We’ve just started designing an akka-based message processing system and there is one aspect that is not fully clear from the documentation.
We have an Actor that effectively is a dynamically configured finite state machine. There are many sources for the messages for this Actor and for certain special states, it should postpone the processing of incoming messages until there is a special message that resets the state. For example, an incoming message may cause the Actor to request an updated configuration from an external system and in that case, the processing of any new messages should be postponed until the external system returns the results (or the timer runs out). Once the data from the external system arrives - the postponed messages should be processed before the new messages from the mailbox are processed.
There is a stashing mechanism described in the documentation, but it is not entirely clear to me if the approach is actually good for what we are trying to do.
Any recommendations on how to properly design an actor to achieve this postponed message processing?
Without diving deeper into your problem it sounds like stashing or an internal buffer in the actor makes sense.
In the Akka Typed actor API the stash looks much like an internal message buffer but with some extra API to process its contents (unstashAll and friends): https://doc.akka.io/docs/akka/current/typed/stash.html#stash