I'm pretty sure this is supposed to be exponential

[wparad]

But here is it multiplicative. Reason and experimentation suggests the formula for the Nth time is:

delay wait = interval Seconds * (backoff rate) ^ (attempt - 1)

But we would expect for the 4 attempts to be

> 30 * 5 ** 0
30
> 30 * 5 ** 1
150
> 30 * 5 ** 2
750
> 30 * 5 ** 3
3750

See similar here: https://exponentialbackoffcalculator.com/

[JesterXL]

@wparad Thanks, I'll take a look.

[wparad]

fwiw, I spent a really long time on not finding any good documented cases on how this works, so it might be valuable to actually run a state machine with the exact configuration to see what the times end up being.

[JesterXL]

Good idea, I'll create some tests with results to see.

[sacredwolf7]

The logic should be like:
(Interval)* Math.pow(backOff, attempt);

[cameroncf]

I just spent more time than I'd like to matching up math to actual results in a Step Function that failed a few times and here is the logic that I came up with:

previousTime + ((Interval)* Math.pow(backOff, attempt));

I validate that it matches results where:

backoffRate: 3,
interval: Duration.seconds(10),
maxAttempts: 5,

Which yields this result:

0+10*(3^0) = 10
10+(10*(3^1)) = 40
40+(10*(3^2)) = 130
130+(10*(3^3)) = 400
400+(10*(3^4)) = 1210

[cameroncf]

@JesterXL I forked the repo and attempted to make this change but I don't know enough Elm to do it quickly. Would love to help get this calculator working though. It's really the only thing I've found via Google that can do this sort of calculation (even if it's a bit off right now).

[cameroncf]

I'm knee deep in Step Functions this week so I'm going to add more test data here so we know what the math and results should looks like.

I built a test function that will intentionally fail since it's trying to hit a DynamoDB table that doesn't exist.

Ran it, and got the following times:

The right math is Interval * (backOff ^ retryCount).

PR inbound shortly.