botobackoff provides a boto3 client wrapper and function decorator that automatically retries failed boto3 calls with exponential backoff and jitter.
The decorator and client wrapper include options to add additional error codes to catch and retry on and
error codes to ignore and fail silently returning None.
Boto3 provides limited retry functionality for failed calls. I wanted to create a simple way to retry boto3 calls with exponential backoff and jitter as well as more flexibility as to what errors can be retried or ignored without needing to add additional try/except blocks around boto3 calls.
install from PyPI: pip install botobackoff
Example with the client wrapper:
from botobackoff import BotoBackoff
# `BotoBackoff` can be instantiated with a boto3 client or service name
wrapped_client = BotoBackoff("s3", max_retries=5)
print(wrapped_client.list_objects(Bucket="my-bucket"))Example as context manager:
from botobackoff import BotoBackoff
wrapped_client = BotoBackoff("s3", max_retries=2)
# use `with_options` to change any of the options for the context manager instance
with wrapped_client.with_options(max_retries=5) as client:
print(client.list_objects(Bucket="my-bucket"))Example with the decorator:
from botobackoff import botobackoff
import boto3
@botobackoff(max_retries=5)
def list_objects():
return boto3.client("s3").list_objects(Bucket="my-bucket")
print(list_objects())The boto3 errors that are retried by default are:
- ThrottlingException
- TooManyRequestsException
- ServiceUnavailableException
- RequestLimitExceeded
- RequestThrottled
- RequestThrottledException
- ProvisionedThroughputExceededException
- LimitExceededException
- EndpointConnectionError
- ConnectTimeoutError
- Unavailable
- InternalFailure
- InternalError
to add additional errors to retry on use the added_error_codes parameter:
from botobackoff import botobackoff
import boto3
@botobackoff(added_error_codes=["NoSuchBucket"])
def list_objects():
return boto3.client("s3").list_objects(Bucket="my-bucket")