Retries do not support backoff

[TimFerrell]

When calling a service with the retry option, it will immediately make a retry request after the first failure:

grpc-caller/lib/client.js

Lines 47 to 64 in 549484c

	if (_.has(options, 'retry')) {
	const retryOpts = options.retry
	const callOpts = options ? _.omit(options, 'retry') : options
	if (_.isFunction(fn)) {
	async.retry(retryOpts, rCb => {
	v.call(this.client, arg, metadata, callOpts, rCb)
	}, fn)
	} else {
	return new Promise((resolve, reject) => {
	async.retry(retryOpts, rCb => {
	v.call(this.client, arg, metadata, callOpts, rCb)
	}, (err, res) => {
	if (err) reject(err)
	else resolve(res)
	})
	})
	}

General guidance is to implement backoffs when retrying to prevent overloading the server.

The suggestion is to support a function that returns the amount of time to delay before attempting the retry. This approach would support different approaches to backoffs (constant value, linear, exponential, jitter, no jitter).

Something similar to this:

/**
 * @param {number} requestAttempts Amount of times that the request has been made.
 */
function backoff(requestAttempts) {
  return (requestAttempts * 20);
}

[bojand]

The retry option is just async.retry(). So you could do:

const res = await client.sayHello({ name: 'Bob' }, {}, { retry: { times: 3, interval: 200 } })
console.log(res)

and that will do retries with 200 milliseconds between retry intervals.

You can also indeed use a function as well:

const res = await client.sayHello({
  name: 'Bob'
}, {}, {
  retry: {
    times: 10,
    interval: function(retryCount) {
      return 50 * Math.pow(2, retryCount);
    }
  }
}) 

console.log(res)

Is this not working for you?

[TimFerrell]

That works! Thanks for the reply, and my apologies for not catching this myself.