Exponential backoff?

[chuckhacker]

It's "good practices" to have exponential backoff for successive retries. Any plans to implement this?

[rubennorte]

We wrote this plugin just to retry network errors and we didn't expect to retry too many times, but that's a feature that could be added (together with #2).

Perhaps a retryInterval (that could be a number or a function that receives the request config) would be a good option to add. Would you consider creating a PR?

[chuckhacker]

Unfortunately, I realized a bit later that the environment that I am running my code in does not support JS promises (old version of SeaMonkey), sorry. This means no more axios for us.

Once I fix this, sure, but right now upgrading things is a "nice to have" for us.

[antony]

I'm also looking for exponential backoff, but even a simple way to control the retry timeout would be useful.

My suggestion is to have two configuration methods:

Simple config

retry: {
  attempts: 5,
  delay: 1000
}

or a function, to allow exponential backoff.

returning -1 means don't retry
returning 1000 * attempt means that on attempt 5, the retry delay is 5 seconds.

retry: (attempt) => {
  if (attempt > 5) { return -1 }
  return 1000 * attempt
}

[TheChuckster]

I'm back, working on this.

[s3rious]

Well, I've basically just implemented this in my fork: https://github.com/s3rious/axios-retry/blob/master/es/index.js#L116

Config now receives 3 arguments, I've added delay, which can be a simple number or a function which receives an attempt count as an argument and returns number.

If someone can help with tests and README.md I think I can make it a PR.

[towfiqi]

@s3rious Just replaced the es/index.js with your one and then tried like this but there was no delay axiosRetry(axios, { retries: 3,delay : 11000 });

[yipcma]

interested in exponential delay as well

[s3rious]

@towfiqi That's strange, have you rebuilt the lib after making changes?
I've just started a simple project with my fork, here is a gist, and delay works ok.

[gregorym]

[edit] Incorrect statement.

@s3rious Why don't you create a PR?

[matthiasfeist]

Is there any progress on that?

[rubennorte]

@moose56 is working on it in #28

[matthiasfeist]

Sorry I must have missed that there is a PR open... Thanks!

[dashmug]

This is great.