This very simple library provides a function for retrying an asynchronous operation until it succeeds. An "asynchronous operation" is embodied by a function that returns a promise or returns synchronously.

It supports regular intervals and exponential backoff with a configurable limit, as well as an overall timeout for the operation that limits the number of retries.

The bluebird library supplies the promise implementation.

var Promise = require('bluebird');
var retry = require('bluebird-retry');

var count = 0;
function myfunc() {
    console.log('myfunc called ' + (++count) + ' times');
    if (count < 3) {
        return Promise.reject(new Error('fail the first two times'));
    } else {
        return Promise.resolve('succeed the third time');
    }
}

retry(myfunc).done(function(result) {
    console.log(result);
});

This will display:

myfunc called 1 times
myfunc called 2 times
myfunc called 3 times
succeed the third time

The function is executed by Promise.attempt, so it can return a simple value or a Promise that resolves successfully to indicate success, or it can throw an Error or a rejected promise to indicate failure.

Note that the rejection messages from the first two failed calls were absorbed by retry.

The maximum number of retries and controls for the interval between retries can be specified via the options parameter:

• interval initial wait time between attempts in milliseconds (default 1000)
• backoff if specified, increase interval by this factor between attempts
• max_interval if specified, maximum amount that interval can increase to
• timeout total time to wait for the operation to succeed in milliseconds
• max_tries maximum number of attempts to try the operation (default 5)
• predicate to be used as bluebird's Filtered Catch. func will be retried only if the predicate expectation is met, it will otherwise fail immediately.

Note that timeout does not actually set a real timeout for the operation, but actually computes a maximum number of attempts based on the interval options. If both timeout and max_tries are specified, then whichever limit comes first applies. If max_tries is set to -1 and no timeout is specified, retry will be performed forever.

For example:

function logFail() {
    console.log(new Date().toISOString());
    throw new Error('bail');
}

retry(logFail, { max_tries: 4, interval: 500 });

Will display:

2014-05-29T23:16:28.941Z
2014-05-29T23:16:29.445Z
2014-05-29T23:16:29.946Z
2014-05-29T23:16:30.447Z
Error: operation timed out

And

retry(logFail, { timeout: 10000, interval: 1000, backoff: 2 });

Will display:

2014-05-29T23:17:29.655Z
2014-05-29T23:17:30.658Z
2014-05-29T23:17:32.660Z
2014-05-29T23:17:36.661Z
Error: operation timed out

The library also supports stopping the retry loop before the timeout occurs by throwing a new instance of retry.StopError from within the called function.

For example:

var retry = require('bluebird-retry');
var i = 0;
var err;
var swing = function() {
    i++;
    console.log('strike ' + i);
    if (i == 3) {
        throw new retry.StopError('yer out');
    }
    throw new Error('still up at bat');
};

retry(swing, {timeout: 10000})
.caught(function(e) {
    console.log(e.message)
});

Will display:

strike 1
strike 2
strike 3
yer out

The StopError constructor accepts one argument. If it is invoked with an instance of Error, then the promise is rejected with that error argument. Otherwise the promise is rejected with the StopError itself.