logs -f \$(docker ps -q)"
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="docker logs -f speedcam-main"
+
+# 최근 로그
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="docker logs --tail 100 speedcam-main"
 ```
 
+### 10.5 IP 변경 시 수정 대상
+
+GCP 계정/프로젝트 변경으로 IP가 바뀌면 아래 파일만 교체하면 된다.
+
+| 파일 | 교체 대상 | 배포 위치 |
+|------|-----------|-----------|
+| `env/backend.env` | `DB_HOST`, `RABBITMQ_HOST`, `CELERY_BROKER_URL`, `OTEL_EXPORTER_OTLP_ENDPOINT` | app, ocr, alert |
+| `config/monitoring/prometheus/prometheus.yml` | 모든 targets IP | mon |
+| `config/monitoring/promtail/promtail-config.yml` | Loki URL (`${MON_HOST}`) | db, mq, app, ocr, alert |
+| `docker-compose.mon.yml` | `CE_BROKER_URL` 의 MQ IP | mon |
+
+Grafana datasources, OTel Collector config, mysqld-exporter .my.cnf는 같은 인스턴스 내 통신(localhost)이므로 IP 변경 영향 없음.
+
 ---
 
-## 6. 트러블슈팅
+## 11. 트러블슈팅
 
-### 6.1 컨테이너 시작 실패
+### 11.1 컨테이너 시작 실패
 
 ```bash
 # 컨테이너 상태 확인
-gcloud compute ssh speedcam-main --zone=${GCP_ZONE} \
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
   --command="docker ps -a"
 
 # 종료된 컨테이너 로그 확인
-gcloud compute ssh speedcam-main --zone=${GCP_ZONE} \
-  --command="docker logs \$(docker ps -aq | head -1)"
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="docker logs speedcam-main"
 ```
 
-### 6.2 DB 연결 실패
+### 11.2 DB 연결 실패
 
 ```bash
 # MySQL 연결 테스트
-gcloud compute ssh speedcam-main --zone=${GCP_ZONE} \
-  --command="docker exec \$(docker ps -q) python -c \"
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="docker exec speedcam-main python -c \"
 import pymysql
-conn = pymysql.connect(host='10.178.0.3', user='sa', password='1234', database='speedcam')
+conn = pymysql.connect(host='${DB_HOST}', user='sa', password='<password>', database='speedcam')
 print('Connected!')
 conn.close()\""
+
+# 방화벽 규칙 확인
+gcloud compute firewall-rules list --filter="name~speedcam"
 ```
 
-### 6.3 MQTT 연결 실패
+### 11.3 MQTT 연결 실패
 
 ```bash
 # RabbitMQ MQTT 플러그인 상태 확인
-gcloud compute ssh speedcam-rabbitmq --zone=${GCP_ZONE} \
-  --command="docker exec \$(docker ps -q) rabbitmq-plugins list | grep mqtt"
+gcloud compute ssh speedcam-mq --zone=${GCP_ZONE} \
+  --command="docker exec speedcam-rabbitmq rabbitmq-plugins list | grep mqtt"
+
+# MQTT 포트 listening 확인
+gcloud compute ssh speedcam-mq --zone=${GCP_ZONE} \
+  --command="netstat -tlnp | grep 1883"
 ```
 
-### 6.4 이미지 Pull 실패
+### 11.4 이미지 Pull 실패
 
 ```bash
 # 서비스 계정 권한 확인
-gcloud compute instances describe speedcam-main --zone=${GCP_ZONE} \
+gcloud compute instances describe speedcam-app --zone=${GCP_ZONE} \
   --format='get(serviceAccounts[0].scopes)'
 
 # cloud-platform 스코프 필요
+# 없으면 인스턴스 재생성 또는 scope 추가
+```
+
+### 11.5 host 모드 네트워크 문제
+
+```bash
+# 포트 사용 확인
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="netstat -tlnp | grep 8000"
+
+# 컨테이너 네트워크 모드 확인
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="docker inspect speedcam-main | grep NetworkMode"
+```
+
+### 11.6 모니터링 메트릭 수집 실패
+
+```bash
+# Prometheus targets 상태 확인
+curl http://${MON_HOST}:9090/api/v1/targets | python3 -m json.tool
+
+# cAdvisor 접근 확인
+curl http://${APP_HOST}:8080/metrics | head
+
+# Promtail 로그 확인
+gcloud compute ssh speedcam-app --zone=${GCP_ZONE} \
+  --command="docker logs speedcam-promtail"
 ```
 
 ---
 
-## 7. 리소스 정리
+## 12. 리소스 정리
 
 ```bash
 # 모든 인스턴스 삭제
 gcloud compute instances delete \
-  speedcam-rabbitmq speedcam-mysql \
-  speedcam-main speedcam-ocr speedcam-alert \
+  speedcam-db speedcam-mq speedcam-app \
+  speedcam-ocr speedcam-alert speedcam-mon \
   --zone=${GCP_ZONE} --quiet
 
 # 방화벽 규칙 삭제
-gcloud compute firewall-rules delete speedcam-internal speedcam-external --quiet
+gcloud compute firewall-rules delete \
+  speedcam-internal \
+  speedcam-api-external \
+  speedcam-mqtt-external \
+  speedcam-grafana-external \
+  --quiet
 
 # Artifact Registry 삭제
-gcloud artifacts repositories delete speedcam --location=${GCP_REGION} --quiet
+gcloud artifacts repositories delete speedcam \
+  --location=${GCP_REGION} --quiet
+
+# VPC 서브넷 삭제
+gcloud compute networks subnets delete speedcam-subnet \
+  --region=${GCP_REGION} --quiet
+
+# VPC 네트워크 삭제
+gcloud compute networks delete speedcam-vpc --quiet
 ```
 
 ---
@@ -536,3 +1465,4 @@ gcloud artifacts repositories delete speedcam --location=${GCP_REGION} --quiet
 | 날짜 | 버전 | 변경 내용 |
 |------|------|----------|
 | 2026-01-23 | 1.0 | 초기 문서 작성 |
+| 2026-02-06 | 2.0 | 멀티 인스턴스 배포 방식으로 전면 재작성 (6개 인스턴스, docker-compose + host mode, 모니터링 스택 추가) |
diff --git a/docs/MONITORING.md b/docs/MONITORING.md
new file mode 100644
index 0000000..65251c4
--- /dev/null
+++ b/docs/MONITORING.md
@@ -0,0 +1,452 @@
+# 모니터링 스택 가이드
+
+## 1. 아키텍처 개요
+
+```
+App Services (main, ocr-worker, alert-worker)
+    │ OTLP gRPC (:4317)
+    ▼
+OTel Collector ──traces──► Jaeger (:16686) ──► Grafana (:3000)
+    │                                              ▲
+    └──metrics──► Prometheus (:9090) ──────────────┘
+                      ▲                            │
+cAdvisor ─────────────┤                    Loki (:3100) ◄── Promtail
+Django /metrics ──────┤                            ▲
+RabbitMQ :15692 ──────┤                     Docker logs
+mysqld-exporter ──────┤
+celery-exporter ──────┘
+
+K6 (부하테스트) ──prometheus remote write──► Prometheus
+```
+
+---
+
+## 2. 서비스 구성
+
+### 2.1 전체 서비스 목록
+
+| 서비스 | 이미지 | 포트 | 역할 |
+|--------|--------|------|------|
+| **otel-collector** | `otel/opentelemetry-collector-contrib:0.98.0` | 4317 (gRPC), 4318 (HTTP), 8889 | 트레이스/메트릭 수집 허브 |
+| **jaeger** | `jaegertracing/all-in-one:1.57` | 16686 (UI), 14250 | 분산 트레이싱 저장/UI |
+| **prometheus** | `prom/prometheus:v2.51.2` | 9090 | 메트릭 수집/저장/쿼리 |
+| **grafana** | `grafana/grafana:10.4.2` | 3000 | 통합 대시보드 |
+| **loki** | `grafana/loki:2.9.6` | 3100 | 로그 집계/저장 |
+| **promtail** | `grafana/promtail:2.9.6` | - | Docker 로그 → Loki 전송 |
+| **cadvisor** | `gcr.io/cadvisor/cadvisor:v0.49.1` | 8080 | 컨테이너 리소스 메트릭 |
+| **mysqld-exporter** | `prom/mysqld-exporter:v0.15.1` | 9104 | MySQL 메트릭 노출 |
+| **celery-exporter** | `danihodovic/celery-exporter:0.10.3` | 9808 | Celery 큐/태스크 메트릭 |
+| **k6** | `grafana/k6:latest` | - | 부하 테스트 (on-demand) |
+
+### 2.2 Prometheus Scrape Targets
+
+| Job | Target | 수집 항목 |
+|-----|--------|-----------|
+| `django` | `main:8000/metrics` | HTTP 요청 수, 응답 시간, DB 쿼리 수 |
+| `otel-collector` | `otel-collector:8889` | OTel에서 변환된 앱 메트릭 |
+| `rabbitmq` | `rabbitmq:15692` | 큐 깊이, 메시지 rate, 커넥션, 채널 |
+| `mysql` | `mysqld-exporter:9104` | 쿼리 수, 커넥션, InnoDB 버퍼, 슬로우 쿼리 |
+| `celery` | `celery-exporter:9808` | 태스크 성공/실패, 실행 시간, 큐 길이 |
+| `cadvisor` | `cadvisor:8080` | 컨테이너 CPU, 메모리, 네트워크 I/O |
+
+---
+
+## 3. 실행 방법
+
+### 3.1 기본 서비스만 (모니터링 없이)
+
+```bash
+cd docker
+docker compose up -d
+```
+
+앱은 모니터링 스택 없이도 정상 동작함. OTel Collector에 연결 실패해도 앱은 죽지 않음 (graceful fallback).
+
+### 3.2 모니터링 포함
+
+```bash
+cd docker
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml up -d
+```
+
+### 3.3 부하 테스트 포함 (k6)
+
+```bash
+# k6 서비스는 profiles: [loadtest] 이므로 명시적 실행 필요
+cd docker
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml \
+  run k6 run --out experimental-prometheus-rw /scripts/load-test.js
+```
+
+### 3.4 모니터링만 재시작 (앱 유지)
+
+```bash
+cd docker
+docker compose -f docker-compose.monitoring.yml restart prometheus grafana
+```
+
+---
+
+## 4. 접속 정보
+
+| 서비스 | URL | 인증 |
+|--------|-----|------|
+| **Grafana** | http://localhost:3000 | admin / admin |
+| **Prometheus** | http://localhost:9090 | 없음 |
+| **Jaeger** | http://localhost:16686 | 없음 |
+| **RabbitMQ Management** | http://localhost:15672 | sa / 1234 |
+| **Flower** | http://localhost:5555 | 없음 |
+| **cAdvisor** | http://localhost:8080 | 없음 |
+| **Django /metrics** | http://localhost:8000/metrics | 없음 |
+
+---
+
+## 5. Prometheus 타겟 확인
+
+### 5.1 UI에서 확인
+
+```
+http://localhost:9090/targets
+```
+
+6개 job이 모두 **UP** (초록색)이면 정상.
+
+### 5.2 API로 확인
+
+```bash
+curl -s http://localhost:9090/api/v1/targets | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+for t in data['data']['activeTargets']:
+    print(f\"{t['labels']['job']:20s} {t['labels']['instance']:30s} {t['health']}\")
+"
+```
+
+### 5.3 타겟이 DOWN일 때
+
+| 증상 | 원인 | 해결 |
+|------|------|------|
+| django DOWN | main 컨테이너 미기동 또는 django-prometheus 미설치 | `docker logs speedcam-main` 확인 |
+| rabbitmq DOWN | rabbitmq_prometheus 플러그인 미활성화 | docker-compose.yml의 command에 `rabbitmq_prometheus` 포함 확인 |
+| mysql DOWN | mysqld-exporter 인증 실패 | `.my.cnf` 파일의 user/password 확인 |
+| celery DOWN | celery-exporter가 broker 연결 실패 | RabbitMQ 기동 여부 확인 |
+
+---
+
+## 6. 설정 파일 구조
+
+```
+docker/monitoring/
+├── otel-collector/
+│   └── otel-collector-config.yml    # OTLP 수신 → Jaeger/Prometheus 내보내기
+├── prometheus/
+│   └── prometheus.yml               # scrape targets 정의
+├── loki/
+│   └── loki-config.yml              # 로그 저장 (7일 보존)
+├── promtail/
+│   └── promtail-config.yml          # Docker 로그 수집 → Loki 전송
+├── grafana/
+│   └── provisioning/
+│       ├── datasources/
+│       │   └── datasources.yml      # Prometheus, Jaeger, Loki 자동 등록
+│       └── dashboards/
+│           └── dashboards.yml       # 대시보드 프로비저닝
+└── mysqld-exporter/
+    └── .my.cnf                      # MySQL 접속 정보
+```
+
+---
+
+## 7. OpenTelemetry 계측
+
+### 7.1 앱 계측 방식
+
+`opentelemetry-instrument` CLI로 자동 계측 (코드 수정 없음):
+
+```bash
+# start_main.sh
+opentelemetry-instrument \
+    --service_name speedcam-api \
+    gunicorn config.wsgi:application ...
+
+# start_ocr_worker.sh
+opentelemetry-instrument \
+    --service_name speedcam-ocr \
+    celery -A config worker ...
+
+# start_alert_worker.sh
+opentelemetry-instrument \
+    --service_name speedcam-alert \
+    celery -A config worker ...
+```
+
+### 7.2 자동 계측 대상
+
+| 패키지 | 계측 대상 |
+|--------|-----------|
+| `opentelemetry-instrumentation-django` | HTTP 요청/응답, 미들웨어 |
+| `opentelemetry-instrumentation-celery` | 태스크 실행, 큐 대기 시간 |
+| `opentelemetry-instrumentation-pymysql` | DB 쿼리, 커넥션 |
+| `opentelemetry-instrumentation-requests` | 외부 HTTP 호출 (GCS, FCM) |
+| `opentelemetry-instrumentation-logging` | 로그에 trace_id/span_id 주입 |
+
+### 7.3 환경변수 (backend.env)
+
+```bash
+OTEL_EXPORTER_OTLP_ENDPOINT=http://otel-collector:4317
+OTEL_EXPORTER_OTLP_PROTOCOL=grpc
+OTEL_RESOURCE_ATTRIBUTES=service.namespace=speedcam,deployment.environment=dev
+OTEL_TRACES_SAMPLER=parentbased_always_on
+OTEL_PYTHON_LOG_CORRELATION=true
+```
+
+### 7.4 데이터 흐름
+
+```
+Django/Celery → (OTLP gRPC) → OTel Collector
+                                  ├── traces → Jaeger
+                                  └── metrics → Prometheus (:8889)
+
+Django /metrics → (HTTP scrape) → Prometheus (django-prometheus 메트릭)
+```
+
+---
+
+## 8. 로그 → 트레이스 연동 (Loki ↔ Jaeger)
+
+### 8.1 동작 원리
+
+1. `opentelemetry-instrumentation-logging`이 로그에 `trace_id`, `span_id` 주입
+2. Django LOGGING 포맷:
+   ```
+   INFO 2024-01-01 12:00:00 views [trace_id=abc123 span_id=def456] Request processed
+   ```
+3. Promtail이 로그에서 `trace_id` 추출 → Loki 라벨로 저장
+4. Grafana Loki 데이터소스의 `derivedFields`가 trace_id → Jaeger 링크 자동 생성
+
+### 8.2 확인 방법
+
+1. Grafana → Explore → Loki 데이터소스 선택
+2. `{service="main"}` 쿼리 실행
+3. 로그 라인의 `trace_id=` 부분 클릭 → Jaeger 트레이스로 이동
+
+---
+
+## 9. 유용한 PromQL 쿼리
+
+### 9.1 Django
+
+```promql
+# 초당 요청 수 (RPS)
+rate(django_http_requests_total_by_method_total[5m])
+
+# 응답 시간 p95
+histogram_quantile(0.95, rate(django_http_requests_latency_seconds_by_view_method_bucket[5m]))
+
+# HTTP 5xx 에러율
+rate(django_http_responses_total_by_status_total{status=~"5.."}[5m])
+/ rate(django_http_responses_total_by_status_total[5m])
+
+# DB 쿼리 수
+rate(django_db_execute_total[5m])
+```
+
+### 9.2 RabbitMQ
+
+```promql
+# 큐별 대기 메시지 수
+rabbitmq_queue_messages{queue=~"ocr_queue|fcm_queue"}
+
+# 초당 메시지 발행율
+rate(rabbitmq_queue_messages_published_total[5m])
+
+# Consumer 수
+rabbitmq_queue_consumers{queue=~"ocr_queue|fcm_queue"}
+```
+
+### 9.3 MySQL
+
+```promql
+# 활성 커넥션 수
+mysql_global_status_threads_connected
+
+# 초당 쿼리 수
+rate(mysql_global_status_questions[5m])
+
+# 슬로우 쿼리 수
+rate(mysql_global_status_slow_queries[5m])
+```
+
+### 9.4 Celery
+
+```promql
+# 태스크 성공/실패 수
+celery_tasks_total{state="SUCCESS"}
+celery_tasks_total{state="FAILURE"}
+
+# 태스크 실행 시간
+celery_tasks_runtime_seconds{quantile="0.95"}
+
+# 큐 길이
+celery_queue_length
+```
+
+### 9.5 컨테이너 리소스
+
+```promql
+# 컨테이너별 CPU 사용률
+rate(container_cpu_usage_seconds_total{name=~"speedcam-.*"}[5m]) * 100
+
+# 컨테이너별 메모리 사용량 (MB)
+container_memory_usage_bytes{name=~"speedcam-.*"} / 1024 / 1024
+
+# 컨테이너별 네트워크 I/O (bytes/sec)
+rate(container_network_receive_bytes_total{name=~"speedcam-.*"}[5m])
+```
+
+---
+
+## 10. K6 부하 테스트 + 모니터링
+
+### 10.1 실행
+
+```bash
+cd docker
+
+# 부하 테스트 실행 (Prometheus에 결과 기록)
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml \
+  run k6 run --out experimental-prometheus-rw /scripts/load-test.js
+```
+
+### 10.2 K6 → Prometheus 메트릭
+
+k6는 `--out experimental-prometheus-rw`로 결과를 Prometheus에 직접 기록. `--web.enable-remote-write-receiver` 플래그가 Prometheus에 설정되어 있음.
+
+| k6 메트릭 | PromQL | 의미 |
+|-----------|--------|------|
+| `k6_http_req_duration_seconds` | `histogram_quantile(0.95, rate(k6_http_req_duration_seconds_bucket[1m]))` | HTTP p95 응답 시간 |
+| `k6_http_reqs_total` | `rate(k6_http_reqs_total[1m])` | 초당 HTTP 요청 수 |
+| `k6_vus` | `k6_vus` | 현재 VU 수 |
+| `k6_http_req_failed_total` | `rate(k6_http_req_failed_total[1m])` | 실패율 |
+
+### 10.3 부하 테스트 중 모니터링 체크리스트
+
+부하 테스트 중 Grafana에서 아래 항목을 실시간 확인:
+
+| 확인 항목 | 보는 곳 | 정상 기준 |
+|-----------|---------|-----------|
+| API 응답 시간 | Prometheus - django 메트릭 | p95 < 500ms |
+| 에러율 | Prometheus - django 5xx rate | < 1% |
+| RabbitMQ 큐 깊이 | Prometheus - rabbitmq 메트릭 | 지속 증가 없음 |
+| Celery 태스크 처리율 | Prometheus - celery 메트릭 | 발행율 ≈ 소비율 |
+| MySQL 커넥션 | Prometheus - mysql 메트릭 | < pool size 80% |
+| 컨테이너 CPU/메모리 | Prometheus - cadvisor 메트릭 | CPU < 80%, Memory < 85% |
+| 분산 트레이스 | Jaeger | 에러 트레이스 없음 |
+| 로그 에러 | Loki | ERROR 로그 급증 없음 |
+
+---
+
+## 11. GCP 멀티 인스턴스 배포 시 고려사항
+
+현재 Docker Compose는 단일 호스트 내 가상 네트워크. 인스턴스를 분리할 경우:
+
+### 11.1 인스턴스 분리 구성 예시
+
+| 인스턴스 | 서비스 | GCP 머신 타입 |
+|----------|--------|---------------|
+| app | main, flower | e2-medium |
+| ocr-worker | ocr-worker | e2-standard-2 (CPU) |
+| alert-worker | alert-worker | e2-small |
+| db | mysql | e2-highmem-2 |
+| mq | rabbitmq | e2-medium |
+| monitoring | prometheus, grafana, jaeger, loki, promtail, otel-collector, cadvisor, exporters | e2-standard-2 |
+
+### 11.2 네트워크 연결 방법
+
+**방법 A: GCP 내부 IP 직접 지정**
+
+```bash
+# 각 인스턴스의 backend.env에서 컨테이너명 대신 내부 IP 사용
+DB_HOST=10.178.0.11                                  # db 인스턴스
+CELERY_BROKER_URL=amqp://sa:1234@10.178.0.12:5672//  # mq 인스턴스
+OTEL_EXPORTER_OTLP_ENDPOINT=http://10.178.0.15:4317  # monitoring 인스턴스
+```
+
+**방법 B: GCP 내부 DNS (같은 VPC)**
+
+```bash
+DB_HOST=db-instance.asia-northeast3-a.c.PROJECT_ID.internal
+```
+
+**방법 C: GKE (Kubernetes) — 서비스 분리가 목적이면 추천**
+
+- Service DNS 자동 부여: `mysql.default.svc.cluster.local`
+- IP 관리 불필요
+- HPA로 worker auto-scaling 가능
+- `kompose convert`로 docker-compose → k8s 변환 가능
+
+### 11.3 Prometheus 멀티 인스턴스 설정
+
+인스턴스가 분리되면 `prometheus.yml`에서 내부 IP 사용:
+
+```yaml
+scrape_configs:
+  - job_name: "django"
+    static_configs:
+      - targets: ["10.178.0.10:8000"]  # app 인스턴스
+
+  - job_name: "rabbitmq"
+    static_configs:
+      - targets: ["10.178.0.12:15692"]  # mq 인스턴스
+
+  - job_name: "mysql"
+    static_configs:
+      - targets: ["10.178.0.11:9104"]  # db 인스턴스 (mysqld-exporter 같이 띄움)
+
+  - job_name: "celery"
+    static_configs:
+      - targets: ["10.178.0.13:9808"]  # celery-exporter를 어디서 띄울지 결정 필요
+```
+
+### 11.4 주의사항
+
+- GCP 방화벽 규칙에서 모니터링 포트 (9090, 4317, 15692, 9104, 9808 등) 내부 허용 필요
+- 외부 노출하면 안 되는 포트: Prometheus (9090), Grafana (3000) → VPN 또는 IAP 터널 사용
+- 각 인스턴스에서 cAdvisor를 로컬로 띄우고, 모니터링 인스턴스의 Prometheus가 모든 cAdvisor를 scrape
+
+---
+
+## 12. 트러블슈팅
+
+### 12.1 OTel Collector 연결 실패
+
+```bash
+docker logs speedcam-otel-collector
+# "connection refused" → Jaeger 미기동 확인
+# "context deadline exceeded" → 네트워크 문제
+```
+
+### 12.2 Grafana 데이터소스 연결 실패
+
+```bash
+# Grafana 컨테이너에서 직접 확인
+docker exec speedcam-grafana curl -s http://prometheus:9090/-/healthy
+docker exec speedcam-grafana curl -s http://loki:3100/ready
+docker exec speedcam-grafana curl -s http://jaeger:16686/
+```
+
+### 12.3 Promtail 로그 수집 안됨
+
+```bash
+docker logs speedcam-promtail
+# Docker socket 접근 권한 확인
+# container name이 speedcam-* 패턴인지 확인
+```
+
+### 12.4 mysqld-exporter 인증 실패
+
+```bash
+docker logs speedcam-mysqld-exporter
+# "Access denied" → .my.cnf의 user/password 확인
+# "no configuration found" → config.my-cnf 마운트 경로 확인
+```
diff --git a/docs/PERFORMANCE_TEST.md b/docs/PERFORMANCE_TEST.md
deleted file mode 100644
index d6be556..0000000
--- a/docs/PERFORMANCE_TEST.md
+++ /dev/null
@@ -1,1094 +0,0 @@
-# 성능 테스트 가이드
-
-## 1. 개요
-
-### 1.1 목적
-Edge Device(Raspberry Pi) 없이 시스템의 성능과 안정성을 검증하기 위한 부하 테스트 수행
-
-### 1.2 테스트 도구
-- **K6**: 부하 테스트 도구
-- **xk6-mqtt**: K6 MQTT 확장 (Edge Device 시뮬레이션)
-- **Docker**: 테스트 환경 구성
-
-### 1.3 테스트 대상
-| 구간 | 설명 |
-|------|------|
-| MQTT → Main Service | Edge Device 메시지 수신 처리 |
-| Main Service → RabbitMQ | Task 발행 성능 |
-| OCR Worker | 이미지 처리 처리량 |
-| Alert Worker | FCM 전송 처리량 |
-| REST API | API 응답 시간 |
-
----
-
-## 2. 테스트 환경 구성
-
-### 2.1 아키텍처
-
-```
-┌─────────────────────────────────────────────────────────────────────┐
-│                        Performance Test Environment                  │
-├─────────────────────────────────────────────────────────────────────┤
-│                                                                     │
-│  ┌─────────────────┐         ┌─────────────────────────────────┐   │
-│  │   K6 Runner     │         │     Application Stack           │   │
-│  │                 │         │                                 │   │
-│  │ - MQTT Publish  │ ──────▶ │  RabbitMQ (MQTT + AMQP)        │   │
-│  │ - HTTP Request  │         │  Main Service (Django)          │   │
-│  │ - Metrics       │ ──────▶ │  OCR Worker (Mock)             │   │
-│  │                 │         │  Alert Worker (Mock)            │   │
-│  └─────────────────┘         │  MySQL                          │   │
-│                              └─────────────────────────────────┘   │
-│                                                                     │
-│  ┌─────────────────┐         ┌─────────────────────────────────┐   │
-│  │   Monitoring    │         │     Mock Services               │   │
-│  │                 │         │                                 │   │
-│  │ - Grafana       │◀────────│  - GCS Mock (MinIO)            │   │
-│  │ - InfluxDB      │         │  - FCM Mock (WireMock)          │   │
-│  │ - DataDog       │         │                                 │   │
-│  └─────────────────┘         └─────────────────────────────────┘   │
-│                                                                     │
-└─────────────────────────────────────────────────────────────────────┘
-```
-
-### 2.2 Docker Compose (테스트 환경)
-
-```yaml
-# docker-compose.test.yml
-version: '3.8'
-
-services:
-  # 애플리케이션 스택
-  mysql:
-    image: mysql:8.0
-    environment:
-      MYSQL_ROOT_PASSWORD: root
-      MYSQL_DATABASE: speedcam_test
-      MYSQL_USER: sa
-      MYSQL_PASSWORD: "1234"
-    ports:
-      - "3306:3306"
-    networks:
-      - test-network
-
-  rabbitmq:
-    image: rabbitmq:3.13-management
-    environment:
-      RABBITMQ_DEFAULT_USER: sa
-      RABBITMQ_DEFAULT_PASS: "1234"
-    ports:
-      - "5672:5672"
-      - "1883:1883"
-      - "15672:15672"
-    volumes:
-      - ./rabbitmq/enabled_plugins:/etc/rabbitmq/enabled_plugins
-      - ./rabbitmq/rabbitmq.conf:/etc/rabbitmq/rabbitmq.conf
-    networks:
-      - test-network
-
-  main:
-    build:
-      context: ..
-      dockerfile: docker/Dockerfile.main
-    environment:
-      - DJANGO_SETTINGS_MODULE=config.settings.dev
-      - DB_HOST=mysql
-      - DB_USER=sa
-      - DB_PASSWORD=1234
-      - CELERY_BROKER_URL=amqp://sa:1234@rabbitmq:5672//
-    ports:
-      - "8000:8000"
-    depends_on:
-      - mysql
-      - rabbitmq
-    networks:
-      - test-network
-
-  ocr-worker:
-    build:
-      context: ..
-      dockerfile: docker/Dockerfile.ocr
-    environment:
-      - DJANGO_SETTINGS_MODULE=config.settings.dev
-      - DB_HOST=mysql
-      - DB_USER=sa
-      - DB_PASSWORD=1234
-      - CELERY_BROKER_URL=amqp://sa:1234@rabbitmq:5672//
-      - GCS_MOCK_URL=http://minio:9000
-      - OCR_MOCK=true  # OCR Mock 모드
-    depends_on:
-      - rabbitmq
-      - minio
-    networks:
-      - test-network
-
-  alert-worker:
-    build:
-      context: ..
-      dockerfile: docker/Dockerfile.alert
-    environment:
-      - DJANGO_SETTINGS_MODULE=config.settings.dev
-      - DB_HOST=mysql
-      - DB_USER=sa
-      - DB_PASSWORD=1234
-      - CELERY_BROKER_URL=amqp://sa:1234@rabbitmq:5672//
-      - FCM_MOCK_URL=http://wiremock:8080
-    depends_on:
-      - rabbitmq
-      - wiremock
-    networks:
-      - test-network
-
-  # Mock Services
-  minio:
-    image: minio/minio
-    command: server /data --console-address ":9001"
-    environment:
-      MINIO_ROOT_USER: minioadmin
-      MINIO_ROOT_PASSWORD: minioadmin
-    ports:
-      - "9000:9000"
-      - "9001:9001"
-    networks:
-      - test-network
-
-  wiremock:
-    image: wiremock/wiremock:3.3.1
-    ports:
-      - "8080:8080"
-    volumes:
-      - ./wiremock:/home/wiremock
-    networks:
-      - test-network
-
-  # Monitoring
-  influxdb:
-    image: influxdb:1.8
-    environment:
-      INFLUXDB_DB: k6
-    ports:
-      - "8086:8086"
-    networks:
-      - test-network
-
-  grafana:
-    image: grafana/grafana:10.2.0
-    environment:
-      - GF_AUTH_ANONYMOUS_ORG_ROLE=Admin
-      - GF_AUTH_ANONYMOUS_ENABLED=true
-      - GF_AUTH_BASIC_ENABLED=false
-    ports:
-      - "3000:3000"
-    volumes:
-      - ./grafana/provisioning:/etc/grafana/provisioning
-      - ./grafana/dashboards:/var/lib/grafana/dashboards
-    depends_on:
-      - influxdb
-    networks:
-      - test-network
-
-  # K6 Runner
-  k6:
-    image: grafana/k6:latest
-    volumes:
-      - ./k6:/scripts
-    environment:
-      - K6_OUT=influxdb=http://influxdb:8086/k6
-    networks:
-      - test-network
-    depends_on:
-      - influxdb
-      - main
-
-networks:
-  test-network:
-    driver: bridge
-```
-
-### 2.3 Mock 서비스 설정
-
-#### WireMock (FCM Mock)
-
-**wiremock/mappings/fcm-send.json**
-```json
-{
-  "request": {
-    "method": "POST",
-    "urlPattern": "/v1/projects/.*/messages:send"
-  },
-  "response": {
-    "status": 200,
-    "headers": {
-      "Content-Type": "application/json"
-    },
-    "jsonBody": {
-      "name": "projects/test/messages/{{randomValue type='UUID'}}"
-    },
-    "transformers": ["response-template"],
-    "fixedDelayMilliseconds": 50
-  }
-}
-```
-
-#### OCR Mock 모드
-
-```python
-# tasks/ocr_tasks.py (테스트 모드)
-import os
-import random
-import string
-
-OCR_MOCK = os.getenv('OCR_MOCK', 'false').lower() == 'true'
-
-def mock_ocr_result():
-    """테스트용 가짜 OCR 결과 생성"""
-    num1 = random.randint(10, 999)
-    char = random.choice('가나다라마바사아자차카타파하')
-    num2 = random.randint(1000, 9999)
-    plate = f"{num1}{char}{num2}"
-    confidence = random.uniform(0.85, 0.99)
-    return plate, confidence
-
-@shared_task(bind=True, max_retries=3, acks_late=True)
-def process_ocr(self, detection_id: int, gcs_uri: str):
-    try:
-        Detection.objects.filter(id=detection_id).update(
-            status='processing'
-        )
-        
-        if OCR_MOCK:
-            # Mock 모드: 실제 OCR 없이 가짜 결과 반환
-            import time
-            time.sleep(random.uniform(0.1, 0.5))  # 처리 시간 시뮬레이션
-            plate_number, confidence = mock_ocr_result()
-        else:
-            # 실제 OCR 처리
-            # ... 기존 코드 ...
-            pass
-        
-        # 이하 동일
-```
-
----
-
-## 3. K6 설치 및 MQTT 확장
-
-### 3.1 xk6-mqtt 빌드
-
-MQTT 테스트를 위해 K6에 xk6-mqtt 확장을 추가합니다.
-
-```bash
-# xk6 설치
-go install go.k6.io/xk6/cmd/xk6@latest
-
-# xk6-mqtt 확장 포함하여 K6 빌드
-xk6 build --with github.com/pmalhaire/xk6-mqtt@latest
-
-# 빌드된 바이너리 확인
-./k6 version
-```
-
-### 3.2 Docker 이미지 빌드
-
-**k6/Dockerfile**
-```dockerfile
-FROM golang:1.21 as builder
-
-RUN go install go.k6.io/xk6/cmd/xk6@latest
-
-RUN xk6 build \
-    --with github.com/pmalhaire/xk6-mqtt@latest \
-    --output /k6
-
-FROM grafana/k6:latest
-COPY --from=builder /k6 /usr/bin/k6
-```
-
----
-
-## 4. 테스트 시나리오
-
-### 4.1 테스트 유형
-
-| 테스트 | 목적 | VU | Duration | 특징 |
-|--------|------|-----|----------|------|
-| **Smoke** | 기본 동작 확인 | 1-5 | 1분 | 최소 부하로 시스템 정상 동작 확인 |
-| **Load** | 예상 부하 검증 | 50-100 | 10분 | 일반적인 운영 환경 시뮬레이션 |
-| **Stress** | 시스템 한계 확인 | 100-500 | 30분 | 점진적 부하 증가로 Breaking Point 탐색 |
-| **Spike** | 급증 대응력 확인 | 10→500→10 | 10분 | 급격한 트래픽 변화 대응 |
-| **Soak** | 장시간 안정성 | 100 | 2-4시간 | 메모리 누수, 리소스 고갈 확인 |
-
-### 4.2 예상 트래픽 기준
-
-| 항목 | 값 | 설명 |
-|------|-----|------|
-| Edge Device 수 | 100대 | 동시 연결 카메라 수 |
-| 이벤트/분/디바이스 | 10건 | 분당 과속 감지 이벤트 |
-| 총 이벤트/분 | 1,000건 | 피크 시간대 |
-| 총 이벤트/초 | ~17건 | 평균 TPS |
-
----
-
-## 5. K6 테스트 스크립트
-
-### 5.1 공통 설정
-
-**k6/common/config.js**
-```javascript
-// 환경 변수 또는 기본값
-export const CONFIG = {
-  // 서비스 URL
-  MAIN_SERVICE_URL: __ENV.MAIN_SERVICE_URL || 'http://main:8000',
-  
-  // MQTT 설정
-  MQTT_BROKER: __ENV.MQTT_BROKER || 'tcp://rabbitmq:1883',
-  MQTT_USER: __ENV.MQTT_USER || 'sa',
-  MQTT_PASS: __ENV.MQTT_PASS || '1234',
-  MQTT_TOPIC: 'detections/new',
-  
-  // 테스트 데이터
-  CAMERAS: ['cam_001', 'cam_002', 'cam_003', 'cam_004', 'cam_005'],
-  LOCATIONS: [
-    '서울시 강남구 테헤란로',
-    '서울시 서초구 반포대로',
-    '서울시 송파구 올림픽로',
-    '경기도 성남시 분당구',
-    '인천시 연수구 센트럴로',
-  ],
-};
-
-// 테스트용 Detection 메시지 생성
-export function generateDetectionMessage() {
-  const camera = CONFIG.CAMERAS[Math.floor(Math.random() * CONFIG.CAMERAS.length)];
-  const location = CONFIG.LOCATIONS[Math.floor(Math.random() * CONFIG.LOCATIONS.length)];
-  const speedLimit = [50, 60, 80, 100][Math.floor(Math.random() * 4)];
-  const detectedSpeed = speedLimit + Math.random() * 40 + 10; // 제한속도 + 10~50
-  
-  return JSON.stringify({
-    camera_id: camera,
-    location: location,
-    detected_speed: Math.round(detectedSpeed * 10) / 10,
-    speed_limit: speedLimit,
-    detected_at: new Date().toISOString(),
-    image_gcs_uri: `gs://test-bucket/${camera}/${Date.now()}.jpg`,
-  });
-}
-
-// 성능 임계값 (Thresholds)
-export const THRESHOLDS = {
-  // HTTP 요청
-  http_req_duration: ['p(95)<500', 'p(99)<1000'],
-  http_req_failed: ['rate<0.01'],
-  
-  // MQTT 발행
-  mqtt_publish_duration: ['p(95)<100'],
-  mqtt_publish_failed: ['rate<0.01'],
-  
-  // 커스텀 메트릭
-  detection_e2e_duration: ['p(95)<30000'], // End-to-End 30초 이내
-};
-```
-
-### 5.2 Smoke Test
-
-**k6/tests/smoke.js**
-```javascript
-import mqtt from 'k6/x/mqtt';
-import http from 'k6/http';
-import { check, sleep } from 'k6';
-import { Counter, Trend } from 'k6/metrics';
-import { CONFIG, generateDetectionMessage, THRESHOLDS } from '../common/config.js';
-
-// 커스텀 메트릭
-const mqttPublishDuration = new Trend('mqtt_publish_duration');
-const mqttPublishFailed = new Counter('mqtt_publish_failed');
-
-export const options = {
-  vus: 3,
-  duration: '1m',
-  thresholds: THRESHOLDS,
-};
-
-// MQTT 클라이언트 (VU당 1개)
-const client = new mqtt.Client(
-  CONFIG.MQTT_BROKER,
-  `k6-smoke-${__VU}-${Date.now()}`
-);
-
-export function setup() {
-  // API 헬스체크
-  const res = http.get(`${CONFIG.MAIN_SERVICE_URL}/health/`);
-  check(res, {
-    'API is healthy': (r) => r.status === 200,
-  });
-  
-  console.log('Smoke Test 시작: 기본 동작 확인');
-}
-
-export default function () {
-  // MQTT 연결
-  client.connect({
-    username: CONFIG.MQTT_USER,
-    password: CONFIG.MQTT_PASS,
-  });
-  
-  // Detection 메시지 발행
-  const message = generateDetectionMessage();
-  const startTime = Date.now();
-  
-  try {
-    client.publish(CONFIG.MQTT_TOPIC, message, 1, false);
-    mqttPublishDuration.add(Date.now() - startTime);
-    
-    check(null, {
-      'MQTT publish successful': () => true,
-    });
-  } catch (e) {
-    mqttPublishFailed.add(1);
-    console.error(`MQTT publish failed: ${e}`);
-  }
-  
-  client.disconnect();
-  
-  // API 조회 테스트
-  const apiRes = http.get(`${CONFIG.MAIN_SERVICE_URL}/api/v1/detections/pending/`);
-  check(apiRes, {
-    'API status 200': (r) => r.status === 200,
-    'API response time < 500ms': (r) => r.timings.duration < 500,
-  });
-  
-  sleep(1);
-}
-
-export function teardown() {
-  console.log('Smoke Test 완료');
-}
-```
-
-### 5.3 Load Test
-
-**k6/tests/load.js**
-```javascript
-import mqtt from 'k6/x/mqtt';
-import http from 'k6/http';
-import { check, sleep } from 'k6';
-import { Counter, Trend, Rate } from 'k6/metrics';
-import { CONFIG, generateDetectionMessage, THRESHOLDS } from '../common/config.js';
-
-// 커스텀 메트릭
-const mqttPublishDuration = new Trend('mqtt_publish_duration');
-const mqttPublishFailed = new Counter('mqtt_publish_failed');
-const detectionCreated = new Counter('detection_created');
-
-export const options = {
-  stages: [
-    { duration: '1m', target: 50 },   // Ramp-up
-    { duration: '8m', target: 50 },   // Steady state
-    { duration: '1m', target: 0 },    // Ramp-down
-  ],
-  thresholds: {
-    ...THRESHOLDS,
-    'detection_created': ['count>400'], // 10분간 최소 400건
-  },
-};
-
-let client;
-
-export function setup() {
-  console.log('Load Test 시작: 예상 부하 검증');
-  console.log(`Target: 50 VUs, 예상 TPS: ~17/s`);
-}
-
-export default function () {
-  // VU별 MQTT 클라이언트 생성
-  if (!client) {
-    client = new mqtt.Client(
-      CONFIG.MQTT_BROKER,
-      `k6-load-${__VU}-${Date.now()}`
-    );
-  }
-  
-  client.connect({
-    username: CONFIG.MQTT_USER,
-    password: CONFIG.MQTT_PASS,
-  });
-  
-  // 1. MQTT 메시지 발행 (Edge Device 시뮬레이션)
-  const message = generateDetectionMessage();
-  const startTime = Date.now();
-  
-  try {
-    client.publish(CONFIG.MQTT_TOPIC, message, 1, false);
-    mqttPublishDuration.add(Date.now() - startTime);
-    detectionCreated.add(1);
-    
-    check(null, { 'MQTT publish OK': () => true });
-  } catch (e) {
-    mqttPublishFailed.add(1);
-  }
-  
-  client.disconnect();
-  
-  // 2. API 부하 (프론트엔드 시뮬레이션)
-  const endpoints = [
-    '/api/v1/detections/',
-    '/api/v1/detections/pending/',
-    '/api/v1/detections/statistics/',
-  ];
-  
-  const endpoint = endpoints[Math.floor(Math.random() * endpoints.length)];
-  const apiRes = http.get(`${CONFIG.MAIN_SERVICE_URL}${endpoint}`);
-  
-  check(apiRes, {
-    'API status 200': (r) => r.status === 200,
-    'API response < 500ms': (r) => r.timings.duration < 500,
-  });
-  
-  // 초당 약 17건 (50 VU * 0.33건/VU/초)
-  sleep(3);
-}
-
-export function teardown() {
-  console.log('Load Test 완료');
-}
-```
-
-### 5.4 Stress Test
-
-**k6/tests/stress.js**
-```javascript
-import mqtt from 'k6/x/mqtt';
-import http from 'k6/http';
-import { check, sleep } from 'k6';
-import { Counter, Trend, Gauge } from 'k6/metrics';
-import { CONFIG, generateDetectionMessage, THRESHOLDS } from '../common/config.js';
-
-// 커스텀 메트릭
-const mqttPublishDuration = new Trend('mqtt_publish_duration');
-const mqttPublishFailed = new Counter('mqtt_publish_failed');
-const activeVUs = new Gauge('active_vus');
-const queueDepth = new Gauge('estimated_queue_depth');
-
-export const options = {
-  stages: [
-    // 점진적 부하 증가
-    { duration: '2m', target: 50 },
-    { duration: '5m', target: 50 },
-    { duration: '2m', target: 100 },
-    { duration: '5m', target: 100 },
-    { duration: '2m', target: 200 },
-    { duration: '5m', target: 200 },
-    { duration: '2m', target: 300 },
-    { duration: '5m', target: 300 },
-    // Breaking point 탐색
-    { duration: '2m', target: 500 },
-    { duration: '5m', target: 500 },
-    // Ramp-down
-    { duration: '2m', target: 0 },
-  ],
-  thresholds: {
-    http_req_duration: ['p(95)<2000'],  // Stress 시 완화
-    http_req_failed: ['rate<0.1'],       // 10% 미만 실패 허용
-    mqtt_publish_failed: ['rate<0.05'],  // 5% 미만 실패 허용
-  },
-};
-
-let client;
-
-export function setup() {
-  console.log('Stress Test 시작: 시스템 한계 확인');
-  console.log('단계: 50 → 100 → 200 → 300 → 500 VUs');
-}
-
-export default function () {
-  activeVUs.add(__VU);
-  
-  if (!client) {
-    client = new mqtt.Client(
-      CONFIG.MQTT_BROKER,
-      `k6-stress-${__VU}-${Date.now()}`
-    );
-  }
-  
-  try {
-    client.connect({
-      username: CONFIG.MQTT_USER,
-      password: CONFIG.MQTT_PASS,
-    });
-    
-    const message = generateDetectionMessage();
-    const startTime = Date.now();
-    
-    client.publish(CONFIG.MQTT_TOPIC, message, 1, false);
-    mqttPublishDuration.add(Date.now() - startTime);
-    
-    client.disconnect();
-  } catch (e) {
-    mqttPublishFailed.add(1);
-    console.error(`Stress error at VU ${__VU}: ${e}`);
-  }
-  
-  // API 부하
-  const apiRes = http.get(`${CONFIG.MAIN_SERVICE_URL}/api/v1/detections/`);
-  check(apiRes, {
-    'API responds': (r) => r.status === 200 || r.status === 503,
-  });
-  
-  // RabbitMQ Queue 깊이 확인 (추정)
-  try {
-    const rmqRes = http.get(
-      'http://rabbitmq:15672/api/queues/%2F/ocr_queue',
-      { auth: 'sa:1234' }
-    );
-    if (rmqRes.status === 200) {
-      const queue = JSON.parse(rmqRes.body);
-      queueDepth.add(queue.messages || 0);
-    }
-  } catch (e) {
-    // Queue 모니터링 실패 무시
-  }
-  
-  sleep(1);
-}
-
-export function teardown() {
-  console.log('Stress Test 완료');
-  console.log('Breaking Point 분석 필요');
-}
-```
-
-### 5.5 Spike Test
-
-**k6/tests/spike.js**
-```javascript
-import mqtt from 'k6/x/mqtt';
-import http from 'k6/http';
-import { check, sleep } from 'k6';
-import { Counter, Trend } from 'k6/metrics';
-import { CONFIG, generateDetectionMessage } from '../common/config.js';
-
-const mqttPublishDuration = new Trend('mqtt_publish_duration');
-const mqttPublishFailed = new Counter('mqtt_publish_failed');
-const recoveryTime = new Trend('recovery_time');
-
-export const options = {
-  stages: [
-    // 정상 상태
-    { duration: '1m', target: 10 },
-    // 급격한 스파이크
-    { duration: '10s', target: 500 },
-    // 스파이크 유지
-    { duration: '1m', target: 500 },
-    // 급격한 감소
-    { duration: '10s', target: 10 },
-    // 정상 상태 복귀
-    { duration: '2m', target: 10 },
-    // 두 번째 스파이크
-    { duration: '10s', target: 300 },
-    { duration: '1m', target: 300 },
-    { duration: '10s', target: 10 },
-    // 복구 확인
-    { duration: '2m', target: 10 },
-    // 종료
-    { duration: '30s', target: 0 },
-  ],
-  thresholds: {
-    http_req_duration: ['p(95)<3000'],  // Spike 시 완화된 임계값
-    mqtt_publish_failed: ['rate<0.1'],   // 10% 미만 실패 허용
-  },
-};
-
-let client;
-let preSpikeDuration = null;
-
-export function setup() {
-  console.log('Spike Test 시작: 급격한 트래픽 변화 대응력 확인');
-  console.log('시나리오: 10 → 500 → 10 → 300 → 10 VUs');
-}
-
-export default function () {
-  if (!client) {
-    client = new mqtt.Client(
-      CONFIG.MQTT_BROKER,
-      `k6-spike-${__VU}-${Date.now()}`
-    );
-  }
-  
-  try {
-    client.connect({
-      username: CONFIG.MQTT_USER,
-      password: CONFIG.MQTT_PASS,
-    });
-    
-    const message = generateDetectionMessage();
-    const startTime = Date.now();
-    
-    client.publish(CONFIG.MQTT_TOPIC, message, 1, false);
-    const duration = Date.now() - startTime;
-    mqttPublishDuration.add(duration);
-    
-    // 스파이크 전 기준 응답 시간 저장
-    if (__ITER < 60 && !preSpikeDuration) {
-      preSpikeDuration = duration;
-    }
-    
-    // 복구 시간 측정 (스파이크 후 정상 응답으로 돌아오는 시간)
-    if (__ITER > 200 && preSpikeDuration) {
-      if (duration <= preSpikeDuration * 1.5) {
-        recoveryTime.add(__ITER);
-      }
-    }
-    
-    client.disconnect();
-  } catch (e) {
-    mqttPublishFailed.add(1);
-  }
-  
-  // API 응답 확인
-  const apiRes = http.get(`${CONFIG.MAIN_SERVICE_URL}/api/v1/detections/pending/`);
-  check(apiRes, {
-    'API responds during spike': (r) => r.status === 200 || r.status === 503,
-  });
-  
-  sleep(0.5);
-}
-
-export function teardown() {
-  console.log('Spike Test 완료');
-  console.log('복구 시간 분석 필요');
-}
-```
-
-### 5.6 Soak Test
-
-**k6/tests/soak.js**
-```javascript
-import mqtt from 'k6/x/mqtt';
-import http from 'k6/http';
-import { check, sleep } from 'k6';
-import { Counter, Trend, Gauge } from 'k6/metrics';
-import { CONFIG, generateDetectionMessage } from '../common/config.js';
-
-const mqttPublishDuration = new Trend('mqtt_publish_duration');
-const mqttPublishFailed = new Counter('mqtt_publish_failed');
-const memoryUsage = new Gauge('memory_usage_estimate');
-const dbConnections = new Gauge('db_connections_estimate');
-
-export const options = {
-  stages: [
-    { duration: '5m', target: 100 },   // Ramp-up
-    { duration: '4h', target: 100 },   // 4시간 유지 (조정 가능: 2h, 8h)
-    { duration: '5m', target: 0 },     // Ramp-down
-  ],
-  thresholds: {
-    http_req_duration: ['p(95)<500', 'p(99)<1000'],
-    http_req_failed: ['rate<0.01'],
-    mqtt_publish_failed: ['rate<0.01'],
-  },
-};
-
-let client;
-let iterationCount = 0;
-
-export function setup() {
-  console.log('Soak Test 시작: 장시간 안정성 확인');
-  console.log('Duration: 4시간, VUs: 100');
-  console.log('확인 항목: 메모리 누수, DB 커넥션 풀 고갈, 성능 저하');
-}
-
-export default function () {
-  iterationCount++;
-  
-  if (!client) {
-    client = new mqtt.Client(
-      CONFIG.MQTT_BROKER,
-      `k6-soak-${__VU}-${Date.now()}`
-    );
-  }
-  
-  try {
-    client.connect({
-      username: CONFIG.MQTT_USER,
-      password: CONFIG.MQTT_PASS,
-    });
-    
-    const message = generateDetectionMessage();
-    const startTime = Date.now();
-    
-    client.publish(CONFIG.MQTT_TOPIC, message, 1, false);
-    mqttPublishDuration.add(Date.now() - startTime);
-    
-    client.disconnect();
-  } catch (e) {
-    mqttPublishFailed.add(1);
-  }
-  
-  // API 호출
-  const apiRes = http.get(`${CONFIG.MAIN_SERVICE_URL}/api/v1/detections/`);
-  check(apiRes, {
-    'API status 200': (r) => r.status === 200,
-    'API response < 500ms': (r) => r.timings.duration < 500,
-  });
-  
-  // 주기적으로 시스템 상태 확인 (10분마다)
-  if (iterationCount % 600 === 0) {
-    console.log(`Checkpoint at iteration ${iterationCount}`);
-    
-    // Health check
-    const healthRes = http.get(`${CONFIG.MAIN_SERVICE_URL}/health/`);
-    if (healthRes.status !== 200) {
-      console.error('Health check failed!');
-    }
-    
-    // RabbitMQ Queue 상태
-    try {
-      const rmqRes = http.get(
-        'http://rabbitmq:15672/api/overview',
-        { auth: 'sa:1234' }
-      );
-      if (rmqRes.status === 200) {
-        const overview = JSON.parse(rmqRes.body);
-        console.log(`RabbitMQ Messages: ${overview.queue_totals?.messages || 0}`);
-      }
-    } catch (e) {
-      // 무시
-    }
-  }
-  
-  sleep(1);
-}
-
-export function teardown() {
-  console.log('Soak Test 완료');
-  console.log(`Total iterations: ${iterationCount}`);
-  console.log('메모리 사용량 그래프 및 성능 추이 분석 필요');
-}
-```
-
----
-
-## 6. 테스트 실행 방법
-
-### 6.1 테스트 환경 시작
-
-```bash
-# 1. 테스트 환경 시작
-cd docs
-docker compose -f docker-compose.test.yml up -d
-
-# 2. 서비스 준비 대기
-sleep 30
-
-# 3. 헬스체크
-curl http://localhost:8000/health/
-```
-
-### 6.2 K6 테스트 실행
-
-```bash
-# Smoke Test (1분)
-docker compose -f docker-compose.test.yml run k6 run /scripts/tests/smoke.js
-
-# Load Test (10분)
-docker compose -f docker-compose.test.yml run k6 run /scripts/tests/load.js
-
-# Stress Test (37분)
-docker compose -f docker-compose.test.yml run k6 run /scripts/tests/stress.js
-
-# Spike Test (9분)
-docker compose -f docker-compose.test.yml run k6 run /scripts/tests/spike.js
-
-# Soak Test (4시간+)
-docker compose -f docker-compose.test.yml run k6 run /scripts/tests/soak.js
-```
-
-### 6.3 결과 확인
-
-- **Grafana Dashboard**: http://localhost:3000
-- **RabbitMQ Management**: http://localhost:15672 (sa/1234)
-- **Flower (Celery)**: http://localhost:5555
-
----
-
-## 7. 메트릭 및 분석
-
-### 7.1 핵심 메트릭
-
-| 메트릭 | 설명 | 목표값 |
-|--------|------|--------|
-| `http_req_duration` (p95) | API 응답 시간 | < 500ms |
-| `mqtt_publish_duration` (p95) | MQTT 발행 시간 | < 100ms |
-| `http_req_failed` | API 실패율 | < 1% |
-| `mqtt_publish_failed` | MQTT 발행 실패율 | < 1% |
-| `detection_e2e_duration` | 감지→알림 전체 시간 | < 30초 |
-
-### 7.2 RabbitMQ 메트릭
-
-| 메트릭 | 설명 | 경고 임계값 |
-|--------|------|-------------|
-| Queue Depth (ocr_queue) | OCR 대기 메시지 수 | > 1000 |
-| Queue Depth (fcm_queue) | FCM 대기 메시지 수 | > 500 |
-| Consumer Count | 활성 Consumer 수 | = 0 (장애) |
-| Message Rate | 초당 메시지 처리량 | 감소 추세 |
-
-### 7.3 시스템 메트릭
-
-| 메트릭 | 설명 | 경고 임계값 |
-|--------|------|-------------|
-| CPU Usage | CPU 사용률 | > 80% |
-| Memory Usage | 메모리 사용률 | > 85% |
-| DB Connections | DB 커넥션 수 | > Pool Size 80% |
-| Network I/O | 네트워크 트래픽 | 급격한 변화 |
-
-### 7.4 Grafana Dashboard JSON
-
-**grafana/dashboards/k6-performance.json**
-```json
-{
-  "dashboard": {
-    "title": "K6 Performance Test Dashboard",
-    "panels": [
-      {
-        "title": "Virtual Users",
-        "type": "graph",
-        "targets": [
-          {
-            "query": "SELECT mean(\"value\") FROM \"k6_vus\" WHERE $timeFilter GROUP BY time(10s)",
-            "alias": "VUs"
-          }
-        ]
-      },
-      {
-        "title": "HTTP Request Duration (p95)",
-        "type": "graph",
-        "targets": [
-          {
-            "query": "SELECT percentile(\"value\", 95) FROM \"k6_http_req_duration\" WHERE $timeFilter GROUP BY time(10s)",
-            "alias": "p95"
-          }
-        ]
-      },
-      {
-        "title": "MQTT Publish Duration",
-        "type": "graph",
-        "targets": [
-          {
-            "query": "SELECT mean(\"value\") FROM \"k6_mqtt_publish_duration\" WHERE $timeFilter GROUP BY time(10s)",
-            "alias": "mean"
-          }
-        ]
-      },
-      {
-        "title": "Error Rate",
-        "type": "graph",
-        "targets": [
-          {
-            "query": "SELECT sum(\"value\") FROM \"k6_http_req_failed\" WHERE $timeFilter GROUP BY time(10s)",
-            "alias": "HTTP Errors"
-          },
-          {
-            "query": "SELECT sum(\"value\") FROM \"k6_mqtt_publish_failed\" WHERE $timeFilter GROUP BY time(10s)",
-            "alias": "MQTT Errors"
-          }
-        ]
-      },
-      {
-        "title": "Requests per Second",
-        "type": "graph",
-        "targets": [
-          {
-            "query": "SELECT count(\"value\") FROM \"k6_http_reqs\" WHERE $timeFilter GROUP BY time(1s)",
-            "alias": "RPS"
-          }
-        ]
-      }
-    ]
-  }
-}
-```
-
----
-
-## 8. 테스트 결과 분석 체크리스트
-
-### 8.1 Smoke Test
-- [ ] 모든 컴포넌트 정상 동작
-- [ ] MQTT → Django → RabbitMQ 흐름 확인
-- [ ] API 응답 정상
-
-### 8.2 Load Test
-- [ ] 목표 TPS 달성 (17건/초)
-- [ ] p95 응답 시간 < 500ms
-- [ ] 에러율 < 1%
-- [ ] Queue 백로그 축적 없음
-
-### 8.3 Stress Test
-- [ ] Breaking Point 식별 (VU 수, TPS)
-- [ ] 장애 발생 지점 확인
-- [ ] 리소스 병목 구간 확인 (CPU/Memory/DB/Queue)
-- [ ] 장애 시 Graceful Degradation 여부
-
-### 8.4 Spike Test
-- [ ] 스파이크 시 시스템 다운 없음
-- [ ] 복구 시간 측정
-- [ ] 메시지 유실 여부 확인
-- [ ] Auto-scaling 동작 확인 (적용 시)
-
-### 8.5 Soak Test
-- [ ] 메모리 누수 없음 (일정한 메모리 사용량)
-- [ ] DB 커넥션 풀 안정
-- [ ] 성능 저하 없음 (시간 경과에 따른 응답 시간)
-- [ ] 로그 파일 사이즈 관리
-
----
-
-## 9. 트러블슈팅
-
-### 9.1 일반적인 문제
-
-| 문제 | 원인 | 해결 |
-|------|------|------|
-| MQTT 연결 실패 | RabbitMQ MQTT Plugin 미활성화 | `rabbitmq-plugins enable rabbitmq_mqtt` |
-| Queue 백로그 증가 | Worker 처리량 부족 | Worker concurrency 증가 |
-| DB 커넥션 고갈 | Pool Size 부족 | `CONN_MAX_AGE`, `pool_size` 증가 |
-| OOM Kill | 메모리 부족 | Container 메모리 제한 증가 |
-
-### 9.2 성능 병목 해결
-
-```bash
-# RabbitMQ Queue 상태 확인
-curl -u sa:1234 http://localhost:15672/api/queues/%2F/ocr_queue
-
-# MySQL 커넥션 상태 확인
-mysql -u sa -p1234 -e "SHOW PROCESSLIST;"
-
-# Celery Worker 상태 확인
-celery -A config inspect active
-
-# Docker 리소스 사용량
-docker stats
-```
-
----
-
-## 10. 권장 테스트 순서
-
-```
-1. Smoke Test (1분)
-   └─ 기본 동작 확인
-   
-2. Load Test (10분)
-   └─ 예상 부하 검증
-   
-3. Stress Test (37분)
-   └─ 시스템 한계 확인
-   
-4. Spike Test (9분)
-   └─ 급증 대응력 확인
-   
-5. Soak Test (4시간)
-   └─ 장시간 안정성 확인
-```
-
-각 테스트 후 결과를 분석하고, 발견된 문제를 해결한 후 다음 테스트를 진행합니다.
-
diff --git a/docs/PERFORMANCE_TEST_GUIDE.md b/docs/PERFORMANCE_TEST_GUIDE.md
new file mode 100644
index 0000000..12d20ff
--- /dev/null
+++ b/docs/PERFORMANCE_TEST_GUIDE.md
@@ -0,0 +1,864 @@
+# 성능 테스트 가이드
+
+SpeedCam IoT 백엔드의 성능, 안정성, 파이프라인 처리 능력을 검증하기 위한 종합 가이드입니다. HTTP API 부하 테스트(k6)와 IoT 파이프라인 부하 테스트(MQTT)를 다룹니다.
+
+---
+
+## 목차
+
+1. [사전 준비](#1-사전-준비)
+2. [모니터링 대시보드](#2-모니터링-대시보드)
+3. [HTTP API 부하 테스트 (k6)](#3-http-api-부하-테스트-k6)
+4. [IoT 파이프라인 부하 테스트 (MQTT)](#4-iot-파이프라인-부하-테스트-mqtt)
+5. [End-to-End 검증 체크리스트](#5-end-to-end-검증-체크리스트)
+6. [트러블슈팅](#6-트러블슈팅)
+7. [정리 및 종료](#7-정리-및-종료)
+
+---
+
+## 1. 사전 준비
+
+### 1.1 필요 도구
+
+| 도구 | 용도 | 설치 방법 |
+|------|------|----------|
+| Docker | 컨테이너 실행 | https://docs.docker.com/get-docker/ |
+| Docker Compose | 다중 컨테이너 관리 | Docker Desktop 포함 |
+| Python 3.x | MQTT 파이프라인 테스트 | 기본 설치됨 |
+| paho-mqtt | MQTT 클라이언트 | `pip install paho-mqtt` |
+| curl | API 요청 테스트 | 기본 설치됨 |
+
+### 1.2 환경 시작
+
+**중요**: `docker-compose.yml`이 `speedcam-network`를 생성하고, `docker-compose.monitoring.yml`은 이를 `external: true`로 참조합니다. 반드시 순서대로 또는 `-f` 플래그로 함께 시작하세요.
+
+```bash
+# 방법 1: 앱 + 모니터링 함께 시작 (권장)
+cd docker
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml up -d
+```
+
+```bash
+# 방법 2: 순차 시작
+cd docker
+docker compose -f docker-compose.yml up -d
+docker compose -f docker-compose.monitoring.yml up -d
+```
+
+### 1.3 macOS 참고사항
+
+- **cAdvisor는 Linux 전용**: `docker-compose.monitoring.yml`에 `profiles: [linux]`가 설정되어 있으므로 macOS에서는 자동 제외됩니다
+- **Linux에서 cAdvisor 포함**:
+  ```bash
+  docker compose -f docker-compose.yml -f docker-compose.monitoring.yml --profile linux up -d
+  ```
+- **권장 Docker Desktop 메모리**: 8GB 이상
+
+### 1.4 서비스 상태 확인
+
+```bash
+# 전체 컨테이너 상태 확인
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml ps
+```
+
+**예상 상태**: 모든 컨테이너가 `Up` 상태
+
+```bash
+# Prometheus 타겟 상태 확인
+curl -s http://localhost:9090/api/v1/targets | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+print('Prometheus Scrape Targets:')
+for t in data['data']['activeTargets']:
+    status = '✓ UP' if t['health'] == 'up' else '✗ DOWN'
+    print(f\"  {t['labels']['job']:20s} {t['health']:5s} {t['labels']['instance']}\")
+"
+```
+
+**예상 출력**:
+```
+Prometheus Scrape Targets:
+  django               up     main:8000
+  otel-collector       up     otel-collector:8889
+  rabbitmq             up     rabbitmq:15692
+  mysql                up     mysqld-exporter:9104
+  celery               up     celery-exporter:9808
+  cadvisor             up     cadvisor:8080          (Linux only)
+```
+
+---
+
+## 2. 모니터링 대시보드
+
+### 2.1 접속 정보
+
+| 서비스 | URL | 인증 | 용도 |
+|--------|-----|------|------|
+| **Grafana** | http://localhost:3000 | admin / admin | 통합 대시보드 (메트릭, 로그, 트레이스) |
+| **Prometheus** | http://localhost:9090 | 없음 | 메트릭 저장소 및 PromQL 쿼리 |
+| **Jaeger** | http://localhost:16686 | 없음 | 분산 트레이싱 UI |
+| **RabbitMQ** | http://localhost:15672 | sa / 1234 | 큐 모니터링 |
+| **Flower** | http://localhost:5555 | 없음 | Celery 태스크 모니터링 |
+
+### 2.2 Grafana 대시보드 Import
+
+시작 시 자동으로 대시보드가 프로비저닝되지만, 추가 대시보드는 수동 import:
+
+1. Grafana 접속: http://localhost:3000
+2. Dashboards → New → Import
+3. Dashboard ID 입력:
+
+| 대시보드 | ID | 데이터소스 | 용도 |
+|---------|-----|-----------|------|
+| Django Prometheus | 17658 | Prometheus | HTTP 요청, 응답시간, 에러율 |
+| Celery Monitoring | 17509 | Prometheus | 태스크 성공/실패, 큐 깊이 |
+| RabbitMQ Overview | 10991 | Prometheus | 메시지 rate, 큐 깊이 |
+| MySQL Overview | 14057 | Prometheus | 쿼리 수, 커넥션, 슬로우 쿼리 |
+| K6 Load Testing | 19665 | Prometheus | k6 부하 테스트 결과 (실시간) |
+
+### 2.3 주요 메트릭 보기
+
+**Django HTTP 메트릭** (자동 수집):
+```
+http://localhost:3000/d/<dashboard-id>
+```
+
+**Jaeger 트레이스** (요청 플로우 추적):
+```
+http://localhost:16686 → Services → speedcam-api → 최근 트레이스 보기
+```
+
+**Loki 로그** (구조화된 로그):
+```
+Grafana → Explore → Data source: Loki
+쿼리: {service="main"}
+```
+
+---
+
+## 3. HTTP API 부하 테스트 (k6)
+
+### 3.1 테스트 대상
+
+REST API 엔드포인트 검증 (IoT 파이프라인 제외):
+
+| 엔드포인트 | 메서드 | 용도 |
+|-----------|--------|------|
+| `/health/` | GET | 헬스 체크 |
+| `/api/v1/vehicles/` | GET, POST, PUT, DELETE | 차량 CRUD |
+| `/api/v1/detections/` | GET | 검출 목록 조회 |
+| `/api/v1/notifications/` | GET | 알림 목록 조회 |
+
+### 3.2 설치
+
+paho-mqtt는 MQTT 테스트에만 필요합니다. k6 테스트는 Docker 컨테이너에서 실행되므로 호스트 설치 불필요합니다.
+
+### 3.3 실행 방법
+
+```bash
+cd docker
+
+# 기본 실행: Prometheus에 결과 기록
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml \
+  run k6 run --out experimental-prometheus-rw /scripts/load-test.js
+```
+
+**선택 사항**: 환경 변수 오버라이드
+
+```bash
+# 커스텀 대상 서버 지정
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml \
+  run -e MAIN_SERVICE_URL=http://localhost:8000 k6 \
+  run --out experimental-prometheus-rw /scripts/load-test.js
+```
+
+### 3.4 시나리오별 테스트
+
+`load-test.js`에 3가지 시나리오가 정의되어 있습니다. 각 시나리오는 startTime이 다르므로 한 번의 실행으로 모두 테스트됩니다.
+
+| 시나리오 | VU 범위 | 시간 | 시작 시간 | 용도 | 기대 결과 |
+|----------|---------|------|----------|------|----------|
+| **smoke** | 1 | 10초 | 0s | 기본 동작 확인 | 에러 0%, 응답 <100ms |
+| **average_load** | 0→10→0 | ~100초 | 15s | 평균 부하 검증 | p95 <500ms, 에러 <1% |
+| **spike** | 0→30→0 | ~25초 | 120s | 스파이크 처리 능력 | p99 <1000ms, 에러 <5% |
+
+**총 실행 시간**: ~2분 30초
+
+### 3.5 실행 중 모니터링
+
+실시간으로 다른 터미널에서 메트릭 확인:
+
+```bash
+# Prometheus UI에서 확인
+open http://localhost:9090/graph
+# 쿼리: rate(k6_http_reqs_total[1m])
+```
+
+```bash
+# Grafana K6 대시보드 (ID: 19665) 보기
+open http://localhost:3000/d/K6-dashboard
+```
+
+### 3.6 결과 해석
+
+k6 실행 완료 후 stdout에 요약이 표시됩니다:
+
+```
+     checks.........................: 100.00% ✓ 5000  ✗ 0
+     data_received..................: 1.2 MB  ✓
+     data_sent.......................: 850 kB  ✓
+     http_req_blocked...............: avg=1.2ms   min=100µs   max=50ms   p(90)=2.1ms   p(95)=3.5ms
+     http_req_connecting............: avg=0.8ms   min=0µs     max=40ms   p(90)=1.5ms   p(95)=2.2ms
+     http_req_duration..............: avg=125ms   min=50ms    max=2s     p(90)=350ms   p(95)=450ms
+     http_req_failed................: 0.00% ✓ 0    ✗ 5000
+     http_req_receiving.............: avg=2.5ms   min=0.5ms   max=20ms   p(90)=4ms     p(95)=5ms
+     http_req_sending...............: avg=0.5ms   min=0.1ms   max=5ms    p(90)=1ms     p(95)=1ms
+     http_req_tls_handshaking.......: avg=0ms     min=0µs     max=0s     p(90)=0s      p(95)=0s
+     http_req_waiting...............: avg=122ms   min=48ms    max=1.9s   p(90)=348ms   p(95)=448ms
+     http_reqs.......................: 5000    199.31/s
+     iteration_duration.............: avg=2.5s    min=2s      max=30s    p(90)=2.8s    p(95)=3.1s
+     iterations......................: 5000    199.31/s
+```
+
+**주요 메트릭**:
+- `checks`: 테스트 검증 통과율 (100%이어야 함)
+- `http_req_duration` (p95): 95% 요청의 응답시간 (목표 <500ms)
+- `http_req_failed`: 실패율 (0%이어야 함)
+- `http_reqs`: 초당 처리한 요청 수 (RPS)
+
+### 3.7 맞춤형 시나리오 작성
+
+`load-test.js`를 수정하여 커스텀 시나리오를 추가할 수 있습니다. 자세한 내용은 [k6 공식 문서](https://k6.io/docs/get-started/running-k6/)를 참고하세요.
+
+---
+
+## 4. IoT 파이프라인 부하 테스트 (MQTT)
+
+### 4.1 테스트 대상
+
+실제 IoT 카메라 동작을 시뮬레이션하여 전체 파이프라인을 검증합니다:
+
+```
+MQTT 메시지 발행 (Raspberry Pi 시뮬)
+    ↓
+RabbitMQ 큐에 저장
+    ↓
+Detection 생성 (pending)
+    ↓
+OCR Worker (이미지 처리)
+    ↓
+Alert Worker (FCM 알림)
+    ↓
+완료 (completed)
+```
+
+### 4.2 사전 준비
+
+**호스트에서 실행하는 경우**:
+```bash
+pip install paho-mqtt
+```
+
+### 4.3 실행 방법
+
+**기본 실행** (호스트):
+```bash
+python docker/k6/mqtt-load-test.py \
+  --workers 5 \
+  --rate 2 \
+  --duration 60
+```
+
+**환경 변수 오버라이드**:
+```bash
+MQTT_HOST=localhost MQTT_PORT=1883 python docker/k6/mqtt-load-test.py \
+  --workers 5 --rate 2 --duration 60
+```
+
+### 4.4 테스트 단계별 파라미터
+
+| 단계 | Workers | Rate(/s) | Duration | 총 메시지 | 용도 | 예상 처리 시간 |
+|------|---------|----------|----------|-----------|------|----------------|
+| **Smoke** | 1 | 1 | 10s | ~10 | 기본 동작 확인 | ~30초 |
+| **Load** | 5 | 2 | 60s | ~600 | 일반 부하 검증 | ~5분 |
+| **Stress** | 20 | 5 | 120s | ~12,000 | 시스템 한계 확인 | ~30분 |
+| **Soak** | 5 | 2 | 3600s | ~36,000 | 장시간 안정성 | ~2시간 |
+
+**추천 시작 순서**:
+1. Smoke 테스트로 연결성 확인
+2. Load 테스트로 정상 동작 확인
+3. Stress 테스트로 한계 확인
+
+### 4.5 메시지 형식
+
+MQTT 메시지는 다음 JSON 형식으로 발행됩니다:
+
+```json
+{
+  "camera_id": "CAM-001",
+  "location": "서울시 강남구 테헤란로",
+  "detected_speed": 95.3,
+  "speed_limit": 60.0,
+  "detected_at": "2024-01-01T12:00:00+09:00",
+  "image_gcs_uri": "gs://speedcam-bucket/detections/1704067200000-1234.jpg"
+}
+```
+
+**필드 설명**:
+- `camera_id`: 카메라 ID (CAM-001 ~ CAM-020)
+- `location`: 카메라 위치 (실제 한국 도로명)
+- `detected_speed`: 감지된 속도 (제한속도 + 5~50km/h 초과)
+- `speed_limit`: 해당 구간 제한속도 (60, 80, 100, 110 중 선택)
+- `detected_at`: ISO 8601 형식의 감지 시간 (한국 표준시)
+- `image_gcs_uri`: GCS에 저장된 이미지 경로 (시뮬레이션용 경로)
+
+### 4.6 실행 중 모니터링
+
+테스트 실행 중 다른 터미널에서 진행 상황을 모니터링합니다:
+
+**RabbitMQ 큐 상태**:
+```bash
+curl -s -u sa:1234 http://localhost:15672/api/queues/%2F | python3 -c "
+import json, sys
+queues = json.load(sys.stdin)
+print('RabbitMQ Queue Status:')
+for q in queues:
+    if q['name'] in ('detections_queue', 'ocr_queue', 'fcm_queue'):
+        print(f\"  {q['name']:20s} messages={q.get('messages', 0):6d} consumers={q.get('consumers', 0)}\")
+"
+```
+
+**Celery 태스크 상태**:
+```bash
+curl -s http://localhost:5555/api/workers | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+print('Celery Workers:')
+for worker, info in data.items():
+    print(f\"  {worker:30s} {info.get('status', 'unknown')}\")
+"
+```
+
+**Jaeger 트레이스 (선택)**:
+```bash
+open http://localhost:16686
+# Services → speedcam-api → Detection 또는 OCR 작업 선택
+```
+
+### 4.7 결과 확인
+
+MQTT 테스트 완료 후 stdout에 통계가 표시됩니다:
+
+```
+=== MQTT Load Test Complete ===
+Total Published: 600
+Failed: 0
+Success Rate: 100.00%
+Avg Latency: 245ms
+Min Latency: 50ms
+Max Latency: 1200ms
+Total Duration: 65 seconds
+Messages/sec: 9.23
+```
+
+**해석**:
+- **Success Rate**: 100%이어야 함 (메시지 발행 성공)
+- **Avg Latency**: MQTT 발행 시간 (네트워크 지연)
+- **Total Duration**: 부하 테스트 총 소요 시간
+
+---
+
+## 5. End-to-End 검증 체크리스트
+
+MQTT 부하 테스트 실행 후 다음 항목들을 확인하여 파이프라인이 정상 동작하는지 검증합니다.
+
+### 5.1 Detection 처리 상태
+
+```bash
+# 전체 Detection 조회
+curl -s http://localhost:8000/api/v1/detections/ | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+print(f\"Total Detections: {data['count']}\")
+print()
+
+# 상태별 카운트 추출
+results = data['results']
+status_counts = {}
+for r in results:
+    status = r.get('ocr_status', 'unknown')
+    status_counts[status] = status_counts.get(status, 0) + 1
+
+print('Status Distribution:')
+for status, count in sorted(status_counts.items()):
+    print(f\"  {status:15s}: {count:4d}\")
+"
+```
+
+**정상 상태**:
+- 대부분이 `completed` 상태
+- 일부 `processing` 또는 `pending` (최근 생성된 건)
+- `failed` 건이 있으면 OCR Worker 로그 확인: `docker logs speedcam-ocr`
+
+```bash
+# 최근 생성된 Detection 확인
+curl -s "http://localhost:8000/api/v1/detections/?ordering=-detected_at&limit=5" | \
+  python3 -m json.tool | head -50
+```
+
+### 5.2 Jaeger 분산 트레이스 확인
+
+트레이스를 통해 요청이 전체 시스템을 거치는 과정을 추적합니다.
+
+```bash
+# 사용 가능한 서비스 확인
+curl -s http://localhost:16686/api/services | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+print('Jaeger Services:')
+for service in data['data']:
+    print(f\"  - {service}\")
+"
+```
+
+**예상 서비스**:
+- `speedcam-api`: Django 메인 애플리케이션
+- `speedcam-ocr`: OCR Worker (Celery)
+- `speedcam-alert`: Alert Worker (Celery)
+
+```bash
+# 최근 트레이스 조회 (speedcam-api)
+curl -s "http://localhost:16686/api/traces?service=speedcam-api&limit=3" | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+print('Recent Traces (speedcam-api):')
+for trace in data['data'][:3]:
+    trace_id = trace['traceID'][:16]
+    num_spans = len(trace['spans'])
+    operation = trace['spans'][0]['operationName']
+    duration_ms = (trace['spans'][0]['endTime'] - trace['spans'][0]['startTime']) / 1000
+    print(f\"  {trace_id}... | Spans: {num_spans:2d} | {operation:30s} | {duration_ms:6.1f}ms\")
+"
+```
+
+**정상 구성**:
+- Health Check: 1-2 spans (빠름)
+- Vehicle Create: 3-5 spans (DB 쿼리 포함)
+- Detection Create: 5-10 spans (MQTT, RabbitMQ, DB)
+- OCR Task: 7-15 spans (GCS, API, DB)
+
+### 5.3 Loki 로그 확인
+
+구조화된 로그를 통해 각 컴포넌트의 동작을 확인합니다.
+
+```bash
+# Loki에서 수집된 로그 스트림 확인
+curl -sG http://localhost:3100/loki/api/v1/labels | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+print('Loki Labels:')
+print(f\"  Available labels: {', '.join(data['data'][:5])}...\")
+"
+```
+
+```bash
+# speedcam 컨테이너의 최근 로그 (Loki)
+curl -sG "http://localhost:3100/loki/api/v1/query" \
+  --data-urlencode 'query={container=~"speedcam.*"}' \
+  --data-urlencode 'limit=10' | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+streams = data['data']['result']
+print(f'Log Streams Found: {len(streams)}')
+for stream in streams[:3]:
+    container = stream['stream'].get('container', 'unknown')
+    num_entries = len(stream['values'])
+    print(f\"  {container:30s}: {num_entries} log entries\")
+"
+```
+
+**Grafana UI에서 로그 보기**:
+1. Grafana → Explore → Loki
+2. 쿼리: `{container=~"speedcam.*"}`
+3. 각 로그 라인의 `trace_id=` 클릭 → Jaeger 트레이스 자동 이동
+
+### 5.4 RabbitMQ 큐 상태
+
+MQTT 메시지 처리 파이프라인의 큐 상태를 확인합니다.
+
+```bash
+# 큐별 메시지 수 확인
+curl -s -u sa:1234 http://localhost:15672/api/queues/%2F | python3 -c "
+import json, sys
+queues = json.load(sys.stdin)
+print('RabbitMQ Queue Status:')
+print(f\"{'Queue Name':<20} {'Messages':>10} {'Consumers':>10} {'Ready':>10} {'Unacked':>10}\")
+print('-' * 60)
+for q in queues:
+    if q['name'] in ('detections_queue', 'ocr_queue', 'fcm_queue', 'dlq_queue'):
+        name = q['name']
+        msgs = q.get('messages', 0)
+        consumers = q.get('consumers', 0)
+        ready = q.get('messages_ready', 0)
+        unacked = q.get('messages_unacknowledged', 0)
+        print(f'{name:<20} {msgs:>10} {consumers:>10} {ready:>10} {unacked:>10}')
+"
+```
+
+**정상 상태**:
+- **detections_queue**: 0 (Detection 생성 후 즉시 처리)
+- **ocr_queue**: 0-10 (처리 중)
+- **fcm_queue**: 0-5 (처리 중)
+- **dlq_queue**: 0 (에러 없음)
+- **consumers**: 각 큐당 1 이상 (worker가 리스닝 중)
+
+### 5.5 Prometheus 메트릭 확인
+
+시스템 성능 메트릭을 Prometheus PromQL로 확인합니다.
+
+```bash
+# Django HTTP 요청 메트릭
+curl -s http://localhost:9090/api/v1/query --data-urlencode \
+  'query=rate(django_http_requests_total[5m])' | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+result = data['data']['result']
+if result:
+    print(f'Django HTTP Request Rate: {len(result)} series found')
+    print(f'  Current RPS: {float(result[0][\"value\"][1]):.1f}')
+else:
+    print('No Django metrics found')
+"
+```
+
+```bash
+# Celery 태스크 메트릭
+curl -s http://localhost:9090/api/v1/query --data-urlencode \
+  'query=rate(celery_tasks_total[5m])' | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+result = data['data']['result']
+if result:
+    print(f'Celery Task Rate: {len(result)} series found')
+    for r in result[:3]:
+        state = r['metric'].get('state', 'unknown')
+        rate = float(r['value'][1])
+        print(f\"  {state:10s}: {rate:.1f} tasks/sec\")
+else:
+    print('No Celery metrics found')
+"
+```
+
+### 5.6 DB 성능 메트릭
+
+```bash
+# MySQL 활성 커넥션 수
+curl -s http://localhost:9090/api/v1/query --data-urlencode \
+  'query=mysql_global_status_threads_connected' | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+result = data['data']['result']
+if result:
+    value = float(result[0]['value'][1])
+    print(f'Active MySQL Connections: {int(value)}')
+else:
+    print('No MySQL metrics found')
+"
+```
+
+### 5.7 컨테이너 리소스 사용률
+
+```bash
+# 각 컨테이너 CPU 사용률 (%) - cAdvisor 필요
+curl -s http://localhost:9090/api/v1/query --data-urlencode \
+  'query=rate(container_cpu_usage_seconds_total{name=~"speedcam-.*"}[5m])*100' | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+result = data['data']['result']
+if result:
+    print('Container CPU Usage (%):')
+    for r in result[:5]:
+        name = r['metric'].get('name', 'unknown')
+        cpu_usage = float(r['value'][1])
+        print(f\"  {name:30s}: {cpu_usage:6.2f}%\")
+else:
+    print('No cAdvisor metrics found (Linux only)')
+"
+```
+
+```bash
+# 각 컨테이너 메모리 사용량 (MB) - cAdvisor 필요
+curl -s http://localhost:9090/api/v1/query --data-urlencode \
+  'query=container_memory_usage_bytes{name=~"speedcam-.*"}/1024/1024' | python3 -c "
+import json, sys
+data = json.load(sys.stdin)
+result = data['data']['result']
+if result:
+    print('Container Memory Usage (MB):')
+    for r in result[:5]:
+        name = r['metric'].get('name', 'unknown')
+        memory_mb = float(r['value'][1])
+        print(f\"  {name:30s}: {memory_mb:7.1f} MB\")
+else:
+    print('No cAdvisor metrics found (Linux only)')
+"
+```
+
+---
+
+## 6. 트러블슈팅
+
+### 6.1 docker-compose 실행 오류
+
+**오류**: `network speedcam-network not found`
+
+**원인**: 모니터링 스택만 단독으로 시작함
+
+**해결**:
+```bash
+# 앱 스택을 먼저 시작
+docker compose -f docker-compose.yml up -d
+
+# 그 다음 모니터링 추가
+docker compose -f docker-compose.monitoring.yml up -d
+
+# 또는 함께 시작
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml up -d
+```
+
+### 6.2 Prometheus 타겟이 DOWN
+
+**오류**: Prometheus 대시보드에서 일부 타겟이 DOWN 상태
+
+**celery-exporter가 재시작되는 경우**:
+- **원인**: RabbitMQ보다 먼저 시작되어 broker 연결 실패
+- **해결**: 자동 복구됨 (`restart: unless-stopped`). 30초 기다린 후 확인
+
+**mysqld-exporter가 DOWN**:
+- **원인**: MySQL보다 먼저 시작됨
+- **해결**: 자동 복구됨. 30초 기다린 후 확인
+
+**django가 DOWN**:
+- **원인**: 앱 시작 실패
+- **해결**:
+  ```bash
+  docker logs speedcam-main
+  ```
+
+### 6.3 cAdvisor 시작 실패 (macOS)
+
+**오류**: `cadvisor: error setting oom score: open /proc/.../oom_score_adj: no such file or directory`
+
+**원인**: cAdvisor는 Linux 전용이며 /proc 파일시스템 필요
+
+**해결**: 예상 동작. macOS에서는 자동으로 제외됨 (`profiles: [linux]`). Linux에서만 실행하세요.
+
+### 6.4 Jaeger에 트레이스가 없음
+
+**오류**: Jaeger UI에서 데이터가 보이지 않음
+
+**원인**: OTEL_EXPORTER_OTLP_ENDPOINT 미설정
+
+**해결**:
+```bash
+# backend.env 확인
+cat docker/backend.env | grep OTEL_EXPORTER_OTLP_ENDPOINT
+
+# 없으면 추가
+echo 'OTEL_EXPORTER_OTLP_ENDPOINT=http://otel-collector:4317' >> docker/backend.env
+
+# 앱 재시작
+docker compose -f docker-compose.yml up -d --force-recreate speedcam-main
+```
+
+### 6.5 Loki 429 Too Many Requests
+
+**오류**: Loki 쿼리 실패 with 429 status
+
+**원인**: 로그 스트림이 너무 많음 (한계 초과)
+
+**해결**:
+```bash
+# loki-config.yml에서 한계 증가
+# docker/monitoring/loki/loki-config.yml 수정
+# limits_config:
+#   max_global_streams_per_user: 20000  # 기본값 10000에서 증가
+```
+
+### 6.6 환경 변수 변경 후 반영 안됨
+
+**오류**: backend.env 변경 후 앱에 반영 안됨
+
+**원인**: Docker restart는 env_file을 다시 읽지 않음
+
+**해결**:
+```bash
+# --force-recreate 사용
+docker compose -f docker-compose.yml up -d --force-recreate speedcam-main
+```
+
+### 6.7 MQTT 연결 실패
+
+**오류**: `python mqtt-load-test.py` 실행 시 연결 실패
+
+**원인**: RabbitMQ MQTT 플러그인 미활성화
+
+**확인**:
+```bash
+docker logs speedcam-rabbitmq | grep -i mqtt
+# "MQTT plugin loaded" 메시지가 있어야 함
+```
+
+**해결** (이미 자동 활성화됨):
+docker-compose.yml의 rabbitmq command에 `rabbitmq_mqtt` 플러그인이 포함되어 있는지 확인
+
+### 6.8 k6 테스트 타임아웃
+
+**오류**: k6 테스트 중 `dial tcp: lookup main: no such host`
+
+**원인**: k6 컨테이너가 speedcam-network에 연결되지 않음
+
+**해결**: docker-compose.monitoring.yml에서 k6 서비스가 올바른 네트워크 설정이 있는지 확인
+
+```yaml
+networks:
+  - speedcam-network  # speedcam-network 참조
+```
+
+---
+
+## 7. 정리 및 종료
+
+### 7.1 전체 종료 및 데이터 제거
+
+```bash
+cd docker
+
+# 컨테이너 + 볼륨 완전 제거
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml down -v
+```
+
+### 7.2 모니터링 데이터만 삭제
+
+런타임 데이터(Prometheus, Grafana, Loki)를 초기화합니다:
+
+```bash
+rm -rf docker/monitoring/prometheus/data \
+       docker/monitoring/loki/data \
+       docker/monitoring/grafana/data
+```
+
+다시 시작하면 초기 상태로 복구됩니다.
+
+### 7.3 모니터링 스택만 종료 (앱 유지)
+
+앱은 계속 실행하고 모니터링만 종료:
+
+```bash
+cd docker
+
+docker compose -f docker-compose.monitoring.yml down
+```
+
+나중에 모니터링을 다시 시작:
+
+```bash
+docker compose -f docker-compose.monitoring.yml up -d
+```
+
+### 7.4 특정 컨테이너만 종료
+
+```bash
+# 개별 서비스 종료
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml \
+  stop speedcam-main speedcam-ocr
+
+# 개별 서비스 재시작
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml \
+  restart speedcam-main
+```
+
+---
+
+## 8. 추가 자료
+
+### 8.1 관련 문서
+
+- [모니터링 스택 가이드](./MONITORING.md): 아키텍처, 설정, 메트릭 상세 설명
+- [배포 가이드](./DEPLOYMENT.md): GCP 멀티 인스턴스 배포 방법
+- [아키텍처 비교](./ARCHITECTURE_COMPARISON.md): 시스템 설계 이유
+
+### 8.2 외부 자료
+
+- [k6 공식 문서](https://k6.io/docs/)
+- [Prometheus PromQL](https://prometheus.io/docs/prometheus/latest/querying/basics/)
+- [Grafana 대시보드](https://grafana.com/grafana/dashboards/)
+- [Jaeger 분산 트레이싱](https://www.jaegertracing.io/docs/)
+- [OpenTelemetry Python](https://opentelemetry.io/docs/instrumentation/python/)
+
+### 8.3 자주 사용하는 명령어
+
+```bash
+# 모니터링 스택 전체 확인
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml ps
+
+# 로그 실시간 추적
+docker logs -f speedcam-main
+docker logs -f speedcam-ocr
+docker logs -f speedcam-alert
+
+# 모니터링 데이터 초기화 후 재시작
+rm -rf docker/monitoring/*/data
+docker compose -f docker-compose.yml -f docker-compose.monitoring.yml up -d
+
+# Prometheus 메트릭 직접 조회
+curl -s http://localhost:9090/api/v1/query --data-urlencode 'query=<PROMQL>'
+
+# RabbitMQ 큐 확인
+curl -s -u sa:1234 http://localhost:15672/api/queues/%2F
+
+# Jaeger 서비스 확인
+curl -s http://localhost:16686/api/services
+
+# Docker 디스크 정리 (주의: 사용하지 않는 모든 이미지/볼륨 제거)
+docker system prune -a --volumes
+```
+
+---
+
+## 9. FAQ
+
+**Q: k6과 MQTT 테스트 중 어느 것을 먼저 실행해야 하나요?**
+
+A: k6을 먼저 실행하세요. k6은 REST API만 테스트하므로 (순수 읽기 작업) 데이터베이스 상태에 영향을 주지 않습니다. MQTT 테스트는 실제 Detection을 생성하므로 나중에 실행하는 것이 좋습니다.
+
+**Q: 부하 테스트 중 시스템이 느려집니다. 어떻게 해야 하나요?**
+
+A: 정상입니다. 먼저 메트릭을 확인하세요:
+1. Prometheus에서 CPU/메모리 사용률 확인
+2. RabbitMQ 큐 깊이 확인 (메시지 밀림)
+3. MySQL 커넥션 풀 상태 확인
+4. 필요하면 docker-compose.yml의 리소스 제한(`resources`) 조정
+
+**Q: 테스트 결과를 저장하고 싶습니다.**
+
+A: k6은 자동으로 Prometheus에 메트릭을 기록합니다. Prometheus → Export로 데이터를 JSON/CSV로 내보낼 수 있습니다. MQTT 테스트의 경우 stdout을 파일로 리다이렉트합니다:
+```bash
+python docker/k6/mqtt-load-test.py ... > test_results.txt
+```
+
+**Q: 모니터링 없이 성능 테스트를 실행할 수 있나요?**
+
+A: 가능합니다. k6 또는 MQTT 테스트 스크립트는 독립적으로 실행할 수 있습니다. 하지만 모니터링 없으면 결과를 측정하고 분석하기 어렵습니다.
+
+**Q: 프로덕션 환경에서 어떻게 테스트하나요?**
+
+A: 이 가이드는 로컬/개발 환경 기준입니다. 프로덕션 배포는 [GCP 멀티 인스턴스 배포 가이드](./DEPLOYMENT.md)를 참고하세요. 프로덕션에서는:
+1. 전용 모니터링 인스턴스 사용
+2. Prometheus 보안 설정 (인증, TLS)
+3. 백그라운드에서 정기적인 스모크 테스트 실행
+4. 알림 규칙(Alert) 설정
+
+---
+
+마지막 업데이트: 2024년 1월
diff --git a/requirements/base.txt b/requirements/base.txt
index 58478e2..0a0c47b 100644
--- a/requirements/base.txt
+++ b/requirements/base.txt
@@ -24,6 +24,14 @@ pytz==2025.1
 drf-yasg==1.21.10
 PyYAML==6.0.2
 
-# APM
-ddtrace
+# Observability (OpenTelemetry)
+opentelemetry-distro
+opentelemetry-exporter-otlp
+opentelemetry-instrumentation-django
+opentelemetry-instrumentation-celery
+opentelemetry-instrumentation-pymysql
+opentelemetry-instrumentation-requests
+opentelemetry-instrumentation-logging
 
+# Prometheus Metrics
+django-prometheus
diff --git a/scripts/start_alert_worker.sh b/scripts/start_alert_worker.sh
index 6c424f7..24bda15 100644
--- a/scripts/start_alert_worker.sh
+++ b/scripts/start_alert_worker.sh
@@ -4,7 +4,9 @@ set -e
 echo "Starting Alert Worker (Celery)..."
 
 # Celery Worker 시작 (gevent pool - I/O 집약적)
-ddtrace-run celery -A config worker \
+opentelemetry-instrument \
+    --service_name speedcam-alert \
+    celery -A config worker \
     --pool=gevent \
     --concurrency=${ALERT_CONCURRENCY:-100} \
     --queues=fcm_queue \
diff --git a/scripts/start_main.sh b/scripts/start_main.sh
index a7a4499..187a61d 100644
--- a/scripts/start_main.sh
+++ b/scripts/start_main.sh
@@ -45,7 +45,9 @@ start_mqtt_subscriber()
 
 # Gunicorn 시작
 echo "Starting Gunicorn..."
-ddtrace-run gunicorn config.wsgi:application \
+opentelemetry-instrument \
+    --service_name speedcam-api \
+    gunicorn config.wsgi:application \
     --bind 0.0.0.0:8000 \
     --workers ${GUNICORN_WORKERS:-4} \
     --threads ${GUNICORN_THREADS:-2} \
diff --git a/scripts/start_ocr_worker.sh b/scripts/start_ocr_worker.sh
index c884272..40557a6 100644
--- a/scripts/start_ocr_worker.sh
+++ b/scripts/start_ocr_worker.sh
@@ -4,7 +4,9 @@ set -e
 echo "Starting OCR Worker (Celery)..."
 
 # Celery Worker 시작 (prefork pool - CPU 집약적)
-ddtrace-run celery -A config worker \
+opentelemetry-instrument \
+    --service_name speedcam-ocr \
+    celery -A config worker \
     --pool=prefork \
     --concurrency=${OCR_CONCURRENCY:-4} \
     --queues=ocr_queue \
diff --git a/terraform/.terraform.lock.hcl b/terraform/.terraform.lock.hcl
deleted file mode 100644
index 8289be8..0000000
--- a/terraform/.terraform.lock.hcl
+++ /dev/null
@@ -1,60 +0,0 @@
-# This file is maintained automatically by "terraform init".
-# Manual edits may be lost in future updates.
-
-provider "registry.terraform.io/hashicorp/google" {
-  version     = "5.45.2"
-  constraints = "~> 5.0"
-  hashes = [
-    "h1:iy2Q9VcnMu4z/bH3v/NmI/nEpgYY7bXgJmT/hVTAUS4=",
-    "zh:0d09c8f20b556305192cdbe0efa6d333ceebba963a8ba91f9f1714b5a20c4b7a",
-    "zh:117143fc91be407874568df416b938a6896f94cb873f26bba279cedab646a804",
-    "zh:16ccf77d18dd2c5ef9c0625f9cf546ebdf3213c0a452f432204c69feed55081e",
-    "zh:3e555cf22a570a4bd247964671f421ed7517970cd9765ceb46f335edc2c6f392",
-    "zh:688bd5b05a75124da7ae6e885b2b92bd29f4261808b2b78bd5f51f525c1052ca",
-    "zh:6db3ef37a05010d82900bfffb3261c59a0c247e0692049cb3eb8c2ef16c9d7bf",
-    "zh:70316fde75f6a15d72749f66d994ccbdde5f5ed4311b6d06b99850f698c9bbf9",
-    "zh:84b8e583771a4f2bd514e519d98ed7fd28dce5efe0634e973170e1cfb5556fb4",
-    "zh:9d4b8ef0a9b6677935c604d94495042e68ff5489932cfd1ec41052e094a279d3",
-    "zh:a2089dd9bd825c107b148dd12d6b286f71aa37dfd4ca9c35157f2dcba7bc19d8",
-    "zh:f03d795c0fd9721e59839255ee7ba7414173017dc530b4ce566daf3802a0d6dd",
-    "zh:f569b65999264a9416862bca5cd2a6177d94ccb0424f3a4ef424428912b9cb3c",
-  ]
-}
-
-provider "registry.terraform.io/hashicorp/null" {
-  version = "3.2.4"
-  hashes = [
-    "h1:L5V05xwp/Gto1leRryuesxjMfgZwjb7oool4WS1UEFQ=",
-    "zh:59f6b52ab4ff35739647f9509ee6d93d7c032985d9f8c6237d1f8a59471bbbe2",
-    "zh:78d5eefdd9e494defcb3c68d282b8f96630502cac21d1ea161f53cfe9bb483b3",
-    "zh:795c897119ff082133150121d39ff26cb5f89a730a2c8c26f3a9c1abf81a9c43",
-    "zh:7b9c7b16f118fbc2b05a983817b8ce2f86df125857966ad356353baf4bff5c0a",
-    "zh:85e33ab43e0e1726e5f97a874b8e24820b6565ff8076523cc2922ba671492991",
-    "zh:9d32ac3619cfc93eb3c4f423492a8e0f79db05fec58e449dee9b2d5873d5f69f",
-    "zh:9e15c3c9dd8e0d1e3731841d44c34571b6c97f5b95e8296a45318b94e5287a6e",
-    "zh:b4c2ab35d1b7696c30b64bf2c0f3a62329107bd1a9121ce70683dec58af19615",
-    "zh:c43723e8cc65bcdf5e0c92581dcbbdcbdcf18b8d2037406a5f2033b1e22de442",
-    "zh:ceb5495d9c31bfb299d246ab333f08c7fb0d67a4f82681fbf47f2a21c3e11ab5",
-    "zh:e171026b3659305c558d9804062762d168f50ba02b88b231d20ec99578a6233f",
-    "zh:ed0fe2acdb61330b01841fa790be00ec6beaac91d41f311fb8254f74eb6a711f",
-  ]
-}
-
-provider "registry.terraform.io/hashicorp/time" {
-  version = "0.13.1"
-  hashes = [
-    "h1:ZT5ppCNIModqk3iOkVt5my8b8yBHmDpl663JtXAIRqM=",
-    "zh:02cb9aab1002f0f2a94a4f85acec8893297dc75915f7404c165983f720a54b74",
-    "zh:04429b2b31a492d19e5ecf999b116d396dac0b24bba0d0fb19ecaefe193fdb8f",
-    "zh:26f8e51bb7c275c404ba6028c1b530312066009194db721a8427a7bc5cdbc83a",
-    "zh:772ff8dbdbef968651ab3ae76d04afd355c32f8a868d03244db3f8496e462690",
-    "zh:78d5eefdd9e494defcb3c68d282b8f96630502cac21d1ea161f53cfe9bb483b3",
-    "zh:898db5d2b6bd6ca5457dccb52eedbc7c5b1a71e4a4658381bcbb38cedbbda328",
-    "zh:8de913bf09a3fa7bedc29fec18c47c571d0c7a3d0644322c46f3aa648cf30cd8",
-    "zh:9402102c86a87bdfe7e501ffbb9c685c32bbcefcfcf897fd7d53df414c36877b",
-    "zh:b18b9bb1726bb8cfbefc0a29cf3657c82578001f514bcf4c079839b6776c47f0",
-    "zh:b9d31fdc4faecb909d7c5ce41d2479dd0536862a963df434be4b16e8e4edc94d",
-    "zh:c951e9f39cca3446c060bd63933ebb89cedde9523904813973fbc3d11863ba75",
-    "zh:e5b773c0d07e962291be0e9b413c7a22c044b8c7b58c76e8aa91d1659990dfb5",
-  ]
-}
diff --git a/terraform/README.md b/terraform/README.md
deleted file mode 100644
index 99ccc8e..0000000
--- a/terraform/README.md
+++ /dev/null
@@ -1,385 +0,0 @@
-# Speedcam MSA - Terraform 배포 가이드
-
-GCP(Google Cloud Platform)에 Speedcam MSA 인프라를 자동으로 배포하기 위한 Terraform 구성입니다.
-
-## 아키텍처
-
-```
-┌─────────────────────────────────────────────────────────────────┐
-│                    Google Cloud Platform                         │
-│                                                                  │
-│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐              │
-│  │    Main     │  │     OCR     │  │    Alert    │              │
-│  │  (Django)   │  │   (Celery)  │  │   (Celery)  │              │
-│  │  e2-medium  │  │  e2-medium  │  │   e2-small  │              │
-│  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘              │
-│         │                │                │                      │
-│         └────────────────┼────────────────┘                      │
-│                          │                                       │
-│              ┌───────────┴───────────┐                          │
-│              │                       │                          │
-│       ┌──────┴──────┐        ┌──────┴──────┐                    │
-│       │  RabbitMQ   │        │    MySQL    │                    │
-│       │  e2-small   │        │   e2-small  │                    │
-│       │  MQTT/AMQP  │        │   4 DBs     │                    │
-│       └─────────────┘        └─────────────┘                    │
-│                                                                  │
-│  ┌─────────────────────────────────────────────────────────────┐ │
-│  │                   Artifact Registry                         │ │
-│  │              speedcam/{main,ocr,alert}                      │ │
-│  └─────────────────────────────────────────────────────────────┘ │
-└─────────────────────────────────────────────────────────────────┘
-```
-
-## 파일 구조
-
-```
-terraform/
-├── main.tf                 # Provider 설정 및 로컬 변수
-├── variables.tf            # 입력 변수 정의
-├── network.tf              # 방화벽 규칙
-├── artifact_registry.tf    # 컨테이너 레지스트리
-├── instances_infra.tf      # RabbitMQ, MySQL 인스턴스
-├── instances_services.tf   # Main, OCR, Alert 인스턴스
-├── outputs.tf              # 출력 변수
-├── terraform.tfvars.example # 변수 예제 파일
-└── README.md               # 이 문서
-```
-
-## 사전 요구사항
-
-1. **Terraform 설치** (v1.0 이상)
-   ```bash
-   # macOS
-   brew install terraform
-
-   # Linux
-   wget https://releases.hashicorp.com/terraform/1.7.0/terraform_1.7.0_linux_amd64.zip
-   unzip terraform_1.7.0_linux_amd64.zip
-   sudo mv terraform /usr/local/bin/
-   ```
-
-2. **Google Cloud SDK 설치 및 인증**
-   ```bash
-   # 인증
-   gcloud auth login
-   gcloud auth application-default login
-
-   # 프로젝트 설정
-   gcloud config set project YOUR_PROJECT_ID
-   ```
-
-3. **필요한 API 활성화**
-   ```bash
-   gcloud services enable compute.googleapis.com
-   gcloud services enable artifactregistry.googleapis.com
-   ```
-
-4. **Docker 이미지 빌드 및 푸시**
-   ```bash
-   # 프로젝트 루트에서
-   make build
-   make push
-   ```
-
-## 빠른 시작
-
-### 1. 변수 파일 생성
-
-```bash
-cd terraform
-cp terraform.tfvars.example terraform.tfvars
-```
-
-### 2. 변수 파일 수정
-
-```hcl
-# terraform.tfvars
-project_id = "your-actual-project-id"
-
-# 보안을 위해 강력한 비밀번호 설정
-db_password      = "your-secure-db-password"
-db_root_password = "your-secure-root-password"
-rabbitmq_password = "your-secure-rabbitmq-password"
-
-# 환경 설정 (dev, staging, prod)
-environment = "dev"
-```
-
-### 3. Terraform 초기화
-
-```bash
-terraform init
-```
-
-### 4. 배포 계획 확인
-
-```bash
-terraform plan
-```
-
-### 5. 인프라 배포
-
-```bash
-terraform apply
-```
-
-## 변수 설명
-
-### 필수 변수
-
-| 변수 | 설명 | 예시 |
-|------|------|------|
-| `project_id` | GCP 프로젝트 ID | `my-project-123` |
-
-### 선택 변수 (기본값 제공)
-
-| 변수 | 기본값 | 설명 |
-|------|--------|------|
-| `region` | `asia-northeast3` | GCP 리전 (서울) |
-| `zone` | `asia-northeast3-a` | GCP 존 |
-| `environment` | `dev` | 환경 (dev/staging/prod) |
-| `db_name` | `speedcam` | 기본 데이터베이스 이름 |
-| `db_user` | `sa` | 데이터베이스 사용자 |
-| `db_password` | `sa` | 데이터베이스 비밀번호 |
-| `rabbitmq_user` | `sa` | RabbitMQ 사용자 |
-| `rabbitmq_password` | `sa` | RabbitMQ 비밀번호 |
-| `machine_type_small` | `e2-small` | 작은 인스턴스 타입 |
-| `machine_type_medium` | `e2-medium` | 중간 인스턴스 타입 |
-| `ocr_concurrency` | `2` | OCR 워커 동시성 |
-| `alert_concurrency` | `50` | Alert 워커 동시성 |
-| `ocr_mock` | `true` | OCR 모킹 여부 |
-| `fcm_mock` | `true` | FCM 모킹 여부 |
-
-## 출력 값
-
-배포 완료 후 다음 정보를 확인할 수 있습니다:
-
-```bash
-# 모든 출력 확인
-terraform output
-
-# 특정 출력 확인
-terraform output api_url
-terraform output swagger_url
-terraform output deployment_summary
-```
-
-### 주요 출력
-
-- `api_url` - API 기본 URL
-- `swagger_url` - Swagger UI URL
-- `health_url` - 헬스 체크 URL
-- `rabbitmq_management_url` - RabbitMQ 관리 UI URL
-- `registry_url` - Artifact Registry URL
-- `deployment_summary` - 전체 배포 요약
-
-## 환경별 배포
-
-### 개발 환경
-
-```hcl
-# terraform.tfvars
-environment = "dev"
-ocr_mock    = true
-fcm_mock    = true
-machine_type_small  = "e2-small"
-machine_type_medium = "e2-medium"
-```
-
-### 스테이징 환경
-
-```hcl
-# terraform.tfvars
-environment = "staging"
-ocr_mock    = false
-fcm_mock    = true
-machine_type_small  = "e2-small"
-machine_type_medium = "e2-medium"
-```
-
-### 프로덕션 환경
-
-```hcl
-# terraform.tfvars
-environment = "prod"
-ocr_mock    = false
-fcm_mock    = false
-machine_type_small  = "e2-medium"
-machine_type_medium = "e2-standard-2"
-ocr_concurrency   = 4
-alert_concurrency = 100
-```
-
-## Workspace 활용
-
-여러 환경을 관리하려면 Terraform Workspace를 사용할 수 있습니다:
-
-```bash
-# Workspace 생성
-terraform workspace new dev
-terraform workspace new staging
-terraform workspace new prod
-
-# Workspace 전환
-terraform workspace select dev
-
-# 현재 Workspace 확인
-terraform workspace show
-
-# Workspace 목록
-terraform workspace list
-```
-
-## 상태 관리
-
-### 로컬 상태 (기본)
-
-기본적으로 상태 파일은 로컬에 저장됩니다:
-- `terraform.tfstate`
-- `terraform.tfstate.backup`
-
-### 원격 상태 (권장)
-
-팀 협업을 위해 GCS 백엔드 사용을 권장합니다:
-
-```hcl
-# main.tf에 추가
-terraform {
-  backend "gcs" {
-    bucket = "your-terraform-state-bucket"
-    prefix = "speedcam/terraform/state"
-  }
-}
-```
-
-백엔드 설정 후:
-```bash
-terraform init -migrate-state
-```
-
-## 인프라 업데이트
-
-### 이미지 태그 변경
-
-```bash
-terraform apply -var="image_tag=v1.2.0"
-```
-
-### 인스턴스 타입 변경
-
-```bash
-terraform apply -var="machine_type_medium=e2-standard-2"
-```
-
-### 특정 리소스만 재생성
-
-```bash
-# Main 서비스만 재생성
-terraform taint google_compute_instance.main
-terraform apply
-
-# 마이그레이션만 재실행
-terraform taint null_resource.run_migrations
-terraform apply
-```
-
-## 인프라 삭제
-
-```bash
-# 전체 삭제 (확인 필요)
-terraform destroy
-
-# 자동 승인으로 삭제
-terraform destroy -auto-approve
-```
-
-## 문제 해결
-
-### 1. 인스턴스 시작 실패
-
-```bash
-# 인스턴스 로그 확인
-gcloud compute instances get-serial-port-output speedcam-main --zone=asia-northeast3-a
-
-# 컨테이너 로그 확인
-gcloud compute ssh speedcam-main --zone=asia-northeast3-a \
-  --command="docker logs \$(docker ps -q)"
-```
-
-### 2. 데이터베이스 연결 실패
-
-```bash
-# MySQL 상태 확인
-gcloud compute ssh speedcam-mysql --zone=asia-northeast3-a \
-  --command="docker exec \$(docker ps -q) mysqladmin -u root -p status"
-```
-
-### 3. RabbitMQ 연결 실패
-
-```bash
-# RabbitMQ 상태 확인
-gcloud compute ssh speedcam-rabbitmq --zone=asia-northeast3-a \
-  --command="docker exec \$(docker ps -q) rabbitmqctl status"
-```
-
-### 4. Terraform 상태 문제
-
-```bash
-# 상태 새로고침
-terraform refresh
-
-# 상태에서 리소스 제거 (실제 리소스는 유지)
-terraform state rm google_compute_instance.main
-
-# 기존 리소스 가져오기
-terraform import google_compute_instance.main speedcam-main
-```
-
-## 비용 최적화
-
-### 예상 월간 비용 (asia-northeast3 기준)
-
-| 리소스 | 타입 | 예상 비용 |
-|--------|------|----------|
-| Main | e2-medium | ~$25 |
-| OCR | e2-medium | ~$25 |
-| Alert | e2-small | ~$13 |
-| RabbitMQ | e2-small | ~$13 |
-| MySQL | e2-small + SSD | ~$18 |
-| **총계** | | **~$94/월** |
-
-### 비용 절감 팁
-
-1. **Preemptible VM 사용** (개발 환경)
-   ```hcl
-   scheduling {
-     preemptible = true
-   }
-   ```
-
-2. **자동 시작/중지 스케줄링**
-   - Cloud Scheduler로 업무 시간 외 인스턴스 중지
-
-3. **Committed Use Discounts**
-   - 1년/3년 약정으로 최대 57% 할인
-
-## 보안 고려사항
-
-1. **비밀번호 관리**
-   - Secret Manager 사용 권장
-   - terraform.tfvars를 .gitignore에 추가
-
-2. **네트워크 보안**
-   - 프로덕션에서는 외부 IP 제거
-   - VPN 또는 IAP 터널 사용
-
-3. **서비스 계정**
-   - 최소 권한 원칙 적용
-   - 전용 서비스 계정 생성
-
-## 참고 자료
-
-- [Terraform GCP Provider 문서](https://registry.terraform.io/providers/hashicorp/google/latest/docs)
-- [GCP Container-Optimized OS](https://cloud.google.com/container-optimized-os/docs)
-- [GCP 가격 계산기](https://cloud.google.com/products/calculator)
diff --git a/terraform/artifact_registry.tf b/terraform/artifact_registry.tf
deleted file mode 100644
index bef100a..0000000
--- a/terraform/artifact_registry.tf
+++ /dev/null
@@ -1,50 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Artifact Registry Configuration
-# =============================================================================
-
-# =============================================================================
-# Artifact Registry Repository
-# =============================================================================
-
-resource "google_artifact_registry_repository" "speedcam" {
-  location      = var.region
-  repository_id = var.registry_name
-  description   = "Speedcam MSA Docker images"
-  format        = "DOCKER"
-  project       = var.project_id
-
-  labels = local.common_labels
-
-  # Cleanup policy (optional)
-  cleanup_policies {
-    id     = "delete-old-images"
-    action = "DELETE"
-
-    condition {
-      tag_state  = "UNTAGGED"
-      older_than = "2592000s" # 30 days
-    }
-  }
-
-  cleanup_policies {
-    id     = "keep-recent-tagged"
-    action = "KEEP"
-
-    most_recent_versions {
-      keep_count = 10
-      package_name_prefixes = ["main", "ocr", "alert"]
-    }
-  }
-}
-
-# =============================================================================
-# IAM Policy for Compute Engine to pull images
-# =============================================================================
-
-resource "google_artifact_registry_repository_iam_member" "compute_reader" {
-  project    = var.project_id
-  location   = var.region
-  repository = google_artifact_registry_repository.speedcam.name
-  role       = "roles/artifactregistry.reader"
-  member     = "serviceAccount:${data.google_compute_default_service_account.default.email}"
-}
diff --git a/terraform/instances_infra.tf b/terraform/instances_infra.tf
deleted file mode 100644
index 1aeb872..0000000
--- a/terraform/instances_infra.tf
+++ /dev/null
@@ -1,204 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Infrastructure Instances (RabbitMQ, MySQL)
-# =============================================================================
-
-# =============================================================================
-# RabbitMQ Instance
-# =============================================================================
-
-resource "google_compute_instance" "rabbitmq" {
-  name         = "speedcam-rabbitmq"
-  machine_type = var.machine_type_small
-  zone         = var.zone
-  project      = var.project_id
-
-  tags = ["speedcam", "speedcam-web"]
-
-  labels = merge(local.common_labels, {
-    service = "rabbitmq"
-  })
-
-  boot_disk {
-    initialize_params {
-      image = "cos-cloud/cos-stable"
-      size  = 20
-      type  = "pd-standard"
-    }
-  }
-
-  network_interface {
-    network = var.network_name
-
-    access_config {
-      // Ephemeral public IP
-    }
-  }
-
-  metadata = {
-    gce-container-declaration = yamlencode({
-      spec = {
-        containers = [{
-          name  = "rabbitmq"
-          image = var.rabbitmq_image
-          env = [
-            { name = "RABBITMQ_DEFAULT_USER", value = var.rabbitmq_user },
-            { name = "RABBITMQ_DEFAULT_PASS", value = var.rabbitmq_password },
-          ]
-          volumeMounts = []
-        }]
-        volumes = []
-        restartPolicy = "Always"
-      }
-    })
-  }
-
-  service_account {
-    email  = data.google_compute_default_service_account.default.email
-    scopes = ["cloud-platform"]
-  }
-
-  scheduling {
-    automatic_restart   = true
-    on_host_maintenance = "MIGRATE"
-    preemptible         = false
-  }
-
-  allow_stopping_for_update = true
-
-  lifecycle {
-    create_before_destroy = true
-  }
-}
-
-# =============================================================================
-# MySQL Instance
-# =============================================================================
-
-resource "google_compute_instance" "mysql" {
-  name         = "speedcam-mysql"
-  machine_type = var.machine_type_small
-  zone         = var.zone
-  project      = var.project_id
-
-  tags = ["speedcam"]
-
-  labels = merge(local.common_labels, {
-    service = "mysql"
-  })
-
-  boot_disk {
-    initialize_params {
-      image = "cos-cloud/cos-stable"
-      size  = 50
-      type  = "pd-ssd"
-    }
-  }
-
-  network_interface {
-    network = var.network_name
-
-    access_config {
-      // Ephemeral public IP
-    }
-  }
-
-  metadata = {
-    gce-container-declaration = yamlencode({
-      spec = {
-        containers = [{
-          name  = "mysql"
-          image = var.mysql_image
-          env = [
-            { name = "MYSQL_ROOT_PASSWORD", value = var.db_root_password },
-            { name = "MYSQL_DATABASE", value = var.db_name },
-            { name = "MYSQL_USER", value = var.db_user },
-            { name = "MYSQL_PASSWORD", value = var.db_password },
-          ]
-          volumeMounts = [{
-            name      = "mysql-data"
-            mountPath = "/var/lib/mysql"
-          }]
-        }]
-        volumes = [{
-          name = "mysql-data"
-          hostPath = {
-            path = "/var/lib/mysql"
-          }
-        }]
-        restartPolicy = "Always"
-      }
-    })
-  }
-
-  service_account {
-    email  = data.google_compute_default_service_account.default.email
-    scopes = ["cloud-platform"]
-  }
-
-  scheduling {
-    automatic_restart   = true
-    on_host_maintenance = "MIGRATE"
-    preemptible         = false
-  }
-
-  allow_stopping_for_update = true
-
-  lifecycle {
-    create_before_destroy = true
-  }
-}
-
-# =============================================================================
-# Null Resource for Infrastructure Initialization
-# =============================================================================
-
-# Wait for instances to be ready
-resource "time_sleep" "wait_for_infra" {
-  depends_on = [
-    google_compute_instance.rabbitmq,
-    google_compute_instance.mysql
-  ]
-
-  create_duration = "90s"
-}
-
-# Initialize RabbitMQ MQTT plugin
-resource "null_resource" "init_rabbitmq" {
-  depends_on = [time_sleep.wait_for_infra]
-
-  provisioner "local-exec" {
-    command = <<-EOT
-      gcloud compute ssh speedcam-rabbitmq --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) rabbitmq-plugins enable rabbitmq_mqtt" \
-        || echo "MQTT plugin may already be enabled"
-    EOT
-  }
-
-  triggers = {
-    instance_id = google_compute_instance.rabbitmq.instance_id
-  }
-}
-
-# Initialize MySQL databases
-resource "null_resource" "init_mysql" {
-  depends_on = [time_sleep.wait_for_infra]
-
-  provisioner "local-exec" {
-    command = <<-EOT
-      gcloud compute ssh speedcam-mysql --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) mysql -u root -p${var.db_root_password} -e \"\
-          CREATE DATABASE IF NOT EXISTS ${var.db_name}_vehicles; \
-          CREATE DATABASE IF NOT EXISTS ${var.db_name}_detections; \
-          CREATE DATABASE IF NOT EXISTS ${var.db_name}_notifications; \
-          GRANT ALL PRIVILEGES ON ${var.db_name}_vehicles.* TO '${var.db_user}'@'%'; \
-          GRANT ALL PRIVILEGES ON ${var.db_name}_detections.* TO '${var.db_user}'@'%'; \
-          GRANT ALL PRIVILEGES ON ${var.db_name}_notifications.* TO '${var.db_user}'@'%'; \
-          FLUSH PRIVILEGES;\"" \
-        || echo "Databases may already exist"
-    EOT
-  }
-
-  triggers = {
-    instance_id = google_compute_instance.mysql.instance_id
-  }
-}
diff --git a/terraform/instances_monitoring.tf b/terraform/instances_monitoring.tf
deleted file mode 100644
index f3988b1..0000000
--- a/terraform/instances_monitoring.tf
+++ /dev/null
@@ -1,109 +0,0 @@
-# =============================================================================
-# DataDog Agent Instance
-# =============================================================================
-
-resource "google_compute_instance" "datadog_agent" {
-  name         = "speedcam-datadog"
-  machine_type = var.machine_type_small
-  zone         = var.zone
-
-  depends_on = [
-    google_compute_instance.rabbitmq,
-    google_compute_instance.mysql
-  ]
-
-  tags = ["speedcam"]
-
-  boot_disk {
-    initialize_params {
-      image = "cos-cloud/cos-stable"
-      size  = 20
-    }
-  }
-
-  network_interface {
-    network = "default"
-    access_config {}
-  }
-
-  metadata = {
-    # cloud-init: 컨테이너 시작 전에 Integration 설정 파일 생성
-    user-data = <<-CLOUDINIT
-      #cloud-config
-      write_files:
-        - path: /tmp/dd-confd/mysql.d/conf.yaml
-          permissions: '0644'
-          content: |
-            init_config:
-            instances:
-              - host: ${google_compute_instance.mysql.network_interface[0].network_ip}
-                port: 3306
-                username: ${var.db_user}
-                password: ${var.db_password}
-                reported_hostname: speedcam-mysql
-                tags:
-                  - env:${var.environment}
-                  - service:speedcam-mysql
-        - path: /tmp/dd-confd/rabbitmq.d/conf.yaml
-          permissions: '0644'
-          content: |
-            init_config:
-            instances:
-              - rabbitmq_api_url: http://${google_compute_instance.rabbitmq.network_interface[0].network_ip}:15672/api/
-                rabbitmq_user: ${var.rabbitmq_user}
-                rabbitmq_pass: ${var.rabbitmq_password}
-                tag_families: true
-                collect_node_metrics: true
-                reported_hostname: speedcam-rabbitmq
-                tags:
-                  - env:${var.environment}
-                  - service:speedcam-rabbitmq
-    CLOUDINIT
-
-    gce-container-declaration = yamlencode({
-      spec = {
-        containers = [{
-          name  = "datadog-agent"
-          image = "gcr.io/datadoghq/agent:7"
-          env = [
-            { name = "DD_API_KEY", value = var.dd_api_key },
-            { name = "DD_SITE", value = var.dd_site },
-            { name = "DD_APM_ENABLED", value = "true" },
-            { name = "DD_APM_NON_LOCAL_TRAFFIC", value = "true" },
-            { name = "DD_DOGSTATSD_NON_LOCAL_TRAFFIC", value = "true" },
-            { name = "DD_LOGS_ENABLED", value = "true" },
-            { name = "DD_ENV", value = var.environment },
-          ]
-          volumeMounts = [
-            { name = "mysql-confd", mountPath = "/etc/datadog-agent/conf.d/mysql.d", readOnly = true },
-            { name = "rabbitmq-confd", mountPath = "/etc/datadog-agent/conf.d/rabbitmq.d", readOnly = true },
-          ]
-        }]
-        volumes = [
-          { name = "mysql-confd", hostPath = { path = "/tmp/dd-confd/mysql.d" } },
-          { name = "rabbitmq-confd", hostPath = { path = "/tmp/dd-confd/rabbitmq.d" } },
-        ]
-        restartPolicy = "Always"
-      }
-    })
-  }
-
-  labels = {
-    project     = "speedcam"
-    environment = var.environment
-    service     = "datadog"
-    managed_by  = "terraform"
-  }
-
-  scheduling {
-    automatic_restart   = true
-    on_host_maintenance = "MIGRATE"
-    preemptible         = false
-  }
-
-  service_account {
-    scopes = ["cloud-platform"]
-  }
-
-  allow_stopping_for_update = true
-}
diff --git a/terraform/instances_services.tf b/terraform/instances_services.tf
deleted file mode 100644
index c6f2cba..0000000
--- a/terraform/instances_services.tf
+++ /dev/null
@@ -1,259 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Service Instances (Main, OCR, Alert)
-# =============================================================================
-
-# =============================================================================
-# Main Service Instance
-# =============================================================================
-
-resource "google_compute_instance" "main" {
-  name         = "speedcam-main"
-  machine_type = var.machine_type_medium
-  zone         = var.zone
-  project      = var.project_id
-
-  depends_on = [
-    null_resource.init_rabbitmq,
-    null_resource.init_mysql,
-    google_compute_instance.datadog_agent
-  ]
-
-  tags = ["speedcam", "speedcam-web"]
-
-  labels = merge(local.common_labels, {
-    service = "main"
-  })
-
-  boot_disk {
-    initialize_params {
-      image = "cos-cloud/cos-stable"
-      size  = 20
-      type  = "pd-standard"
-    }
-  }
-
-  network_interface {
-    network = var.network_name
-
-    access_config {
-      // Ephemeral public IP
-    }
-  }
-
-  metadata = {
-    gce-container-declaration = yamlencode({
-      spec = {
-        containers = [{
-          name  = "main"
-          image = "${local.registry}/main:${var.image_tag}"
-          env = concat([
-            for k, v in local.common_env : { name = k, value = v }
-          ], [
-            { name = "DD_SERVICE", value = "speedcam-api" },
-            { name = "RABBITMQ_HOST", value = google_compute_instance.rabbitmq.network_interface[0].network_ip },
-            { name = "MQTT_PORT", value = "1883" },
-            { name = "MQTT_USER", value = var.rabbitmq_user },
-            { name = "MQTT_PASS", value = var.rabbitmq_password },
-            { name = "OCR_MOCK", value = tostring(var.ocr_mock) },
-            { name = "FCM_MOCK", value = tostring(var.fcm_mock) },
-          ])
-        }]
-        restartPolicy = "Always"
-      }
-    })
-  }
-
-  service_account {
-    email  = data.google_compute_default_service_account.default.email
-    scopes = ["cloud-platform"]
-  }
-
-  scheduling {
-    automatic_restart   = true
-    on_host_maintenance = "MIGRATE"
-    preemptible         = false
-  }
-
-  allow_stopping_for_update = true
-}
-
-# =============================================================================
-# OCR Worker Instance
-# =============================================================================
-
-resource "google_compute_instance" "ocr" {
-  name         = "speedcam-ocr"
-  machine_type = var.machine_type_medium
-  zone         = var.zone
-  project      = var.project_id
-
-  depends_on = [
-    null_resource.init_rabbitmq,
-    null_resource.init_mysql,
-    google_compute_instance.datadog_agent
-  ]
-
-  tags = ["speedcam"]
-
-  labels = merge(local.common_labels, {
-    service = "ocr"
-  })
-
-  boot_disk {
-    initialize_params {
-      image = "cos-cloud/cos-stable"
-      size  = 30
-      type  = "pd-standard"
-    }
-  }
-
-  network_interface {
-    network = var.network_name
-
-    access_config {
-      // Ephemeral public IP
-    }
-  }
-
-  metadata = {
-    gce-container-declaration = yamlencode({
-      spec = {
-        containers = [{
-          name  = "ocr"
-          image = "${local.registry}/ocr:${var.image_tag}"
-          env = concat([
-            for k, v in local.common_env : { name = k, value = v }
-          ], [
-            { name = "DD_SERVICE", value = "speedcam-ocr" },
-            { name = "OCR_CONCURRENCY", value = tostring(var.ocr_concurrency) },
-            { name = "OCR_MOCK", value = tostring(var.ocr_mock) },
-          ])
-        }]
-        restartPolicy = "Always"
-      }
-    })
-  }
-
-  service_account {
-    email  = data.google_compute_default_service_account.default.email
-    scopes = ["cloud-platform"]
-  }
-
-  scheduling {
-    automatic_restart   = true
-    on_host_maintenance = "MIGRATE"
-    preemptible         = false
-  }
-
-  allow_stopping_for_update = true
-}
-
-# =============================================================================
-# Alert Worker Instance
-# =============================================================================
-
-resource "google_compute_instance" "alert" {
-  name         = "speedcam-alert"
-  machine_type = var.machine_type_small
-  zone         = var.zone
-  project      = var.project_id
-
-  depends_on = [
-    null_resource.init_rabbitmq,
-    null_resource.init_mysql,
-    google_compute_instance.datadog_agent
-  ]
-
-  tags = ["speedcam"]
-
-  labels = merge(local.common_labels, {
-    service = "alert"
-  })
-
-  boot_disk {
-    initialize_params {
-      image = "cos-cloud/cos-stable"
-      size  = 20
-      type  = "pd-standard"
-    }
-  }
-
-  network_interface {
-    network = var.network_name
-
-    access_config {
-      // Ephemeral public IP
-    }
-  }
-
-  metadata = {
-    gce-container-declaration = yamlencode({
-      spec = {
-        containers = [{
-          name  = "alert"
-          image = "${local.registry}/alert:${var.image_tag}"
-          env = concat([
-            for k, v in local.common_env : { name = k, value = v }
-          ], [
-            { name = "DD_SERVICE", value = "speedcam-alert" },
-            { name = "ALERT_CONCURRENCY", value = tostring(var.alert_concurrency) },
-            { name = "FCM_MOCK", value = tostring(var.fcm_mock) },
-          ])
-        }]
-        restartPolicy = "Always"
-      }
-    })
-  }
-
-  service_account {
-    email  = data.google_compute_default_service_account.default.email
-    scopes = ["cloud-platform"]
-  }
-
-  scheduling {
-    automatic_restart   = true
-    on_host_maintenance = "MIGRATE"
-    preemptible         = false
-  }
-
-  allow_stopping_for_update = true
-}
-
-# =============================================================================
-# Django Migrations
-# =============================================================================
-
-resource "time_sleep" "wait_for_main" {
-  depends_on = [google_compute_instance.main]
-
-  create_duration = "60s"
-}
-
-resource "null_resource" "run_migrations" {
-  depends_on = [time_sleep.wait_for_main]
-
-  provisioner "local-exec" {
-    command = <<-EOT
-      # Create migrations
-      gcloud compute ssh speedcam-main --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) python manage.py makemigrations vehicles detections notifications 2>/dev/null || true"
-
-      # Run migrations
-      gcloud compute ssh speedcam-main --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) python manage.py migrate --database=default --noinput"
-
-      gcloud compute ssh speedcam-main --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) python manage.py migrate vehicles --database=vehicles_db --noinput"
-
-      gcloud compute ssh speedcam-main --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) python manage.py migrate detections --database=detections_db --noinput"
-
-      gcloud compute ssh speedcam-main --zone=${var.zone} --project=${var.project_id} \
-        --command="docker exec \$(docker ps -q) python manage.py migrate notifications --database=notifications_db --noinput"
-    EOT
-  }
-
-  triggers = {
-    main_instance_id = google_compute_instance.main.instance_id
-  }
-}
diff --git a/terraform/main.tf b/terraform/main.tf
deleted file mode 100644
index d527cee..0000000
--- a/terraform/main.tf
+++ /dev/null
@@ -1,79 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Terraform Main Configuration
-# =============================================================================
-
-terraform {
-  required_version = ">= 1.5.0"
-
-  required_providers {
-    google = {
-      source  = "hashicorp/google"
-      version = "~> 5.0"
-    }
-  }
-
-  # Optional: Configure backend for state management
-  # backend "gcs" {
-  #   bucket = "your-terraform-state-bucket"
-  #   prefix = "speedcam/state"
-  # }
-}
-
-# =============================================================================
-# Provider Configuration
-# =============================================================================
-
-provider "google" {
-  project = var.project_id
-  region  = var.region
-  zone    = var.zone
-}
-
-# =============================================================================
-# Data Sources
-# =============================================================================
-
-data "google_project" "current" {
-  project_id = var.project_id
-}
-
-data "google_compute_default_service_account" "default" {
-  project = var.project_id
-}
-
-# =============================================================================
-# Local Values
-# =============================================================================
-
-locals {
-  # Common labels for all resources
-  common_labels = {
-    project     = "speedcam"
-    environment = var.environment
-    managed_by  = "terraform"
-  }
-
-  # Container registry path
-  registry = "${var.region}-docker.pkg.dev/${var.project_id}/${var.registry_name}"
-
-  # Service environment variables (common)
-  common_env = {
-    DJANGO_SETTINGS_MODULE   = "config.settings.${var.environment}"
-    DB_HOST                  = google_compute_instance.mysql.network_interface[0].network_ip
-    DB_PORT                  = "3306"
-    DB_NAME                  = var.db_name
-    DB_NAME_VEHICLES         = "${var.db_name}_vehicles"
-    DB_NAME_DETECTIONS       = "${var.db_name}_detections"
-    DB_NAME_NOTIFICATIONS    = "${var.db_name}_notifications"
-    DB_USER                  = var.db_user
-    DB_PASSWORD              = var.db_password
-    CELERY_BROKER_URL        = "amqp://${var.rabbitmq_user}:${var.rabbitmq_password}@${google_compute_instance.rabbitmq.network_interface[0].network_ip}:5672//"
-    DD_AGENT_HOST            = google_compute_instance.datadog_agent.network_interface[0].network_ip
-    DD_TRACE_AGENT_PORT      = "8126"
-    DD_ENV                   = var.environment
-    DD_LOGS_INJECTION        = "true"
-    DD_TRACE_SAMPLE_RATE     = "1"
-    DD_PROFILING_ENABLED     = "true"
-    _DD_TRACE_WRITER_NATIVE  = "false"
-  }
-}
diff --git a/terraform/network.tf b/terraform/network.tf
deleted file mode 100644
index 102ccbb..0000000
--- a/terraform/network.tf
+++ /dev/null
@@ -1,69 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Network Configuration
-# =============================================================================
-
-# =============================================================================
-# Firewall Rules
-# =============================================================================
-
-# Internal communication between services
-resource "google_compute_firewall" "speedcam_internal" {
-  name    = "speedcam-internal"
-  network = var.network_name
-  project = var.project_id
-
-  description = "Allow internal communication for Speedcam MSA"
-
-  allow {
-    protocol = "tcp"
-    ports    = ["3306", "5672", "1883", "15672", "8000", "8126"]
-  }
-
-  allow {
-    protocol = "udp"
-    ports    = ["8125"]
-  }
-
-  source_ranges = ["10.0.0.0/8"]
-  target_tags   = ["speedcam"]
-
-  priority = 1000
-}
-
-# External access for API and RabbitMQ Management
-resource "google_compute_firewall" "speedcam_external" {
-  name    = "speedcam-external"
-  network = var.network_name
-  project = var.project_id
-
-  description = "Allow external access for Speedcam API and RabbitMQ Management"
-
-  allow {
-    protocol = "tcp"
-    ports    = ["8000", "15672"]
-  }
-
-  source_ranges = ["0.0.0.0/0"]
-  target_tags   = ["speedcam-web"]
-
-  priority = 1000
-}
-
-# SSH access (optional - for debugging)
-resource "google_compute_firewall" "speedcam_ssh" {
-  name    = "speedcam-ssh"
-  network = var.network_name
-  project = var.project_id
-
-  description = "Allow SSH access to Speedcam instances"
-
-  allow {
-    protocol = "tcp"
-    ports    = ["22"]
-  }
-
-  source_ranges = ["0.0.0.0/0"]
-  target_tags   = ["speedcam"]
-
-  priority = 1000
-}
diff --git a/terraform/outputs.tf b/terraform/outputs.tf
deleted file mode 100644
index 8c1f65c..0000000
--- a/terraform/outputs.tf
+++ /dev/null
@@ -1,160 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Terraform Outputs
-# =============================================================================
-
-# =============================================================================
-# Instance IPs
-# =============================================================================
-
-output "rabbitmq_internal_ip" {
-  description = "RabbitMQ internal IP"
-  value       = google_compute_instance.rabbitmq.network_interface[0].network_ip
-}
-
-output "rabbitmq_external_ip" {
-  description = "RabbitMQ external IP"
-  value       = google_compute_instance.rabbitmq.network_interface[0].access_config[0].nat_ip
-}
-
-output "mysql_internal_ip" {
-  description = "MySQL internal IP"
-  value       = google_compute_instance.mysql.network_interface[0].network_ip
-}
-
-output "mysql_external_ip" {
-  description = "MySQL external IP"
-  value       = google_compute_instance.mysql.network_interface[0].access_config[0].nat_ip
-}
-
-output "main_internal_ip" {
-  description = "Main service internal IP"
-  value       = google_compute_instance.main.network_interface[0].network_ip
-}
-
-output "main_external_ip" {
-  description = "Main service external IP"
-  value       = google_compute_instance.main.network_interface[0].access_config[0].nat_ip
-}
-
-output "ocr_internal_ip" {
-  description = "OCR worker internal IP"
-  value       = google_compute_instance.ocr.network_interface[0].network_ip
-}
-
-output "ocr_external_ip" {
-  description = "OCR worker external IP"
-  value       = google_compute_instance.ocr.network_interface[0].access_config[0].nat_ip
-}
-
-output "alert_internal_ip" {
-  description = "Alert worker internal IP"
-  value       = google_compute_instance.alert.network_interface[0].network_ip
-}
-
-output "alert_external_ip" {
-  description = "Alert worker external IP"
-  value       = google_compute_instance.alert.network_interface[0].access_config[0].nat_ip
-}
-
-output "datadog_internal_ip" {
-  description = "DataDog Agent internal IP"
-  value       = google_compute_instance.datadog_agent.network_interface[0].network_ip
-}
-
-output "datadog_external_ip" {
-  description = "DataDog Agent external IP"
-  value       = google_compute_instance.datadog_agent.network_interface[0].access_config[0].nat_ip
-}
-
-# =============================================================================
-# Service URLs
-# =============================================================================
-
-output "api_url" {
-  description = "API base URL"
-  value       = "http://${google_compute_instance.main.network_interface[0].access_config[0].nat_ip}:8000"
-}
-
-output "swagger_url" {
-  description = "Swagger UI URL"
-  value       = "http://${google_compute_instance.main.network_interface[0].access_config[0].nat_ip}:8000/swagger/"
-}
-
-output "health_url" {
-  description = "Health check URL"
-  value       = "http://${google_compute_instance.main.network_interface[0].access_config[0].nat_ip}:8000/health/"
-}
-
-output "rabbitmq_management_url" {
-  description = "RabbitMQ Management UI URL"
-  value       = "http://${google_compute_instance.rabbitmq.network_interface[0].access_config[0].nat_ip}:15672"
-}
-
-# =============================================================================
-# Registry
-# =============================================================================
-
-output "registry_url" {
-  description = "Artifact Registry URL"
-  value       = "${var.region}-docker.pkg.dev/${var.project_id}/${var.registry_name}"
-}
-
-# =============================================================================
-# Connection Strings
-# =============================================================================
-
-output "celery_broker_url" {
-  description = "Celery broker URL (internal)"
-  value       = "amqp://${var.rabbitmq_user}:****@${google_compute_instance.rabbitmq.network_interface[0].network_ip}:5672//"
-  sensitive   = false
-}
-
-output "mysql_connection_string" {
-  description = "MySQL connection string (internal)"
-  value       = "mysql://${var.db_user}:****@${google_compute_instance.mysql.network_interface[0].network_ip}:3306/${var.db_name}"
-  sensitive   = false
-}
-
-# =============================================================================
-# Summary
-# =============================================================================
-
-output "deployment_summary" {
-  description = "Deployment summary"
-  value = <<-EOT
-
-    ===========================================
-    Speedcam MSA Deployment Summary
-    ===========================================
-
-    Project:     ${var.project_id}
-    Region:      ${var.region}
-    Zone:        ${var.zone}
-    Environment: ${var.environment}
-
-    -------------------------------------------
-    Infrastructure
-    -------------------------------------------
-    RabbitMQ:    ${google_compute_instance.rabbitmq.network_interface[0].network_ip} (${google_compute_instance.rabbitmq.network_interface[0].access_config[0].nat_ip})
-    MySQL:       ${google_compute_instance.mysql.network_interface[0].network_ip} (${google_compute_instance.mysql.network_interface[0].access_config[0].nat_ip})
-    DataDog:     ${google_compute_instance.datadog_agent.network_interface[0].network_ip} (${google_compute_instance.datadog_agent.network_interface[0].access_config[0].nat_ip})
-
-    -------------------------------------------
-    Services
-    -------------------------------------------
-    Main:        ${google_compute_instance.main.network_interface[0].network_ip} (${google_compute_instance.main.network_interface[0].access_config[0].nat_ip})
-    OCR:         ${google_compute_instance.ocr.network_interface[0].network_ip} (${google_compute_instance.ocr.network_interface[0].access_config[0].nat_ip})
-    Alert:       ${google_compute_instance.alert.network_interface[0].network_ip} (${google_compute_instance.alert.network_interface[0].access_config[0].nat_ip})
-
-    -------------------------------------------
-    URLs
-    -------------------------------------------
-    API:         http://${google_compute_instance.main.network_interface[0].access_config[0].nat_ip}:8000/
-    Swagger:     http://${google_compute_instance.main.network_interface[0].access_config[0].nat_ip}:8000/swagger/
-    Health:      http://${google_compute_instance.main.network_interface[0].access_config[0].nat_ip}:8000/health/
-    RabbitMQ:    http://${google_compute_instance.rabbitmq.network_interface[0].access_config[0].nat_ip}:15672/
-
-    ===========================================
-
-  EOT
-}
diff --git a/terraform/terraform.tfvars.example b/terraform/terraform.tfvars.example
deleted file mode 100644
index 1c062d5..0000000
--- a/terraform/terraform.tfvars.example
+++ /dev/null
@@ -1,54 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Terraform Variables Example
-# =============================================================================
-# Copy this file to terraform.tfvars and customize the values
-# =============================================================================
-
-# =============================================================================
-# Required Variables
-# =============================================================================
-
-# GCP Project ID (required)
-project_id = "your-project-id"
-
-# =============================================================================
-# Optional Variables (defaults provided)
-# =============================================================================
-
-# Region and Zone
-region = "asia-northeast3"
-zone   = "asia-northeast3-a"
-
-# Environment
-environment = "dev"  # dev, staging, prod
-
-# Database Configuration
-db_name          = "speedcam"
-db_user          = "sa"
-db_password      = "your-secure-password"
-db_root_password = "your-secure-root-password"
-
-# RabbitMQ Configuration
-rabbitmq_user     = "sa"
-rabbitmq_password = "your-secure-password"
-
-# Instance Types
-machine_type_small  = "e2-small"
-machine_type_medium = "e2-medium"
-
-# Container Images
-image_tag      = "latest"
-mysql_image    = "mysql:8.0"
-rabbitmq_image = "rabbitmq:3.13-management"
-
-# Application Configuration
-ocr_concurrency   = 2
-alert_concurrency = 50
-ocr_mock          = true
-fcm_mock          = true
-
-# =============================================================================
-# DataDog
-# =============================================================================
-dd_api_key = "your-datadog-api-key"
-# dd_site  = "ap1.datadoghq.com"  # default
diff --git a/terraform/variables.tf b/terraform/variables.tf
deleted file mode 100644
index e63e018..0000000
--- a/terraform/variables.tf
+++ /dev/null
@@ -1,184 +0,0 @@
-# =============================================================================
-# Speedcam MSA - Terraform Variables
-# =============================================================================
-
-# =============================================================================
-# Project Configuration
-# =============================================================================
-
-variable "project_id" {
-  description = "GCP Project ID"
-  type        = string
-}
-
-variable "region" {
-  description = "GCP Region"
-  type        = string
-  default     = "asia-northeast3"
-}
-
-variable "zone" {
-  description = "GCP Zone"
-  type        = string
-  default     = "asia-northeast3-a"
-}
-
-variable "environment" {
-  description = "Environment (dev, staging, prod)"
-  type        = string
-  default     = "dev"
-
-  validation {
-    condition     = contains(["dev", "staging", "prod"], var.environment)
-    error_message = "Environment must be one of: dev, staging, prod."
-  }
-}
-
-# =============================================================================
-# Network Configuration
-# =============================================================================
-
-variable "network_name" {
-  description = "VPC Network name"
-  type        = string
-  default     = "default"
-}
-
-# =============================================================================
-# Artifact Registry Configuration
-# =============================================================================
-
-variable "registry_name" {
-  description = "Artifact Registry repository name"
-  type        = string
-  default     = "speedcam"
-}
-
-# =============================================================================
-# Database Configuration
-# =============================================================================
-
-variable "db_name" {
-  description = "Base database name"
-  type        = string
-  default     = "speedcam"
-}
-
-variable "db_user" {
-  description = "Database user"
-  type        = string
-  default     = "sa"
-}
-
-variable "db_password" {
-  description = "Database password"
-  type        = string
-  sensitive   = true
-  default     = "1234"
-}
-
-variable "db_root_password" {
-  description = "Database root password"
-  type        = string
-  sensitive   = true
-  default     = "root"
-}
-
-# =============================================================================
-# RabbitMQ Configuration
-# =============================================================================
-
-variable "rabbitmq_user" {
-  description = "RabbitMQ user"
-  type        = string
-  default     = "sa"
-}
-
-variable "rabbitmq_password" {
-  description = "RabbitMQ password"
-  type        = string
-  sensitive   = true
-  default     = "1234"
-}
-
-# =============================================================================
-# Instance Configuration
-# =============================================================================
-
-variable "machine_type_small" {
-  description = "Machine type for small instances"
-  type        = string
-  default     = "e2-small"
-}
-
-variable "machine_type_medium" {
-  description = "Machine type for medium instances"
-  type        = string
-  default     = "e2-medium"
-}
-
-# =============================================================================
-# Container Images
-# =============================================================================
-
-variable "image_tag" {
-  description = "Docker image tag"
-  type        = string
-  default     = "latest"
-}
-
-variable "mysql_image" {
-  description = "MySQL Docker image"
-  type        = string
-  default     = "mysql:8.0"
-}
-
-variable "rabbitmq_image" {
-  description = "RabbitMQ Docker image"
-  type        = string
-  default     = "rabbitmq:3.13-management"
-}
-
-# =============================================================================
-# Application Configuration
-# =============================================================================
-
-variable "ocr_concurrency" {
-  description = "OCR worker concurrency"
-  type        = number
-  default     = 2
-}
-
-variable "alert_concurrency" {
-  description = "Alert worker concurrency"
-  type        = number
-  default     = 50
-}
-
-variable "ocr_mock" {
-  description = "Enable OCR mock mode"
-  type        = bool
-  default     = true
-}
-
-variable "fcm_mock" {
-  description = "Enable FCM mock mode"
-  type        = bool
-  default     = true
-}
-
-# =============================================================================
-# DataDog
-# =============================================================================
-
-variable "dd_api_key" {
-  description = "DataDog API Key"
-  type        = string
-  sensitive   = true
-}
-
-variable "dd_site" {
-  description = "DataDog site (e.g., ap1.datadoghq.com)"
-  type        = string
-  default     = "ap1.datadoghq.com"
-}