request-retry - HTTP(s) request retry on recoverable errors.
When the connection fails with one of ECONNRESET
, ENOTFOUND
, ESOCKETTIMEDOUT
, ETIMEDOUT
, ECONNREFUSED
, EHOSTUNREACH
, EPIPE
, EAI_AGAIN
or when an HTTP 5xx or 429 error occurrs, the request will automatically be re-attempted as these are often recoverable errors and will go away on retry.
❤️ Shameless plug
- Charts, simple as a URL. No more server-side rendering pain, 1 url = 1 chart
- Looking for a free Redis GUI? Or for real-time alerting & monitoring for Redis?
- Mailpopin - Stripe payment emails you can actually use
Installation
Install with npm.
npm install --save requestretry
Usage
Request-retry is a drop-in replacement for request but adds two new options maxAttempts
and retryDelay
. It also adds one property to the response (or the error object, upon a network error), attempts
. It supports callbacks or promises.
With callbacks
var request = require('requestretry');
request({
url: 'https://api.domain.com/v1/a/b',
json: true,
// The below parameters are specific to request-retry
maxAttempts: 5, // (default) try 5 times
retryDelay: 5000, // (default) wait for 5s before trying again
retryStrategy: request.RetryStrategies.HTTPOrNetworkError // (default) retry on 5xx or network errors
}, function(err, response, body){
// this callback will only be called when the request succeeded or after maxAttempts or on error
if (response) {
console.log('The number of request attempts: ' + response.attempts);
}
});
With promises
When you're using promises, you can pass the two following options:
-
fullResponse
(default true) - To resolve the promise with the full response or just the body -
promiseFactory
(default whenjs) - A function to allow the usage of a different promise implementation library
request({
url: 'https://api.domain.com/v1/a/b',
json: true,
fullResponse: true // (default) To resolve the promise with the full response or just the body
})
.then(function (response) {
// response = The full response object or just the body
})
.catch(function(error) {
// error = Any occurred error
})
Using promiseFactory
option to use a different promise implementation library
// See the tests for different libraries usage examples
/**
* @param {Function} resolver The promise resolver function
* @return {Object} The promise instance
*/
function customPromiseFactory(resolver) {
// With when.js
return require('when').promise(resolver);
// With RSVP.js
var Promise = require('rsvp').Promise;
return new Promise(resolver);
}
request({
url: 'https://api.domain.com/v1/a/b',
json: true,
// Custom promise factory function
promiseFactory: customPromiseFactory
})
.then(function (response) {
// response = The full response object or just the body
})
.catch(function(error) {
// error = Any occurred error
})
How to define your own retry strategy
A retry strategy let you specify when request-retry should retry a request
/**
* @param {Null | Object} err
* @param {Object} response
* @param {Object} body
* @param {Object} options copy
* @return {Boolean} true if the request should be retried
*/
function myRetryStrategy(err, response, body, options){
// retry the request if we had an error or if the response was a 'Bad Gateway'
return !!err || response.statusCode === 502;
}
/**
* @param {Null | Object} err
* @param {Object} response
* @param {Object} body
* @param {Object} options copy
* @return {Object} mustRetry: {Boolean} true if the request should be retried
* options: {Object} new options for request
*/
function myRetryStrategy(err, response, body, options){
options.url = 'new url'; //you can overwrite some attributes or create new object
return {
mustRetry: !!err || response.statusCode === 502,
options: options, //then it should be passed back, it will be used for new requests
}
}
/**
* With an asynchronous retry strategy
* @param {Null | Object} err
* @param {Object} response
* @param {Object} body
* @param {Object} options copy
* @return {Object} mustRetry: {Boolean} true if the request should be retried
* options: {Object} new options for request
*/
async function myRetryStrategy(err, response, body, options){
let token = await getNewApiAuthToken();
options.headers = {'Authorization': `Bearer ${token}`}
return {
mustRetry: true,
options: options, // retry with new auth token
}
}
request({
url: 'https://api.domain.com/v1/a/b'
json:true,
retryStrategy: myRetryStrategy
}, function(err, response, body){
// this callback will only be called when the request succeeded or after maxAttempts or on error
});
How to define your own delay strategy
A delay strategy let you specify how long request-retry should wait before trying again the request
/**
* @param {Null | Object} err
* @param {Object} response
* @param {Object} body
* @return {Number} number of milliseconds to wait before trying again the request
*/
function myDelayStrategy(err, response, body){
// set delay of retry to a random number between 500 and 3500 ms
return Math.floor(Math.random() * (3500 - 500 + 1) + 500);
}
request({
url: 'https://api.domain.com/v1/a/b'
json:true,
delayStrategy: myDelayStrategy // delayStrategy is called 1 less times than the maxAttempts set
}, function(err, response, body){
// this callback will only be called when the request succeeded or after maxAttempts or on error
});
Here is how to implement an exponential backoff strategy:
/**
* @param {Number} attempts The number of times that the request has been attempted.
* @return {Number} number of milliseconds to wait before retrying again the request.
*/
function getExponentialBackoff(attempts) {
return (Math.pow(2, attempts) * 100) + Math.floor(Math.random() * 50);
}
function constructExponentialBackoffStrategy() {
let attempts = 0;
return () => {
attempts += 1;
return getExponentialBackoff(attempts);
};
}
request({
url: 'https://api.domain.com/v1/a/b'
json:true,
delayStrategy: constructExponentialBackoffStrategy() // need to invoke the function to return the closure.
}, function(err, response, body){
// this callback will only be called when the request succeeded or after maxAttempts or on error
});
How to access the underlying request library
You can access to the underlying request
library thanks to request.Request
:
const request = require('requestretry');
console.log(request.Request); // original request library
Thus, if needed, it's possible to monkey-patch or extend the underlying Request library:
request.Request = class extends request.Request {
constructor(url, options, f, retryConfig) {
super(url, options, f, retryConfig);
// this constructor will be called for every requestretry call,
// and give you global logging
console.log('Request', url, options, f, retryConfig);
}
}
request
options
Modifying You can use the defaults
method to provide default options like so:
var request = require('requestretry').defaults({ json: true, retryStrategy: myRetryStrategy });
API surface
As with request
, several helpers are provided for various HTTP methods and usage:
-
request(options [, callback])
. -
request(url [, callback])
- same asrequest(options [, callback])
. -
request(url, options [, callback])
- same asrequest(options [, callback])
. -
request.get(url [, callback])
- same asrequest(options [, callback])
, defaultsoptions.method
toGET
. -
request.get(url, options [, callback])
- same asrequest(options [, callback])
, defaultsoptions.method
toGET
. -
request.head(url)
- same asrequest(options [, callback])
, defaultsoptions.method
toHEAD
. -
request.post(url)
- same asrequest(options [, callback])
, defaultsoptions.method
toPOST
. -
request.put(url)
- same asrequest(options [, callback])
, defaultsoptions.method
toPUT
. -
request.patch(url)
- same asrequest(options [, callback])
, defaultsoptions.method
toPATCH
. -
request.del(url)
- same asrequest(options [, callback])
, defaultsoptions.method
toDELETE
. -
request.delete(url)
- same asrequest(options [, callback])
, defaultsoptions.method
toDELETE
.
Changelog
You want to support my work?
I maintain this project in my free time, if it helped you, well, I would be grateful to buy a beer thanks to your paypal or Bitcoins, donation!