akkadotnet/akka.net

Port of Akka actors for .NET

Bart de Boer

@boekabart

(note that your 'outer' actor won't get confirmation directly anymore)

Arsene

@Tochemey

@boekabart The reason I used forward is that I do not want to pass the SenderRef into the message.

@boekabart and the Ask-PipeTo is not too recommended in Actors.

Bart de Boer

@boekabart

yes, but you do want to the loader to respond to the manager, right?

Arsene

@Tochemey

@boekabart Yes

Bart de Boer

@boekabart

separate your 'outer contract' from your 'inner contract'

Arsene

@Tochemey

@boekabart How?

Bartosz Sypytkowski

@Horusiath

@Tochemey if ModulesLoaded is addresses specifically to manager, you don't need to check for it using ExpectMsg

Arsene

@Tochemey

However It still did not work

Bart de Boer

@boekabart

@Tochemey try drawing a sequence diagram with 'outside requester', 'manager' and 'loader' as actors

Bartosz Sypytkowski

@Horusiath

@Tochemey you either can use Ask (which is not recommended, because it's slow) or change behavior after LoadModules call

Bart de Boer

@boekabart

Ask won't work, because the manager has to update its state in the continuation (recipes)

Bartosz Sypytkowski

@Horusiath

I'll send an example in a sec

https://gist.github.com/Horusiath/4f252ceef5eae489928142da98192e4d

@boekabart I was talking about using Ask inside ReceiveAsync

Arsene

@Tochemey

@Horusiath So please how do I then proceed with the test code because even the code change I am now getting ModuleLoaded message instead of the expected result: This the new test code:

            ActorSystemRefs.ActorSystem = Sys;
            var actorSystem = ActorSystemRefs.ActorSystem;
            var moduleManager = actorSystem.ActorOf(Props.Create(() => new ModulesActor()));
            moduleManager.Tell(new LoadModules());
            moduleManager.Tell(new CheckModuleState("mtn-gh"));
            var moduleMeta = ExpectMsg<ModuleMeta>();
            Assert.NotNull(moduleMeta);
            Assert.True(moduleMeta.State == ModuleState.ONLINE);

Bartosz Sypytkowski

@Horusiath

@Tochemey have you applied changes from my sample gist?

Arsene

@Tochemey

@Horusiath Yes I have done that exactly.

@Horusiath A mn I am checking something

Still not working

Bartosz Sypytkowski

@Horusiath

it looks like you have Forward or Tell(..., Sender) somewhere in your ModulesActor code

Arsene

@Tochemey

@Horusiath The ModulesActor has a child Actor call ModulesLoader. Its jobs is to load the modules and returns a ModulesLoaded info to the ModulesActors. ModulesActors use forward to communicate to the child.

Bartosz Sypytkowski

@Horusiath

why are you using Forward if you need to get reply back to ModulesActor?

Arsene

@Tochemey

So what do you suggest then?

Bartosz Sypytkowski

@Horusiath

just as I've written in the sample, you can use tell and asynchronously wait for the reply

Arsene

@Tochemey

@Horusiath How I do design the test code then? I have grabbed the architecture.

Should I send Tell(LoadModules) and another Tell(CheckState)?

Bartosz Sypytkowski

@Horusiath

Test code stays the same:

            moduleManager.Tell(new LoadModules());
            moduleManager.Tell(new CheckModuleState("mtn-gh"));
            var moduleMeta = ExpectMsg<ModuleMeta>();

Garrard Kitchen

@garrardkitchen

Hi @Aaronontheweb, I wasn't able to get mono working with F# solution in windows environment so not able to tell whether the sockets error is because of mono or akka. Did any of your tyre kicking turn up anything?

Garrard Kitchen

@garrardkitchen

Hi @Aaronontheweb, I’ve now ruled out F#. Created new sln in win env and all ok but when running on mac, same error.

Garrard Kitchen

@garrardkitchen

Has anybody had success with using akka cluster (1.1.1) in docker container?

Weston

@ronnyek

whats new with akka.net in the past couple mo

are the cluster/lighthouse things working a bit more smoothly (eg, so that if lighthouse drops out after nodes discover one another, and come back in, it doesnt leave you with a cluster thats unusable?)

Garrard Kitchen

@garrardkitchen

@ronnyek cluster yes, can't comment about lighthouse as we don't use it. Before we pushed 1.1 release out we sequenced through a known set of edge cases where we knew our solution would break. Happy to say, it didn't break. Our multi-tenanted solution consists of several windows services and a few IIS sites. We tested the cluster in isolation (1 server) and over multiple servers. All good.