Actor to actor calls
The ability for one actor to call another actor is critical to being able to create composable actor systems. wasmCloud supports RPC-style communication between actors, even if those actors are running in hosts scattered across disparate infrastructure, connected only via the lattice.
The single most important task when calling other actors is to identify the target actor. wasmCloud supports two ways of identifying another actor for invocation purposes:
- Public Key (Subject)
- Call Alias
Public key (Subject)
Every single actor must have a subject, which is a 56-character public key made of up uppercase letters. These subject strings always start with the capital letter M (module). When actors are signed with embedded capability claims, they are always required to have a subject claim. This means that invocation by public key is the most reliable means for locating target actors, though it can often impose some amount of friction on the part of calling actors.
You must be cautious when hard-coding public keys in your actors. If your software pipeline involves signing actors with different private keys across different environments, then that will change the public keys of those actors and result in target actors not being found during invocations at runtime.
One potentially appealing alternative to the guaranteed uniqueness and reliability of public keys is the use of a call alias. When actors are signed, they can optionally be signed with a
call_alias claim. This alias must be an alphanumeric string that can optionally have separator/hierarchy characters like
When an actor with a call alias is started, the entire lattice in which that actor resides will be informed of the claim on this alias. If the alias is already claimed, then the newly started actor will not be able to use it for identification purposes. This is the downside to using aliases: it is up to you to ensure that they are unique per lattice. They do not need to be globally unique.
Call aliases, when used appropriately, can provide consistent, developer-friendly naming conventions for locating target actors and often serve a similar purpose as the concept of "discovery servers" or "discovery services" in more traditional microservice environments.
Using the Actor Core API
All actors in the wasmCloud ecosystem are capable of making calls to other actors. The wasmCloud host runtime will, by default, allow any two actors to communicate with each other.
The following code snippet illustrates invoking a
ping method (in other words, sending a ping message) to an actor that implements a Ponger interface:
// import the interface library
// obtain the public key of the other actor ("M...")
let actor_id = get_other_actor_key();
// obtain an interface handle for sending message to a Ponger actor
let ponger = PongerSender::for_actor(&actor_id)?;
// call the actor and wait for the result
let result = ponger.ping().await?;
When you design your distributed application to allow actors to communicate with each other, you need some way to declare the shape or schema of the data they will use to communicate. This can easily be done using smithy models that declare only "actor receive" services. One actor can then invoke the "actor receive" function on another actor, enabling RPC between actors.