Services advertisement and discovery
Overview
The example shows one of the important Registry use-cases — services advertisement and discovery. So it is about advertisement by providers, discovery and usage by clients with the same Aqua code transparently without any knowledge about particular peer and service ids.
In the beginning, we will deploy a Rust echo service and call it with the Fluence CLI. Then we will start a JS/TS client with echo service and run it.
Finally, we will use Registry to call different types of services with exactly one piece of code.
Requirements
node: >= 16
rust: rustc 1.63.0-nightly
@fluencelabs/cli: 0.2.7
Set up environment
-
Install
fluence
cli:npm i -g @fluencelabs/cli@0.2.7
-
Initialize Fluence project:
fluence init .
Output:
Successfully initialized Fluence project template at <your-path-to-registry-repo>/example
-
You can use VSCode with Aqua extension for syntax highlighting and better developer experience.
Add echo service in Rust
-
If you have Apple Silicon you should install cargo manually, or you can ignore this step otherwise:
# for m1: cargo install marine --version '0.12.1' --root ~/.fluence/cargo
-
Add echo service:
fluence service add echoService
output:
$ fluence service add echoService Added echoService to fluence.yaml
-
Following code returns a peer id of the host and an initial message:
#[marine] pub fn echo(msg: String) -> String { format!("{}: {}", marine_rs_sdk::get_call_parameters().host_id, msg) }
Test echo service locally with REPL
-
If you have Apple Silicon you should install mrepl manually, or you can ignore this step otherwise:
# for m1: cargo install mrepl --version '0.18.0' --root ~/.fluence/cargo
-
Run the following command to start REPL:
fluence service repl echoService
output:
$ fluence service repl echoService Making sure service and modules are downloaded and built... done Welcome to the Marine REPL (version 0.18.0) Minimal supported versions sdk: 0.6.0 interface-types: 0.20.0 app service was created with service id = 070fd906-a9e4-44af-963d-81a75526379c elapsed time 55.397166ms 1>
-
Call
echo
function ofechoService
module and passmsg
to it:call echoService echo ["msg"]
output:
1> call echoService echo ["msg"] result: String(": msg") elapsed time: 9.207542ms
Peer id is empty because that is default for Marine REPL.
You can always test your services before deployment with REPL. Check out documentation for more details.
Deploy service and test it on remote peer
-
Deploy service:
fluence deploy
Output, please note that peer ids and service ids may differ:
$ fluence deploy Making sure all services are downloaded... done Making sure all modules are downloaded and built... done Going to deploy services described in ~/Documents/dev/fluencelabs/registry/example/fluence.yaml: echoService: get: echoService deploy: - deployId: default ? Do you want to deploy all of these services? Yes Deploying: service: echoService deployId: default on: 12D3KooWCMr9mU894i8JXAFqpgoFtx6qnV1LFPSfVc3Y34N4h4LS ... done Compiling ~/Documents/dev/fluencelabs/registry/example/.fluence/aqua/deployed.app.aqua... done Currently deployed services listed in ~/Documents/dev/fluencelabs/registry/example/.fluence/app.yaml: echoService: default: - blueprintId: fd21793bae96d1519599346591bfe7ea75614d9fe7f7c824b279e5a2dc841ee0 serviceId: 785487e4-de08-4fa4-bdfd-df981ec66da7 peerId: 12D3KooWCMr9mU894i8JXAFqpgoFtx6qnV1LFPSfVc3Y34N4h4LS
-
The following code declares a module, imports App and EchoService, and defines a method to call
echo
:module Main import App from "deployed.app.aqua" import EchoService from "services/echoService.aqua" export App, echo func echo(msg: string) -> string: services <- App.services() on services.echoService.default[0].peerId: EchoService services.echoService.default[0].serviceId res <- EchoService.echo(msg) <- res
-
Let’s test our echo service:
fluence run -f 'echo("hi")'
Output:
$ fluence run -f 'echo("hi")' Running: function: echo("hi") relay: /dns4/kras-08.fluence.dev/tcp/19001/wss/p2p/12D3KooWFtf3rfCDAfWwt6oLZYZbDfn9Vn7bv7g6QjjQxUUEFVBt ... done Result: "12D3KooWCMr9mU894i8JXAFqpgoFtx6qnV1LFPSfVc3Y34N4h4LS: hi"
We've successfully built and deployed a service written in Rust, and called it from our Aqua code using the Fluence CLI.
Run echo service in JS/TS:
-
Install dependencies:
npm i
-
Check out src/aqua/export.aqua with an export of EchoService API to JS/TS:
module Export import "services/echoService.aqua" export EchoService
-
The following code registers a local EchoService:
// register local service with service id "echo" await registerEchoService(serviceId, {echo: (msg) => { console.log("Received message:", msg); return peerId + ": " + msg }});
-
The following code is calling our local JS/TS peer with
EchoService
:func echoJS(peer: string, relay: string, serviceId: string, msg: string) -> string: on peer via relay: EchoService serviceId res <- EchoService.echo(msg) <- res
-
Start the client with our service:
npm run start
... > example@1.0.0 start > node src/echo.js 📗 created a fluence peer 12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN with relay 12D3KooWFEwNWcHqi9rtsmDhsYcDbRUCDXH84RC4FW6UfsFWaoHi Copy this code: fluence run -f 'echoJS("12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN", "12D3KooWFEwNWcHqi9rtsmDhsYcDbRUCDXH84RC4FW6UfsFWaoHi", "echo", "msg")'
-
Open a new terminal in the same registry example directory, and execute the following command to check echo service:
$ fluence run -f 'echoJS("12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN", "12D3KooWFEwNWcHqi9rtsmDhsYcDbRUCDXH84RC4FW6UfsFWaoHi", "echo", "msg")' Running: function: echoJS("12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN", "12D3KooWFEwNWcHqi9rtsmDhsYcDbRUCDXH84RC4FW6UfsFWaoHi", "echo", "msg") relay: /dns4/kras-09.fluence.dev/tcp/19001/wss/p2p/12D3KooWD7CvsYcpF9HE9CCV9aY3SJ317tkXVykjtZnht2EbzDPm ... done Result: "12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN: msg"
We've successfully started JS/TS peer with an EchoService and tested it with Fluence CLI.
Register both services in Registry
As service providers we would like all our echo services to be discoverable without specifying particular peer and service ids, in order to achieve that we should use Registry to advertise our services.
Firstly, we need to create a resource. A resource represents a group of service providers and has a corresponding resource id for its discovery. Secondly, we need to register the service providers on this resource id. So the echo services can be discovered and resolved by the resource id only.
-
The following code registers resource with label
echo
:func registerResource() -> ?string: resource_id, error <- createResource("echo") <- resource_id
-
fluence run -f 'registerResource()'
output:
$ fluence run -f 'registerResource()' Running: function: registerResource() relay: /dns4/kras-01.fluence.dev/tcp/19001/wss/p2p/12D3KooWKnEqMfYo9zvfHmqTLpLdiHXPe4SVqUWcWHDJdFGrSmcA ... done Result: [ "echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77" ]
So the resource id is
echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77
. Please note that the resource id might be different in your case.Using the resource id we can access any registered provider of the resource. There can be more registered echo services on different peers that we can use transparently.
-
This code registers deployed service by given
resource_id
:func registerService(resource_id: string) -> bool: services <- App.services() echoService = services.echoService.default[0] success, error <- registerNodeProvider(echoService.peerId, resource_id, "", ?[echoService.serviceId]) <- success
-
Register service for resource_id from the step 2:
fluence run -f 'registerService("echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77")'
output:
$ fluence run -f 'registerService("echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77")' Running: function: registerService("echo12D3KooWCWsJLrG6evQKPEsDmKNbaW8NGj5JhVZfHumSvyx1Zw4c") relay: /dns4/kras-05.fluence.dev/tcp/19001/wss/p2p/12D3KooWCMr9mU894i8JXAFqpgoFtx6qnV1LFPSfVc3Y34N4h4LS ... done Result: [ [ true ] ]
-
Next, we need to register JS service. For that we have Aqua imports and exports in src/aqua/export.aqua:
import registerProvider from "@fluencelabs/registry/resources-api.aqua" export registerProvider
-
In src/echo.ts we should pass resource id as cmd argument:
let [success, error] = await registerProvider(process.argv[2], "echo", serviceId); console.log("registration result: ", success);
-
So, stop the previous JS client from the step 5 and run
npm run start echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77
(Note: resource_id is from the step 2) output:... Copy this code to call this service: fluence run -f 'echoJS("12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN", "12D3KooWFEwNWcHqi9rtsmDhsYcDbRUCDXH84RC4FW6UfsFWaoHi", "echo", "msg")' registration result: true
We've successfully registered both services, JS/TS and Rust, in Registry and now we can access them only with resource_id
without knowledge of particular peer and service ids.
Call services with Registry (using resource_id)
-
In src/aqua/main.aqua defined
echoAll
to resolve services and call sequentially:func echoAll(resource_id: string, msg: string) -> *string: -- 2 is the min number of providers we want to find providers <- resolveProviders(resource_id, 2) results: *string for p <- providers: on p.peer_id via p.relay_id: EchoService p.service_id! results <- EchoService.echo(msg) <- results
-
Let’s run all registered echo services with only resource id:
fluence run -f 'echoAll("echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77", "hi")'
output:
$ fluence run -f 'echoAll("echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77", "hi")' Running: function: echoAll("echo12D3KooWRgEgxP4qAyUR5jerwKE1rwTLZoAhJ3YwAhpeSMC5pi77", "hi") relay: /dns4/kras-06.fluence.dev/tcp/19001/wss/p2p/12D3KooWDUszU2NeWyUVjCXhGEt1MoZrhvdmaQQwtZUriuGN1jTr ... done Result: [ [ "12D3KooWFEwNWcHqi9rtsmDhsYcDbRUCDXH84RC4FW6UfsFWaoHi: hi", "12D3KooWCmnhnGvKTqEXpVLzdrYu3TkQ3HcLyArGJpLPooJQ69dN: hi" ] ]