perf: API 성능 개선 Round 2 — 9개 항목 (P6-P14)

Tier 1:
- P6: 미들웨어 request_info 지연 추출 (성공 경로에서 body/headers 읽기 제거)
- P7: GZip 압축 미들웨어 (minimum_size=1000, ~4x 압축)
- P8: Redis maxmemory 512MB + allkeys-lru 퇴거 정책

Tier 2:
- P9: _store_ticker_data() batch upsert 전환 (ON CONFLICT DO NOTHING)
- P10: DB 커넥션 풀 증가 (pool_size=20, max_overflow=30, 환경변수 설정)
- P11: Attention 엔드포인트 Redis 캐시 추가 (@with_cache ttl=3600)

Tier 3:
- P12: Dockerfile 멀티 워커 (--workers 4, dev는 --reload 오버라이드)
- P13: aiohttp ClientSession 싱글턴 공유 (4개 파일 7곳 TCP/TLS 재사용)
- P14: Pydantic model_validate → model_construct (bulk 경로 validation 스킵)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Dockerfile

@@ -24,4 +24,4 @@ RUN mkdir -p /app/data
 EXPOSE 18000
 
 # Run the application
-CMD ["python", "-m", "uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "18000", "--reload"]
+CMD ["python", "-m", "uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "18000", "--workers", "4"]

app/api/v1/endpoints/attention.py

@@ -16,11 +16,12 @@ Admin endpoints:
 import logging
 from datetime import date, timedelta
 
-from fastapi import APIRouter, Depends, HTTPException, Query
+from fastapi import APIRouter, Depends, HTTPException, Query, Response
 from sqlalchemy import func, select
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from app.core.database import get_db
+from app.utils.cache import with_cache
 from app.models.attention import AttentionFeaturesDaily, CompanyEntityMap, GdeltArticleRaw
 from app.services.attention.gdelt_collector import GDELT_EARLIEST_DATE
 from app.schemas.attention import (
@@ -237,8 +238,10 @@ async def admin_collect_gdelt(
     **Example**: `GET /attention/entity/AAPL`
     """,
 )
+@with_cache(namespace="attention:entity", ttl=3600, key_params=["ticker"])
 async def get_entity(
     ticker: str,
+    response: Response = None,
     db: AsyncSession = Depends(get_db),
 ) -> EntityResolveResponse:
     ticker = ticker.upper()
@@ -297,9 +300,11 @@ async def get_entity(
         500: {"description": "Feature materialization or collection error"},
     },
 )
+@with_cache(namespace="attention:event", ttl=3600, key_params=["ticker", "event_date"])
 async def get_event_attention(
     ticker: str,
     event_date: date = Query(..., description="Event date in YYYY-MM-DD format"),
+    response: Response = None,
     db: AsyncSession = Depends(get_db),
 ) -> EventAttentionResponse:
     ticker = ticker.upper()

app/core/config.py

@@ -41,6 +41,8 @@ class Settings(BaseSettings):
         "postgresql+asyncpg://stockoracle:stockoracle2024@localhost:15433/stock_oracle"
     )
     DATABASE_ECHO: bool = False
+    DB_POOL_SIZE: int = int(os.getenv("DB_POOL_SIZE", "20"))
+    DB_MAX_OVERFLOW: int = int(os.getenv("DB_MAX_OVERFLOW", "30"))
     
     # Redis
     REDIS_URL: str = os.getenv("REDIS_URL", "redis://localhost:16380/0")

app/core/database.py

@@ -11,8 +11,8 @@ engine = create_async_engine(
     settings.DATABASE_URL,
     echo=settings.DATABASE_ECHO,
     future=True,
-    pool_size=10,
+    pool_size=settings.DB_POOL_SIZE,
-    max_overflow=20,
+    max_overflow=settings.DB_MAX_OVERFLOW,
     pool_timeout=30,
     pool_pre_ping=True,
     pool_recycle=3600,

app/core/http_client.py

@@ -0,0 +1,33 @@
+"""
+Shared aiohttp ClientSession singleton.
+
+Reuses a single TCP connection pool across all HTTP clients,
+avoiding repeated TCP/TLS handshake overhead per request.
+"""
+
+import logging
+from typing import Optional
+
+import aiohttp
+
+logger = logging.getLogger(__name__)
+
+_session: Optional[aiohttp.ClientSession] = None
+
+
+async def get_http_session() -> aiohttp.ClientSession:
+    """Get or create the shared aiohttp ClientSession."""
+    global _session
+    if _session is None or _session.closed:
+        _session = aiohttp.ClientSession(
+            timeout=aiohttp.ClientTimeout(total=30),
+        )
+    return _session
+
+
+async def close_http_session() -> None:
+    """Close the shared session. Call on app shutdown."""
+    global _session
+    if _session and not _session.closed:
+        await _session.close()
+        _session = None

app/main.py

@@ -6,6 +6,7 @@ import os
 from contextlib import asynccontextmanager
 from fastapi import FastAPI, HTTPException
 from fastapi.middleware.cors import CORSMiddleware
+from starlette.middleware.gzip import GZipMiddleware
 from fastapi.responses import RedirectResponse
 
 from app.core.config import settings
@@ -36,6 +37,11 @@ async def lifespan(app: FastAPI):
         stop_scheduler()
     except Exception:
         pass
+    try:
+        from app.core.http_client import close_http_session
+        await close_http_session()
+    except Exception:
+        pass
     await engine.dispose()
 
 # Create FastAPI app
@@ -61,6 +67,9 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
+# GZip compression for responses > 1KB
+app.add_middleware(GZipMiddleware, minimum_size=1000)
+
 # Include API router
 app.include_router(api_router, prefix=settings.API_PREFIX)
 

app/middleware/error_logger.py

@@ -105,7 +105,7 @@ class ErrorLoggingMiddleware(BaseHTTPMiddleware):
         start_time = time.time()
         
         # Store request details for potential error logging
-        request_info = await self._extract_request_info(request)
+        request_info = self._extract_request_info_minimal(request)
 
         logger.info(f"Processing request {request_id}: {request.method} {request.url.path}")
         
@@ -124,6 +124,9 @@ class ErrorLoggingMiddleware(BaseHTTPMiddleware):
             if response.status_code >= 400:
                 logger.info(f"Error response detected: {response.status_code} for request {request_id}")
 
+                # Full extraction for error logging (body + headers)
+                request_info = await self._extract_request_info(request)
+
                 # Try to capture response body for errors
                 # We need to consume the response body and recreate it
                 from starlette.responses import Response
@@ -184,6 +187,9 @@ class ErrorLoggingMiddleware(BaseHTTPMiddleware):
             # Log unexpected errors
             response_time_ms = (time.time() - start_time) * 1000
 
+            # Full extraction for error logging
+            request_info = await self._extract_request_info(request)
+
             await self._log_error(
                 request_id=request_id,
                 request_info=request_info,
@@ -206,8 +212,21 @@ class ErrorLoggingMiddleware(BaseHTTPMiddleware):
                 headers={"X-Request-ID": request_id}
             )
     
+    def _extract_request_info_minimal(self, request: Request) -> dict:
+        """Extract minimal request info (sync, no body read) for success-path logging."""
+        return {
+            "endpoint": str(request.url.path),
+            "method": request.method,
+            "path": str(request.url),
+            "query_params": dict(request.query_params) if request.query_params else None,
+            "request_body": None,
+            "headers": None,
+            "user_agent": request.headers.get("user-agent"),
+            "client_ip": request.client.host if request.client else None,
+        }
+
     async def _extract_request_info(self, request: Request) -> dict:
-        """Extract request information for logging"""
+        """Extract full request information for error logging"""
         
         # Get request body if present
         body = None

app/services/finra_short_volume_service.py

@@ -10,6 +10,8 @@ import aiohttp
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy import select, and_, func, desc
 
+from app.core.http_client import get_http_session
+
 from app.models.finra_short_volume import FinraShortVolume
 
 logger = logging.getLogger(__name__)
@@ -34,7 +36,7 @@ class FinraShortVolumeService:
         url = f"{FINRA_CDN_BASE}/{filename}"
 
         try:
-            async with aiohttp.ClientSession() as session:
+            session = await get_http_session()
             async with session.get(url, timeout=aiohttp.ClientTimeout(total=30)) as resp:
                 if resp.status == 404:
                     logger.debug(f"FINRA file not found (holiday/weekend): {filename}")

app/services/news_social_service.py

@@ -10,6 +10,8 @@ Aggregates news and social media data from multiple sources for sentiment analys
 import asyncio
 import aiohttp
 import logging
+
+from app.core.http_client import get_http_session
 from datetime import datetime, timedelta
 from typing import Dict, List, Optional, Any, Union
 from dataclasses import dataclass
@@ -326,7 +328,7 @@ class NewsSocialService:
                 "language": "en"
             }
             
-            async with aiohttp.ClientSession() as session:
+            session = await get_http_session()
             async with session.get(url, params=params) as response:
                 if response.status == 200:
                     data = await response.json()
@@ -405,7 +407,7 @@ class NewsSocialService:
                         "User-Agent": "StockOracle/1.0.0"
                     }
                     
-                    async with aiohttp.ClientSession() as session:
+                    session = await get_http_session()
                     async with session.get(url, params=params, headers=headers) as response:
                         if response.status == 200:
                             data = await response.json()
@@ -497,7 +499,7 @@ class NewsSocialService:
             
             auth = aiohttp.BasicAuth(self.reddit_client_id, self.reddit_client_secret)
             
-            async with aiohttp.ClientSession() as session:
+            session = await get_http_session()
             async with session.post(auth_url, data=auth_data, auth=auth, headers=headers) as response:
                 if response.status == 200:
                     token_data = await response.json()

app/services/overlay/yahoo_rss_adapter.py

@@ -11,6 +11,8 @@ import aiohttp
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy import select
 
+from app.core.http_client import get_http_session
+
 from app.models.overlay_raw_event import OverlayHeadlineEvent
 from app.services.overlay.entity_resolver import EntityResolver
 from app.core.overlay_config import YAHOO_RSS_FEEDS, TOP_50_SYMBOLS
@@ -33,7 +35,7 @@ class YahooRSSAdapter:
             return []
 
         try:
-            async with aiohttp.ClientSession() as session:
+            session = await get_http_session()
             headers = {
                 "User-Agent": "Mozilla/5.0 StockOracle/1.0",
                 "Accept": "application/rss+xml, application/xml, text/xml",

app/services/overlay/youtube_adapter.py

@@ -9,6 +9,8 @@ from typing import Dict, List, Optional
 
 import aiohttp
 from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.core.http_client import get_http_session
 from sqlalchemy import select
 
 from app.core.config import settings
@@ -53,7 +55,7 @@ class YouTubeAdapter:
             "key": self.api_key,
         }
         try:
-            async with aiohttp.ClientSession() as session:
+            session = await get_http_session()
             async with session.get(
                 YOUTUBE_SEARCH_URL, params=params, timeout=aiohttp.ClientTimeout(total=30)
             ) as resp:
@@ -77,7 +79,7 @@ class YouTubeAdapter:
             "key": self.api_key,
         }
         try:
-            async with aiohttp.ClientSession() as session:
+            session = await get_http_session()
             async with session.get(
                 YOUTUBE_VIDEOS_URL, params=params, timeout=aiohttp.ClientTimeout(total=30)
             ) as resp:

app/services/price_data_service.py

@@ -588,7 +588,12 @@ class PriceDataService:
                         
                         # Convert to response models
                         price_points = [
-                            PriceDataPoint.model_validate(pd) for pd in price_data
+                            PriceDataPoint.model_construct(
+                                date=pd.date.date() if isinstance(pd.date, datetime) else pd.date,
+                                open=pd.open, high=pd.high, low=pd.low, close=pd.close,
+                                volume=pd.volume, adjusted_close=pd.adjusted_close,
+                                data_source=pd.data_source.value if hasattr(pd.data_source, 'value') else pd.data_source,
+                            ) for pd in price_data
                         ]
                         
                         # Calculate actual date range from returned data
@@ -821,39 +826,48 @@ class PriceDataService:
         ticker_data,
         interval: str
     ):
-        """Store individual ticker data with optimized batch operations"""
+        """Store individual ticker data using batch upsert (INSERT ... ON CONFLICT DO NOTHING)."""
-        # Batch check existing dates to avoid individual DB queries
+        from sqlalchemy.dialects.postgresql import insert as pg_insert
-        existing_dates = await self._get_existing_dates_for_ticker(db, ticker)
+        import uuid as _uuid
-        
+
-        new_records = []
+        def _safe(val):
-        for date, row in ticker_data.iterrows():
+            if val is None:
-            # Convert pandas timestamp to datetime
+                return None
-            price_date = date.to_pydatetime()
+            try:
+                v = float(val)
+                return None if pd.isna(v) else v
+            except (TypeError, ValueError):
+                return None
+
+        now = datetime.now(timezone.utc)
+        rows = []
+        for date_idx, row in ticker_data.iterrows():
+            price_date = date_idx.to_pydatetime()
             if price_date.tzinfo is None:
                 price_date = price_date.replace(tzinfo=timezone.utc)
 
-            # Skip if already exists
+            rows.append({
-            if price_date.date() in existing_dates:
+                'id': _uuid.uuid4(),
-                continue
+                'ticker': ticker,
+                'date': price_date,
+                'open': _safe(row.get('Open')),
+                'high': _safe(row.get('High')),
+                'low': _safe(row.get('Low')),
+                'close': _safe(row.get('Close')) or 0.0,
+                'volume': _safe(row.get('Volume')),
+                'adjusted_close': _safe(row.get('Close')),
+                'data_source': DataSource.YAHOO_FINANCE.value,
+                'created_at': now,
+                'updated_at': now,
+            })
 
-            # Prepare new record
+        if not rows:
-            price_record = PriceData(
+            return
-                ticker=ticker,
+
-                date=price_date,
+        stmt = pg_insert(PriceData).values(rows)
-                open=float(row.get('Open', 0)) if not pd.isna(row.get('Open')) else None,
+        stmt = stmt.on_conflict_do_nothing(constraint='uq_price_data')
-                high=float(row.get('High', 0)) if not pd.isna(row.get('High')) else None,
+        await db.execute(stmt)
-                low=float(row.get('Low', 0)) if not pd.isna(row.get('Low')) else None,
+        logger.info(f"Batch upserted {len(rows)} price records for {ticker}")
-                close=float(row.get('Close', 0)) if not pd.isna(row.get('Close')) else 0,
-                volume=float(row.get('Volume', 0)) if not pd.isna(row.get('Volume')) else None,
-                adjusted_close=float(row.get('Close', 0)) if not pd.isna(row.get('Close')) else None,
-                data_source=DataSource.YAHOO_FINANCE
-            )
-            new_records.append(price_record)
-        
-        # Batch insert new records
-        if new_records:
-            db.add_all(new_records)
-            logger.info(f"Added {len(new_records)} new price records for {ticker}")
     
     async def _get_existing_dates_for_ticker(
         self,
@@ -927,7 +941,12 @@ class PriceDataService:
                 
                 # Convert to response models
                 price_points = [
-                    PriceDataPoint.model_validate(pd) for pd in price_data
+                    PriceDataPoint.model_construct(
+                        date=pd.date.date() if isinstance(pd.date, datetime) else pd.date,
+                        open=pd.open, high=pd.high, low=pd.low, close=pd.close,
+                        volume=pd.volume, adjusted_close=pd.adjusted_close,
+                        data_source=pd.data_source.value if hasattr(pd.data_source, 'value') else pd.data_source,
+                    ) for pd in price_data
                 ]
                 
                 # Calculate actual date range from returned data

docker-compose.yml

@@ -23,7 +23,7 @@ services:
   redis:
     image: redis:7-alpine
     container_name: stock_oracle_cache
-    command: redis-server --appendonly yes
+    command: redis-server --appendonly yes --maxmemory 512mb --maxmemory-policy allkeys-lru
     ports:
       - "16380:6379"  # Unique port to avoid conflicts
     volumes: