Add per-module log level overrides for staging/production

Support LOG_LEVEL_{MODULE} env vars (GAME, AI, HANDLERS, ROOM, AUTH,
STORES) to override the global log level for specific modules. Active
overrides are logged at startup. Includes staging/production presets
in .env.example files and a V3.18 stub doc for PostgreSQL storage
efficiency investigation.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
adlee-was-taken 2026-02-23 17:12:11 -05:00
parent 1b923838e0
commit e463d929e3
4 changed files with 133 additions and 0 deletions

View File

@ -20,6 +20,24 @@ DEBUG=false
# Logging level: DEBUG, INFO, WARNING, ERROR, CRITICAL
LOG_LEVEL=INFO
# Per-module log level overrides (optional)
# These override LOG_LEVEL for specific modules.
# LOG_LEVEL_GAME=DEBUG # Core game logic
# LOG_LEVEL_AI=DEBUG # AI decisions (very verbose at DEBUG)
# LOG_LEVEL_HANDLERS=DEBUG # WebSocket message handlers
# LOG_LEVEL_ROOM=DEBUG # Room/lobby management
# LOG_LEVEL_AUTH=DEBUG # Auth stack (auth, routers.auth, services.auth_service)
# LOG_LEVEL_STORES=DEBUG # Database/Redis operations
# --- Preset examples ---
# Staging (debug game logic, quiet everything else):
# LOG_LEVEL=INFO
# LOG_LEVEL_GAME=DEBUG
# LOG_LEVEL_AI=DEBUG
#
# Production (minimal logging):
# LOG_LEVEL=WARNING
# Environment name (development, staging, production)
ENVIRONMENT=development

View File

@ -0,0 +1,57 @@
# V3.18: PostgreSQL Game Data Storage Efficiency
**Status:** Planning
**Priority:** Medium
**Category:** Infrastructure / Performance
## Problem
Per-move game logging stores full `hand_state` and `visible_opponents` JSONB on every move. For a typical 6-player, 9-hole game this generates significant redundant data since most of each player's hand doesn't change between moves.
## Areas to Investigate
### 1. Delta Encoding for Move Data
Store only what changed from the previous move instead of full state snapshots.
- First move of each round stores full state (baseline)
- Subsequent moves store only changed positions (e.g., `{"player_0": {"pos_2": "5H"}}`)
- Replay reconstruction applies deltas sequentially
- Trade-off: simpler queries vs. storage savings
### 2. PostgreSQL TOAST and Compression
- TOAST already compresses large JSONB values automatically
- Measure actual on-disk size vs. logical size for typical game data
- Consider whether explicit compression (e.g., storing gzipped blobs) adds meaningful savings over TOAST
### 3. Retention Policy
- Archive completed games older than N days to a separate table or cold storage
- Configurable retention period via env var (e.g., `GAME_LOG_RETENTION_DAYS`)
- Keep aggregate stats even after pruning raw move data
### 4. Move Logging Toggle
- Env var `GAME_LOGGING_ENABLED=true|false` to disable move-level logging entirely
- Useful for non-analysis environments (dev, load testing)
- Game outcomes and stats would still be recorded
### 5. Batch Inserts
- Buffer moves in memory and flush periodically instead of per-move INSERT
- Reduces database round-trips during active games
- Risk: data loss if server crashes mid-game (acceptable for non-critical move logs)
## Measurements Needed
Before optimizing, measure current impact:
- Average JSONB size per move (bytes)
- Average moves per game
- Total storage per game (moves + overhead)
- Query patterns: how often is per-move data actually read?
## Dependencies
- None (independent infrastructure improvement)

View File

@ -7,6 +7,24 @@ PORT=8000
DEBUG=true
LOG_LEVEL=DEBUG
# Per-module log level overrides (optional)
# These override LOG_LEVEL for specific modules.
# LOG_LEVEL_GAME=DEBUG # Core game logic
# LOG_LEVEL_AI=DEBUG # AI decisions (very verbose at DEBUG)
# LOG_LEVEL_HANDLERS=DEBUG # WebSocket message handlers
# LOG_LEVEL_ROOM=DEBUG # Room/lobby management
# LOG_LEVEL_AUTH=DEBUG # Auth stack (auth, routers.auth, services.auth_service)
# LOG_LEVEL_STORES=DEBUG # Database/Redis operations
# --- Preset examples ---
# Staging (debug game logic, quiet everything else):
# LOG_LEVEL=INFO
# LOG_LEVEL_GAME=DEBUG
# LOG_LEVEL_AI=DEBUG
#
# Production (minimal logging):
# LOG_LEVEL=WARNING
# Environment (development, staging, production)
# Affects logging format, security headers (HSTS), etc.
ENVIRONMENT=development

View File

@ -148,6 +148,39 @@ class DevelopmentFormatter(logging.Formatter):
return output
# Per-module log level overrides via env vars.
# Key: env var suffix, Value: list of Python logger names to apply to.
MODULE_LOGGER_MAP = {
"GAME": ["game"],
"AI": ["ai"],
"HANDLERS": ["handlers"],
"ROOM": ["room"],
"AUTH": ["auth", "routers.auth", "services.auth_service"],
"STORES": ["stores"],
}
def _apply_module_overrides() -> dict[str, str]:
"""
Apply per-module log level overrides from LOG_LEVEL_{MODULE} env vars.
Returns:
Dict of module name -> level for any overrides that were applied.
"""
active = {}
for module, logger_names in MODULE_LOGGER_MAP.items():
env_val = os.environ.get(f"LOG_LEVEL_{module}", "").upper()
if not env_val:
continue
level = getattr(logging, env_val, None)
if level is None:
continue
active[module] = env_val
for name in logger_names:
logging.getLogger(name).setLevel(level)
return active
def setup_logging(
level: str = "INFO",
environment: str = "development",
@ -182,12 +215,19 @@ def setup_logging(
logging.getLogger("websockets").setLevel(logging.WARNING)
logging.getLogger("asyncio").setLevel(logging.WARNING)
# Apply per-module overrides from env vars
overrides = _apply_module_overrides()
# Log startup
logger = logging.getLogger(__name__)
logger.info(
f"Logging configured: level={level}, environment={environment}",
extra={"level": level, "environment": environment},
)
if overrides:
logger.info(
f"Per-module log level overrides: {', '.join(f'{m}={l}' for m, l in overrides.items())}",
)
class ContextLogger(logging.LoggerAdapter):