backoff

What is backoff?

The 'backoff' npm package provides a mechanism to handle retry logic with exponential backoff and other backoff strategies. It is useful for scenarios where you need to retry operations that may fail intermittently, such as network requests or database queries.

What are backoff's main functionalities?

Exponential Backoff

This feature allows you to implement an exponential backoff strategy using the Fibonacci sequence. The code sample demonstrates how to set up a Fibonacci backoff with a randomisation factor, initial delay, and maximum delay. It also shows how to handle the 'ready' and 'fail' events.

const backoff = require('backoff');

const fibonacciBackoff = backoff.fibonacci({ randomisationFactor: 0.2, initialDelay: 10, maxDelay: 1000 });

fibonacciBackoff.on('ready', function(number, delay) {
  console.log(`Retrying after ${delay} ms`);
  // Perform the operation that needs to be retried
});

fibonacciBackoff.on('fail', function() {
  console.log('Operation failed after maximum retries');
});

// Start the backoff process
fibonacciBackoff.backoff();

Retry with Custom Strategy

This feature allows you to implement a custom retry strategy using exponential backoff. The code sample demonstrates how to set up an exponential backoff with an initial delay and maximum delay. It also shows how to handle the 'ready' and 'fail' events.

const backoff = require('backoff');

const customStrategy = backoff.exponential({ initialDelay: 100, maxDelay: 10000 });

customStrategy.on('ready', function(number, delay) {
  console.log(`Retry attempt #${number} after ${delay} ms`);
  // Perform the operation that needs to be retried
});

customStrategy.on('fail', function() {
  console.log('Operation failed after maximum retries');
});

// Start the backoff process
customStrategy.backoff();

Handling Errors

This feature demonstrates how to handle errors during the retry process. The code sample shows how to simulate an operation that may fail and how to retry the operation using exponential backoff until it succeeds or reaches the maximum number of retries.

const backoff = require('backoff');

const exponentialBackoff = backoff.exponential({ initialDelay: 100, maxDelay: 10000 });

exponentialBackoff.on('ready', function(number, delay) {
  console.log(`Retry attempt #${number} after ${delay} ms`);
  // Simulate an operation that may fail
  const success = Math.random() > 0.5;
  if (!success) {
    exponentialBackoff.backoff();
  } else {
    console.log('Operation succeeded');
  }
});

exponentialBackoff.on('fail', function() {
  console.log('Operation failed after maximum retries');
});

// Start the backoff process
exponentialBackoff.backoff();

Other packages similar to backoff

retry

The 'retry' package provides a similar functionality to 'backoff' by allowing you to retry operations with configurable strategies. It supports exponential backoff, custom retry strategies, and error handling. Compared to 'backoff', 'retry' offers a more flexible API for defining custom retry logic.

promise-retry

The 'promise-retry' package is designed for retrying operations that return promises. It provides a simple API for retrying promise-based operations with configurable retry strategies, including exponential backoff. Compared to 'backoff', 'promise-retry' is more focused on promise-based workflows.

async-retry

The 'async-retry' package is another alternative that supports retrying asynchronous operations with configurable strategies. It works well with both promises and async/await syntax. Compared to 'backoff', 'async-retry' offers a more modern API that integrates seamlessly with async/await.

README

Backoff for Node.js

Fibonacci and exponential backoffs for Node.js.

Installation

npm install backoff

Unit tests

npm test

Usage

Object Oriented

The usual way to instantiate a new Backoff object is to use one predefined factory method: backoff.fibonacci([options]), backoff.exponential([options]).

Backoff inherits from EventEmitter. When a backoff starts, a backoff event is emitted and, when a backoff ends, a ready event is emitted. Handlers for these two events are called with the current backoff number and delay.

var backoff = require('backoff');

var fibonacciBackoff = backoff.fibonacci({
    randomisationFactor: 0,
    initialDelay: 10,
    maxDelay: 300
});

fibonacciBackoff.failAfter(10);

fibonacciBackoff.on('backoff', function(number, delay) {
    // Do something when backoff starts, e.g. show to the
    // user the delay before next reconnection attempt.
    console.log(number + ' ' + delay + 'ms');
});

fibonacciBackoff.on('ready', function(number, delay) {
    // Do something when backoff ends, e.g. retry a failed
    // operation (DNS lookup, API call, etc.). If it fails
    // again then backoff, otherwise reset the backoff
    // instance.
    fibonacciBackoff.backoff();
});

fibonacciBackoff.on('fail', function() {
    // Do something when the maximum number of backoffs is
    // reached, e.g. ask the user to check its connection.
    console.log('fail');
});

fibonacciBackoff.backoff();

The previous example would print the following.

0 10ms
1 10ms
2 20ms
3 30ms
4 50ms
5 80ms
6 130ms
7 210ms
8 300ms
9 300ms
fail

Note that Backoff objects are meant to be instantiated once and reused several times by calling reset after a successful "retry".

Functional

It's also possible to avoid some boilerplate code when invoking an asynchronous function in a backoff loop by using backoff.call(fn, [args, ...], callback).

Typical usage looks like the following.

var call = backoff.call(get, 'https://duplika.ca/', function(err, res) {
    console.log('Retries: ' + call.getResults().length);

    if (err) {
        console.log('Error: ' + err.message);
    } else {
        console.log('Status: ' + res.statusCode);
    }
});

call.setStrategy(new backoff.ExponentialStrategy());
call.failAfter(10);

API

backoff.fibonacci([options])

Constructs a Fibonacci backoff (10, 10, 20, 30, 50, etc.).

See bellow for options description.

backoff.exponential([options])

Constructs an exponential backoff (10, 20, 40, 80, etc.).

The options are the following.

• randomisationFactor: defaults to 0, must be between 0 and 1
• initialDelay: defaults to 100 ms
• maxDelay: defaults to 10000 ms

With these values, the backoff delay will increase from 100 ms to 10000 ms. The randomisation factor controls the range of randomness and must be between 0 and 1. By default, no randomisation is applied on the backoff delay.

backoff.call(fn, [args, ...], callback)

• fn: function to call in a backoff handler
• args: function's arguments
• callback: function's callback accepting an error as its first argument

Calls an asynchronous function in a backoff handler so that it gets automatically retried on error. The wrapped function will get retried until it succeds or reaches the maximum number of backoffs. In both cases, the callback function will be invoked with the last result returned by the wrapped function.

This function returns a FunctionCall instance that is going to be invoked on next tick and can be used to configure and/or abort the call.

Class Backoff

new Backoff(strategy)

• strategy: the backoff strategy to use

Constructs a new backoff object from a specific backoff strategy. The backoff strategy must implement the BackoffStrategyinterface defined bellow.

backoff.failAfter(numberOfBackoffs)

• numberOfBackoffs: maximum number of backoffs before the fail event gets emitted, must be greater than 0

Sets a limit on the maximum number of backoffs that can be performed before a fail event gets emitted and the backoff instance is reset. By default, there is no limit on the number of backoffs that can be performed.

backoff.backoff()

Starts a backoff operation. Will throw an error if a backoff operation is already in progress.

In practice, this method should be called after a failed attempt to perform a sensitive operation (connecting to a database, downloading a resource over the network, etc.).

backoff.reset()

Resets the backoff delay to the initial backoff delay and stop any backoff operation in progress. After reset, a backoff instance can and should be reused.

In practice, this method should be called after having successfully completed the sensitive operation guarded by the backoff instance or if the client code request to stop any reconnection attempt.

Event: 'backoff'

• number: number of backoffs since last reset, starting at 0
• delay: backoff delay in milliseconds

Emitted when a backoff operation is started. Signals to the client how long the next backoff delay will be.

Event: 'ready'

• number: number of backoffs since last reset, starting at 0
• delay: backoff delay in milliseconds

Emitted when a backoff operation is done. Signals that the failing operation should be retried.

Event: 'fail'

Emitted when the maximum number of backoffs is reached. This event will only be emitted if the client has set a limit on the number of backoffs by calling backoff.failAfter(numberOfBackoffs). The backoff instance is automatically reset after this event is emitted.

Interface BackoffStrategy

A backoff strategy must provide the following methods.

strategy.next()

Computes and returns the next backoff delay.

strategy.reset()

Resets the backoff delay to its initial value.

Class ExponentialStrategy

Exponential (10, 20, 40, 80, etc.) backoff strategy implementation.

new ExponentialStrategy([options])

The options are the following.

• randomisationFactor: defaults to 0, must be between 0 and 1
• initialDelay: defaults to 100 ms
• maxDelay: defaults to 10000 ms

Class FibonacciStrategy

Fibonnaci (10, 10, 20, 30, 50, etc.) backoff strategy implementation.

new FibonacciStrategy([options])

The options are the following.

• randomisationFactor: defaults to 0, must be between 0 and 1
• initialDelay: defaults to 100 ms
• maxDelay: defaults to 10000 ms

Class FunctionCall

This class manages the calling of an asynchronous function within a backoff loop.

This class should rarely be instantiated directly since the factory method backoff.call(fn, [args, ...], callback) offers a more convenient and safer way to create FunctionCall instances.

new FunctionCall(fn, args, callback)

• fn: asynchronous function to call
• args: an array containing fn's args
• callback: fn's callback

Constructs a function handler for the given asynchronous function.

call.setStrategy(strategy)

• strategy: strategy instance to use, defaults to FibonacciStrategy.

Sets the backoff strategy to use. This method should be called before call.call().

call.failAfter(maxNumberOfBackoffs)

• maxNumberOfBackoffs: maximum number of backoffs before the call is aborted

Sets the maximum number of backoffs before the call is aborted. This method should be called before call.call().

call.getResults()

Retrieves all intermediary results returned by the wrapped function. This method can be called at any point in time during the call life cycle, i.e. before, during and after the wrapped function invocation.

Returns an array of arrays containing the results returned by the wrapped function for each call. For example, to get the error code returned by the second call, one would do the following.

var results = call.getResults();
var error = results[1][0];

call.call()

Calls the wrapped function. This method should only be called once.

call.abort()

Aborts the call.

Past results can be retrieved using call.getResults(). This method can be called at any point in time during the call life cycle, i.e. before, during and after the wrapped function invocation.

Event: 'call'

• args: wrapped function's arguments

Emitted each time the wrapped function is called.

Event: 'callback'

• results: wrapped function's return values

Emitted each time the wrapped function invokes its callback.

Event: 'backoff'

• number: backoff number, starts at 0
• delay: backoff delay in milliseconds

Emitted each time a backoff operation is started.

License

This code is free to use under the terms of the MIT license.

Changelog

1.2.1

• Make FunctionCall.backoffFactory a private member.