[FEATURE] GPU 정보 수집을 위한 모니터링 대상 변경

app/main.py

@@ -2,7 +2,6 @@
 from app.core.lifespan import lifespan
 from app.api.endpoints import predictions
 from app.core.logging_config import setup_logging
-from prometheus_fastapi_instrumentator import Instrumentator
 
 setup_logging()
 app = FastAPI(
@@ -12,8 +11,6 @@
     lifespan=lifespan
 )
 
-instrumentator = Instrumentator().instrument(app)
-instrumentator.expose(app, include_in_schema=False, endpoint="/actuator/prometheus")
 app.include_router(predictions.router, prefix="/api/v1", tags=["Prediction"])
 
 @app.get("/health")

deployment/docker-compose.monitoring.yml

@@ -28,6 +28,25 @@ services:
     networks:
       - mynetwork
 
+  dcgm-exporter:
+      image: nvcr.io/nvidia/k8s/dcgm-exporter:latest
+      container_name: dcgm-exporter
+      ports:
+        - "9400:9400"
+      deploy:
+        resources:
+          reservations:
+            devices:
+              - driver: nvidia
+                count: all
+                capabilities: [gpu]
+      environment:
+        - NVIDIA_VISIBLE_DEVICES=all
+        - NVIDIA_DRIVER_CAPABILITIES=utility,compute
+      restart: unless-stopped
+      networks:
+        - mynetwork
+
 networks:
   mynetwork:
     driver: bridge

deployment/docker-compose.yml

@@ -13,6 +13,8 @@ services:
               count: 1
               capabilities: [gpu]
     container_name: app-blue
+    networks:
+      - mynetwork
 
   app-green:
     image: ${ECR_URI}/dearbelly-cv:latest
@@ -27,4 +29,10 @@ services:
             - driver: nvidia
               count: 1
               capabilities: [ gpu ]
-    container_name: app-green
+    container_name: app-green
+    networks:
+      - mynetwork
+
+networks:
+  mynetwork:
+    driver: bridge

deployment/generate_review.py

@@ -59,7 +59,7 @@ def send_prompt():
 
     # Gemini에 전달할 프롬프트 구성
     prompt = f"""
-    당신은 시니어 개발자입니다. 제출된 Pull Request(PR)에 대해 동료 개발자에게 건설적이고 상세한 코드 리뷰를 제공합니다.
+    너는 시니어 개발자다. 제출된 Pull Request(PR)에 대해 건설적이고 상세한 코드 리뷰를 제공해주세요.
     리뷰는 반드시 '우선순위 레벨(P1~P5)'로 분류해 주세요.
 
     [우선순위 정의]

deployment/prometheus.yml

@@ -16,19 +16,15 @@ scrape_configs:
     honor_timestamps: false
     scheme: 'http'
     static_configs:
-      - targets: ['${REMOTE_HOST}:9090']
+      - targets: ['127.0.0.1:9090']
         labels:
           service: 'monitor-1'
   - job_name: 'node'
     static_configs:
-      - targets: ['${REMOTE_HOST}:9090']
-  - job_name: 'fastapi-actuator-blue'
-    metrics_path: '/actuator/prometheus'
-    scrape_interval: 1m
+      - targets: ['127.0.0.1:9090']
+  - job_name: 'dcgm'
+    scrape_interval: 15s
     static_configs:
-      - targets: [ 'app-blue:8000' ]
-  - job_name: 'fastapi-actuator-green'
-    metrics_path: '/actuator/prometheus'
-    scrape_interval: 1m
-    static_configs:
-      - targets: [ 'app-green:8001' ]
+      - targets: ['dcgm-exporter:9400']
+        labels:
+          exporter: 'dcgm'

requirements.txt

@@ -10,6 +10,4 @@ Pillow==11.3.0
 dotenv
 openai
 timm
-logging
-prometheus-client==0.19.0
-prometheus-fastapi-instrumentator==6.1.0
+logging