Source code for sertit.network

# -*- coding: utf-8 -*-
# Copyright 2021, SERTIT-ICube - France, https://sertit.unistra.fr/
# This file is part of sertit-utils project
#     https://github.com/sertit/sertit-utils
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
""" Network control utils """

import logging
import os
import time
from random import Random
from typing import Callable

logger = logging.getLogger(__name__)


# pylint: disable=R0913
# Too many arguments (7/5) (too-many-arguments)
[docs]def exponential_backoff( network_request: Callable, wait_time_slot: float, increase_factor: float, max_wait: float, max_retries: int, desc: str, random_state=None, ) -> None: """ Implementation of the standard Exponential Backoff algorithm (https://en.wikipedia.org/wiki/Exponential_backoff) This algorithm is useful in networking setups where one has to use a potentially unreliable resource over the network, where by unreliable we mean not always available. Every major service provided over the network can't guarantee 100% availability all the time. Therefore if a service fails one should retry a certain number of time and a maximum amount of times. This algorithm is designed to try using the services multiple times with exponentially increasing delays between the tries. The time delays are chosen at random to ensure better theoretical behaviour. No default value is provided on purpose Args: network_request (Callable): Python function taking no arguments which represents a network request which may fail. If it fails it must raise an exception. wait_time_slot (float): Smallest amount of time to wait between retries. Recommended range [0.1 - 10.0] seconds increase_factor (float): Exponent of the expected exponential increase in waiting time. In other words on average by how much should the delay in time increase. Recommended range [1.5 - 3.0] max_wait (float): Maximum of time one will wait for the network request to perform successfully. If the maximum amount of time is reached a timeout exception is thrown. max_retries (int): Number of total tries to perform the network request. Must be at least 2 and maximum `EXP_BACK_OFF_ABS_MAX_RETRIES` (or 100 if the environment value is not defined). Recommended range [5 - 25]. If the value exceeds `EXP_BACK_OFF_ABS_MAX_RETRIES` (or 100 if the environment value is not defined). The value will be set to `EXP_BACK_OFF_ABS_MAX_RETRIES` (or 100 if the environment value is not defined). desc (str): Description of the network request being attempted random_state (int): Seed to the random number generator (optional) """ abs_max_tries = int(os.environ.get("EXP_BACK_OFF_ABS_MAX_RETRIES", "100")) if abs_max_tries <= 2: raise Exception( "Environment variable 'EXP_BACK_OFF_ABS_MAX_RETRIES' must be positive finite integer greater " "than 2" ) if not float(wait_time_slot) == wait_time_slot or wait_time_slot <= 0.0: raise TypeError( f"Variable 'wait_time_slot' " f"(current value {wait_time_slot}) must be float strictly greater than 0.0" ) if not float(increase_factor) == increase_factor or increase_factor <= 0.0: raise TypeError( f"Variable 'increase_factor' (current value {increase_factor}) must be float strictly greater " f"than 0.0" ) if not float(max_wait) == max_wait or max_wait <= 0.0: raise TypeError( f"Variable 'max_wait' (current value {max_wait}) must be float strictly greater " f"than 0.0" ) if not int(max_retries) == max_retries or max_retries <= 2: raise TypeError( f"Variable 'max_retries' (current value {max_retries}) must be integer strictly greater " f"than 2" ) if random_state is not None and not int(random_state) == random_state: raise TypeError( f"Variable 'random_state' (current value {random_state}) must be 'None' or integer value" ) if desc == "": raise TypeError( f"Variable 'desc' (current value {desc}) must be non empty string" ) real_max_tries = min(max_retries, abs_max_tries) rng = Random(random_state) cumulated_wait_time = 0.0 # First try with no wait # pylint: disable=W0703 # W0703: Catching too general exception Exception (broad-except) try: return network_request() except Exception as ex: logger.error("Action '%s' failed with exception: %s", desc, ex, exc_info=True) for i in range(real_max_tries - 2): # Avoids infinite loop random_scale = rng.randint(0, 2 ** i - 1) curr_wait_time = wait_time_slot * increase_factor ** random_scale logger.info("Retrying action %s in %s seconds...", desc, curr_wait_time) cumulated_wait_time += curr_wait_time if cumulated_wait_time > max_wait: logger.error( "Waited %s seconds : Maximum wait time (%s) reached\n !!! Aborting !!!", cumulated_wait_time, max_wait, ) raise TimeoutError( f"About to exceed maximum wait time {max_wait} for action {desc}" ) time.sleep(curr_wait_time) # pylint: disable=W0703 # W0703: Catching too general exception Exception (broad-except) try: return network_request() except Exception as ex: logger.error( "Action '%s' failed with exception: %s", desc, ex, exc_info=True ) # Final blind try and hope for the best return network_request()