feat: add async retry mechanism with exponential backoff

- Add app/utils/retry.py with configurable async retry decorator
- Update DeliveryLog model to track attempt_count and latency_seconds
- Apply @http_retry to engine._exec_forward and _exec_notify methods
- Update save_logs to record retry metadata
- Add comprehensive unit tests for retry functionality
- Support configuration via environment variables (RETRY_*)

This improves reliability for downstream HTTP calls by automatically
retrying transient failures with exponential backoff and jitter.
This commit is contained in:
auto-bot
2025-12-24 11:04:41 +08:00
parent 0def77dc30
commit b11c39f3bf
5 changed files with 269 additions and 33 deletions
+99
View File
@@ -0,0 +1,99 @@
"""
Async retry decorator with exponential backoff and configurable parameters.
"""
import asyncio
import logging
from functools import wraps
from typing import Callable, Any, Optional
import time
import os
logger = logging.getLogger(__name__)
def async_retry(
max_attempts: int = 3,
initial_delay: float = 1.0,
backoff_factor: float = 2.0,
max_delay: float = 60.0,
retry_on: tuple = (Exception,),
jitter: bool = True
):
"""
Decorator for async functions that implements exponential backoff retry logic.
Args:
max_attempts: Maximum number of retry attempts (including initial call)
initial_delay: Initial delay in seconds before first retry
backoff_factor: Factor by which delay increases each retry
max_delay: Maximum delay between retries
retry_on: Tuple of exception types to retry on
jitter: Add random jitter to delay to prevent thundering herd
"""
def decorator(func: Callable) -> Callable:
@wraps(func)
async def wrapper(*args, **kwargs) -> tuple[Any, dict]:
"""
Returns:
tuple: (result, metadata_dict)
metadata_dict contains: attempts, total_latency, last_error
"""
last_error = None
start_time = time.time()
for attempt in range(max_attempts):
try:
result = await func(*args, **kwargs)
total_latency = time.time() - start_time
return result, {
'attempts': attempt + 1,
'total_latency': round(total_latency, 3),
'last_error': None,
'success': True
}
except retry_on as e:
last_error = str(e)
if attempt < max_attempts - 1: # Don't sleep after last attempt
delay = min(initial_delay * (backoff_factor ** attempt), max_delay)
if jitter:
# Add random jitter (±25% of delay)
import random
jitter_range = delay * 0.25
delay += random.uniform(-jitter_range, jitter_range)
logger.warning(f"Attempt {attempt + 1}/{max_attempts} failed for {func.__name__}: {e}. Retrying in {delay:.2f}s")
await asyncio.sleep(delay)
else:
logger.error(f"All {max_attempts} attempts failed for {func.__name__}: {e}")
total_latency = time.time() - start_time
return None, {
'attempts': max_attempts,
'total_latency': round(total_latency, 3),
'last_error': last_error,
'success': False
}
return wrapper
return decorator
# Configuration from environment
def get_retry_config():
"""Get retry configuration from environment variables."""
return {
'max_attempts': int(os.getenv('RETRY_MAX_ATTEMPTS', '3')),
'initial_delay': float(os.getenv('RETRY_INITIAL_DELAY', '1.0')),
'backoff_factor': float(os.getenv('RETRY_BACKOFF_FACTOR', '2.0')),
'max_delay': float(os.getenv('RETRY_MAX_DELAY', '30.0')),
}
# Pre-configured decorators for common use cases
def http_retry(**kwargs):
"""Retry decorator specifically for HTTP operations."""
config = get_retry_config()
config.update(kwargs)
return async_retry(
retry_on=(Exception,), # Retry on any exception for HTTP calls
**config
)