How to implement retry with exponential backoff?

[tschuchortdev]

In the documentation it says that stateful retry strategies like exponential backoff can be implemented by persisting the retry state:

Use custom persistent state to track retry metadata and query it in the retry handler

The retry function allows me to query the State but it does not allow me to write to it since the lambda function can only return an IO[Option[Instant]]:

val withRetryFull = doSomethingFull
  .retry { (error, state, now) =>
    error match {
      case _: TimeoutException => IO.pure(Some(now.plus(Duration.ofMinutes(2))))
      case _                   => IO.pure(None) // Don't retry other errors
    }
  }

or does this sentence in the documentation mean that I should persist the retry state elsewhere (in a database for example)? That seems like something the workflow library should handle for me.

I know I could also implement exponential backoff using a WIO.repeat and WIO.await but that's a lot of work, a lot of additional code and I'm only retrying technical errors where I don't care if the backoff resets if the workflow instance is lost and recovered.

[Krever]

or does this sentence in the documentation mean that I should persist the retry state elsewhere (in a database for example)?

Unfortunately yes, current support for retries is quite minimal, but you can manage the retry state outside the workflow (e.g. in db table) and access it within IO.

That seems like something the workflow library should handle for me.

I totally agree, its just not implemented (yet).

Adding state to retries should be relatively simple to implement. For now I created #163 to express the desire, its definitely a feature I'd like to have.
I cant say when I will find some time to work on it but I'm more than happy to guide volunteers through the implementation.