Dectinc / tenacity

Retrying library for Python

Tenacity

Tenacity is an Apache 2.0 licensed general-purpose retrying library, written in Python, to simplify the task of adding retry behavior to just about anything. It originates from a fork of retrying which is sadly no longer maintained. Tenacity isn't api compatible with retrying but adds significant new functionality and fixes a number of longstanding bugs.

The simplest use case is retrying a flaky function whenever an Exception occurs until a value is returned.

import random from tenacity import retry

@retry def do_something_unreliable(): if random.randint(0, 10) > 1: raise IOError("Broken sauce, everything is hosed!!!111one") else: return "Awesome sauce!"

print(do_something_unreliable())

Awesome sauce!

changelog api

Features

• Generic Decorator API
• Specify stop condition (i.e. limit by number of attempts)
• Specify wait condition (i.e. exponential backoff sleeping between attempts)
• Customize retrying on Exceptions
• Customize retrying on expected returned result
• Retry on coroutines

Installation

To install tenacity, simply:

$ pip install tenacity

Examples

Basic Retry

import logging from tenacity import *

class MyException(Exception):

pass

As you saw above, the default behavior is to retry forever without waiting when an exception is raised.

@retry def never_give_up_never_surrender(): print("Retry forever ignoring Exceptions, don't wait between retries") raise Exception

Stopping

Let's be a little less persistent and set some boundaries, such as the number of attempts before giving up.

@retry(stop=stop_after_attempt(7)) def stop_after_7_attempts(): print("Stopping after 7 attempts") raise Exception

We don't have all day, so let's set a boundary for how long we should be retrying stuff.

@retry(stop=stop_after_delay(10)) def stop_after_10_s(): print("Stopping after 10 seconds") raise Exception

You can combine several stop conditions by using the | operator:

@retry(stop=(stop_after_delay(10) | stop_after_attempt(5))) def stop_after_10_s_or_5_retries(): print("Stopping after 10 seconds or 5 retries") raise Exception

Waiting before retrying

Most things don't like to be polled as fast as possible, so let's just wait 2 seconds between retries.

@retry(wait=wait_fixed(2)) def wait_2_s(): print("Wait 2 second between retries") raise Exception

Some things perform best with a bit of randomness injected.

@retry(wait=wait_random(min=1, max=2)) def wait_random_1_to_2_s(): print("Randomly wait 1 to 2 seconds between retries") raise Exception

Then again, it's hard to beat exponential backoff when retrying distributed services and other remote endpoints.

@retry(wait=wait_exponential(multiplier=1, min=4, max=10)) def wait_exponential_1(): print("Wait 2^x * 1 second between each retry starting with 4 seconds, then up to 10 seconds, then 10 seconds afterwards") raise Exception

Then again, it's also hard to beat combining fixed waits and jitter (to help avoid thundering herds) when retrying distributed services and other remote endpoints.

@retry(wait=wait_fixed(3) + wait_random(0, 2)) def wait_fixed_jitter(): print("Wait at least 3 seconds, and add up to 2 seconds of random delay") raise Exception

When multiple processes are in contention for a shared resource, exponentially increasing jitter helps minimise collisions.

@retry(wait=wait_random_exponential(multiplier=1, max=60)) def wait_exponential_jitter(): print("Randomly wait up to 2^x * 1 seconds between each retry until the range reaches 60 seconds, then randomly up to 60 seconds afterwards") raise Exception

Sometimes it's necessary to build a chain of backoffs.

@retry(wait=wait_chain(*[wait_fixed(3) for i in range(3)] +

[wait_fixed(7) for i in range(2)] + [wait_fixed(9)]))

def wait_fixed_chained():

print("Wait 3s for 3 attempts, 7s for the next 2 attempts and 9s for all attempts thereafter") raise Exception

Whether to retry

We have a few options for dealing with retries that raise specific or general exceptions, as in the cases here.

@retry(retry=retry_if_exception_type(IOError)) def might_io_error(): print("Retry forever with no wait if an IOError occurs, raise any other errors") raise Exception

We can also use the result of the function to alter the behavior of retrying.

def is_none_p(value):

"""Return True if value is None""" return value is None

@retry(retry=retry_if_result(is_none_p)) def might_return_none(): print("Retry with no wait if return value is None")

We can also combine several conditions:

def is_none_p(value):

"""Return True if value is None""" return value is None

@retry(retry=(retry_if_result(is_none_p) | retry_if_exception_type())) def might_return_none(): print("Retry forever ignoring Exceptions with no wait if return value is None")

Any combination of stop, wait, etc. is also supported to give you the freedom to mix and match.

It's also possible to retry explicitly at any time by raising the TryAgain exception:

@retry def do_something(): result = something_else() if result == 23: raise TryAgain

Error Handling

While callables that "timeout" retrying raise a RetryError by default, we can reraise the last attempt's exception if needed:

@retry(reraise=True, stop=stop_after_attempt(3)) def raise_my_exception(): raise MyException("Fail")

try:

raise_my_exception()

except MyException:

# timed out retrying pass

Before and After Retry, and Logging

It's possible to execute an action before any attempt of calling the function by using the before callback function:

import logging

logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)

logger = logging.getLogger(__name__)

@retry(stop=stop_after_attempt(3), before=before_log(logger, logging.DEBUG)) def raise_my_exception(): raise MyException("Fail")

In the same spirit, It's possible to execute after a call that failed:

import logging

logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)

logger = logging.getLogger(__name__)

@retry(stop=stop_after_attempt(3), after=after_log(logger, logging.DEBUG)) def raise_my_exception(): raise MyException("Fail")

It's also possible to only log failures that are going to be retried. Normally retries happen after a wait interval, so the keyword argument is called before_sleep:

import logging

logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)

logger = logging.getLogger(__name__)

@retry(stop=stop_after_attempt(3),

before_sleep=before_sleep_log(logger, logging.DEBUG))

def raise_my_exception():

raise MyException("Fail")

Statistics

You can access the statistics about the retry made over a function by using the retry attribute attached to the function and its statistics attribute:

@retry(stop=stop_after_attempt(3)) def raise_my_exception(): raise MyException("Fail")

try:

raise_my_exception()

except Exception:

pass

print(raise_my_exception.retry.statistics)

...

Custom Callbacks

You can also define your own callbacks. The callback should accept one parameter called retry_state that contains all information about current retry invocation.

For example, you can call a custom callback function after all retries failed, without raising an exception (or you can re-raise or do anything really)

def return_last_value(retry_state):

"""return the result of the last call attempt""" return retry_state.outcome.result()

def is_false(value):

"""Return True if value is False""" return value is False

# will return False after trying 3 times to get a different result @retry(stop=stop_after_attempt(3), retry_error_callback=return_last_value, retry=retry_if_result(is_false)) def eventually_return_false(): return False

Note

Calling the parameter retry_state is important, because this is how tenacity internally distinguishes callbacks from their deprecated counterparts <deprecated-callbacks>.

RetryCallState

retry_state argument is an object of RetryCallState class:

tenacity.RetryCallState

Constant attributes:

start_time(float)

retry_object(BaseRetrying)

fn(callable)

args(tuple)

kwargs(dict)

Variable attributes:

attempt_number(int)

outcome(tenacity.Future or None)

outcome_timestamp(float or None)

idle_for(float)

next_action(tenacity.RetryAction or None)

Other Custom Callbacks

It's also possible to define custom callbacks for other keyword arguments.

my_stop(retry_state)

param RetryState retry_state

info about current retry invocation

return

whether or not retrying should stop

rtype

bool

my_wait(retry_state)

param RetryState retry_state

info about current retry invocation

return

number of seconds to wait before next retry

rtype

float

my_retry(retry_state)

param RetryState retry_state

info about current retry invocation

return

whether or not retrying should continue

rtype

bool

my_before(retry_state)

param RetryState retry_state

info about current retry invocation

my_after(retry_state)

param RetryState retry_state

info about current retry invocation

my_before_sleep(retry_state)

param RetryState retry_state

info about current retry invocation

Here's an example with a custom before_sleep function:

import logging

logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)

logger = logging.getLogger(__name__)

def my_before_sleep(retry_state):

if retry_state.attempt_number < 1:

loglevel = logging.INFO

else:

loglevel = logging.WARNING

logger.log(

loglevel, 'Retrying %s: attempt %s ended with: %s', retry_state.fn, retry_state.attempt_number, retry_state.outcome)

@retry(stop=stop_after_attempt(3), before_sleep=my_before_sleep) def raise_my_exception(): raise MyException("Fail")

try:

raise_my_exception()

except RetryError:

pass

Note

It was also possible to define custom callbacks before, but they accepted varying parameter sets and none of those provided full state. The old way is deprecated, but kept for backward compatibility.

my_stop(previous_attempt_number, delay_since_first_attempt)

deprecated

param previous_attempt_number

the number of current attempt

type previous_attempt_number

int

param delay_since_first_attempt

interval in seconds between the beginning of first attempt and current time

type delay_since_first_attempt

float

rtype

bool

my_wait(previous_attempt_number, delay_since_first_attempt [, last_result])

deprecated

param previous_attempt_number

the number of current attempt

type previous_attempt_number

int

param delay_since_first_attempt

interval in seconds between the beginning of first attempt and current time

type delay_since_first_attempt

float

param tenacity.Future last_result

current outcome

return

number of seconds to wait before next retry

rtype

float

my_retry(attempt)

deprecated

param tenacity.Future attempt

current outcome

return

whether or not retrying should continue

rtype

bool

my_before(func, trial_number)

deprecated

param callable func

function whose outcome is to be retried

param int trial_number

the number of current attempt

my_after(func, trial_number, trial_time_taken)

deprecated

param callable func

function whose outcome is to be retried

param int trial_number

the number of current attempt

param float trial_time_taken

interval in seconds between the beginning of first attempt and current time

my_before_sleep(func, sleep, last_result)

deprecated

param callable func

function whose outcome is to be retried

param float sleep

number of seconds to wait until next retry

param tenacity.Future last_result

current outcome

import logging

logging.basicConfig(stream=sys.stderr, level=logging.DEBUG)

logger = logging.getLogger(__name__)

def my_before_sleep(retry_object, sleep, last_result):

logger.warning(

'Retrying %s: last_result=%s, retrying in %s seconds...', retry_object.fn, last_result, sleep)

@retry(stop=stop_after_attempt(3), before_sleep=my_before_sleep) def raise_my_exception(): raise MyException("Fail")

try:

raise_my_exception()

except RetryError:

pass

Changing Arguments at Run Time

You can change the arguments of a retry decorator as needed when calling it by using the retry_with function attached to the wrapped function:

@retry(stop=stop_after_attempt(3)) def raise_my_exception(): raise MyException("Fail")

try:

raise_my_exception.retry_with(stop=stop_after_attempt(4))()

except Exception:

pass

print(raise_my_exception.retry.statistics)

...

If you want to use variables to set up the retry parameters, you don't have to use the retry decorator - you can instead use Retrying directly:

def never_good_enough(arg1):

raise Exception('Invalid argument: {}'.format(arg1))

def try_never_good_enough(max_attempts=3):

retryer = Retrying(stop=stop_after_attempt(max_attempts), reraise=True) retryer(never_good_enough, 'I really do try')

Async and retry

Finally, retry works also on asyncio and Tornado (>= 4.5) coroutines. Sleeps are done asynchronously too.

@retry
async def my_async_function(loop):
    await loop.getaddrinfo('8.8.8.8', 53)

@retry
@tornado.gen.coroutine
def my_async_function(http_client, url):
    yield http_client.fetch(url)

You can even use alternative event loops such as curio or Trio by passing the correct sleep function:

@retry(sleep=trio.sleep)
async def my_async_function(loop):
    await asks.get('https://example.org')

Contribute

1. Check for open issues or open a fresh issue to start a discussion around a feature idea or a bug.
2. Fork the repository on GitHub to start making your changes to the master branch (or branch off of it).
3. Write a test which shows that the bug was fixed or that the feature works as expected.
4. Add a changelog
5. Make the docs better (or more detailed, or more easier to read, or ...)

Changelogs

reno is used for managing changelogs. Take a look at their usage docs.

The doc generation will automatically compile the changelogs. You just need to add them.

# Opens a template file in an editor
tox -e reno -- new some-slug-for-my-change --edit