These are chat archives for akkadotnet/akka.net

10th
Oct 2016
Corneliu
@corneliutusnea
Oct 10 2016 02:40
@csmorley did you try to do it via the DI? The DI can do custom construction
Aaron Stannard
@Aaronontheweb
Oct 10 2016 07:16
@object yep, haven't quite finished it yet
using Streams heavily in a real project right now and that's informed my tastes somewhat
so I'll be refining it before I publish it, as well as deleting much of the old code
@mmisztal1980 I need to update it to 1.1.2 Akka
that should fix it
DNS / IPv6 stuff
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 13:42
hey guys, from http://getakka.net/docs/clustering/cluster-overview I read: "NOTE: Akka.Cluster is currently a pre-release nuget package, so you'll need to set the pre flag when you install it into your applications. Akka.Cluster will be considered "fully released" once Akka.NET v1.1 ships." -- I thought both Akka .NET 1.1 and Akka.Cluster were already out, so is this just a case of outdated documentation? :)
Arjen Smits
@Danthar
Oct 10 2016 13:43
yes.
You are more then welcome to do a PR :)
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 13:44
I'm not sure what workflow you use for these updates :)
Arjen Smits
@Danthar
Oct 10 2016 13:46
Simple. Fork the getakka.net repo. Make the changes in the .md file as needed.
Make a commit to your fork, and do a PR from there
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 13:47
oh, PR works across repos on github?
I'll give it a try later, thanks :)
Arjen Smits
@Danthar
Oct 10 2016 13:47
And in general this doc applies: http://getakka.net/docs/Contributing%20to%20Akka
But yeah if you fork from a repo. Github tracks that your repo is a Fork, and that it has an upstream repository linked to it. It then allows you to make a PR to that upstream repo.
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 14:04
I'm reading up on clustering, and was just wondering whether it's possible to assign roles dynamically, rather than fixed from config
e.g. have something that tells a node that it needs to adopt a certain role, according to arbitrary demands
Damian Reeves
@DamianReeves
Oct 10 2016 14:29

So I'm working on an application that will have Aggregate Roots like in DDD. I currently see the need to have an actor that given a message, figures out the target entity type and then "activate" the persistent actor for that given entity (by id).

In doing research I came across this statement in the Akka.Cluster.Sharding module:

Cluster sharding is useful in cases when you want to contact with cluster actors using their logical id's, but don't want to care about their physical location inside the cluster or manage their creation. Moreover it's able to rebalance them, as nodes join/leave the cluster. It's often used to represent i.e. Aggregate Roots in Domain Driven Design terminology.

Is my use a canonical example for using cluster sharding? What if I have usages of this pattern that aren't clustered but only have 1 node?

Bartosz Sypytkowski
@Horusiath
Oct 10 2016 14:31
@DamianReeves you can use cluster sharding for only 1 node
Damian Reeves
@DamianReeves
Oct 10 2016 14:32
Cool, does my description of my usecase fit sharding?
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 14:34
yes ;)
Damian Reeves
@DamianReeves
Oct 10 2016 14:35
:+1:
Do you have any usage examples?
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 14:38
There is an example in the core repo. I think it's missing use case of passivation
but this one is present in docs
Damian Reeves
@DamianReeves
Oct 10 2016 14:39
cool... thanks again
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 15:14
hey guys, I'm trying a really bare-bones test of Akka.Cluster based on the documentation here: http://getakka.net/docs/clustering/cluster-overview
akka {
    actor.provider = "Akka.Cluster.ClusterActorRefProvider, Akka.Cluster"
    remote {
        helios.tcp {
            port = 8081
            hostname = localhost
        }
    }
    cluster {
        seed-nodes = ["akka.tcp://ClusterSystem@127.0.0.1:8081"]
    }
}
This doesn't actually work
but if I change both host names to localhost, or both to 127.0.0.1, it works
it looks like Akka doesn't realise that they are actually the same thing
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 15:36
Another thing: If I run a standalone node with itself configured as seed node, I see no errors. If I set up a non-seed node using the aforementioned node as seed node, I get the following error:
blob
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 15:55
OK disregard the last error, had the wrong actorsystem name configured
Damian Reeves
@DamianReeves
Oct 10 2016 16:15
For cluster sharding it would seem that I need to keep track of the ShardRegions I've created. If I call Start(...) or StarAsync(...) twice for the same type it throws and if I call ShardRegion(type) without previously starting a type it throws... am I missing something?
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 16:43
@DamianReeves shard region is singleton in actor system scope, if you're using one actor system for your application, you may set shard region as static property. Also there is other method ClusterSharding.ShardRegion(typeName) which will return shard region if it exists.
Damian Reeves
@DamianReeves
Oct 10 2016 16:45
I get that, but lets say I have 2 AggregateRoots: Inventory and Customer it seems I'd have 2 regions and if I added Product it would be a third, is this the wrong approach?
I'm looking to do something like this:
        ReceiveAsync<AggregateCommandEnvelope>(async envelope =>
        {
            IActorRef shardRegion;
            var aggregateType = envelope.AggregateType;
            var aggregateTypeName = envelope.AggregateType.FullName; //TODO: May need to safe guard this in a different manner
            if(!_aggregateShardRegions.TryGetValue(aggregateTypeName, out shardRegion))
            {
                shardRegion =
                    await _sharding.StartAsync(
                        aggregateTypeName, 
                        Props.Create(aggregateType),
                        ClusterShardingSettings.Create(Context.System), 
                        new ShardingMessageExtractor(10)
                    );
            }
        });
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 16:51
yes
in general identifier of your entity (sharded actor) is something like (string typeName, string shardId, string entityId)
Damian Reeves
@DamianReeves
Oct 10 2016 16:56
Ok, so in my example the typeName I pass to the sharding.Start is a broader concept like InventoryManagementSystem
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 16:58
ClusterSharding.ShardRegion(typeName) matches the ClusterSharding.Start(typeName: ...) parameter
Francis Paulin
@paulinfrancis
Oct 10 2016 17:05

Hi all, I'm just getting into akka .net, and have a question I hope somebody can shed some light on. I've got an ASP .NET Core app, where I have wired SignalR and Akka together.

One of my actors parses some files, and tells my SignalR bridge to notify the clients with the resulting object.

I got an error:

The calling thread cannot access this object because a different thread owns

From json.net when it tried to serialize the object before SignalR sent it to the connected clients.

Without changing anything, the exception is no longer thrown, and everything works.

Am I experiencing issues with the dispatcher? If yes, what is the recommended configuration?

Thanks :)

Bartosz Sypytkowski
@Horusiath
Oct 10 2016 17:17
@paulinfrancis how are you calling a signalr api?
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 17:18
Where do I find the source for the Akka .NET documentation exactly?
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 17:21
@dandago2_twitter http://getakka.net/docs/ . You can also suplement it with docs for akka jvm in many situations
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 17:21
I just want to edit it :)
Francis Paulin
@paulinfrancis
Oct 10 2016 17:22

In my actor system init

public static void Create(IHubContext context, ISimulator simulator)
{
    _signalRSimulatorEventsPusher = new SignalRSimulatorEventsPusher(context);

    _actorSystem = ActorSystem.Create("SimulatorSystem");

    var signalRBridgeProps = Props.Create(() => new SignalRBridgeActor(_signalRSimulatorEventsPusher, ActorReferences.Simulator));
    ActorReferences.SignalRBridge = _actorSystem.ActorOf(signalRBridgeProps, "SignalRBridge");

    //Kosher?
    var simulatorProps = Props.Create(() => new SimulatorActor(simulator, ActorReferences.SignalRBridge));
    ActorReferences.Simulator = _actorSystem.ActorOf(simulatorProps, "Simulator");
}

In my SimulatorActor:

Receive<DiagramLoadedMessage>(message => signalRBridgeActor.Tell(message));

In SignalRBridgeActor. This is where it sometimes blows up.

Receive<DiagramLoadedMessage>(message =>
{
        _simulatorEventsPusher.NotifyLoadedDiagram(message.DiagramRootObject);
});
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 17:25
@Horusiath what do you use to preview the markdown output when you edit the docs?
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 17:25
you've abstracted meat away ;) I was using signalr with actors by wrapping signalr hub/connection with an actor (like Props.Create(() => new SignalRActor(signalrHub)))
@dandago2_twitter vscode with markdown plugin
Francis Paulin
@paulinfrancis
Oct 10 2016 17:33

I had a look here: http://getakka.net/docs/working-with-actors/Dispatchers, so I tried:

var signalRBridgeProps = Props.Create(() => new SignalRBridgeActor(_signalRSimulatorEventsPusher, ActorReferences.Simulator))
                                                                  .WithDispatcher("akka.actor.synchronized-dispatcher");

But it errors: "The current SynchronizationContext may not be used as a TaskScheduler."

Bartosz Sypytkowski
@Horusiath
Oct 10 2016 17:35
SynchornizationContext binds actor to a foreground thread, but it's thread only, which means you cannot use stuff like async/await in actors dispatched by it
Arjen Smits
@Danthar
Oct 10 2016 17:49
Most akka projects can be found under https://github.com/akkadotnet
Francis Paulin
@paulinfrancis
Oct 10 2016 18:53
@Horusiath In your app that uses SignalR, did you configure the dispatcher, or did you go with the defaults?
Damian Reeves
@DamianReeves
Oct 10 2016 18:54
Hey just double checking, but it is now safe for me to use async/await in my actor code?
Bartosz Sypytkowski
@Horusiath
Oct 10 2016 19:58
@paulinfrancis afaik, I was using defaults. I've also written a blog post about using signalr for event logging: http://bartoszsypytkowski.com/log-your-akka-net-application-in-your-browser/
@DamianReeves if you're using ReceiveAsync with default dispatcher, there should be no problem with that
Francis Paulin
@paulinfrancis
Oct 10 2016 20:03
@Horusiath I did some more digging, and the error does indeed only happen when the object is instantiated on another thread, so I'll need to figure out how to marshal to the correct thread somehow.
Daniel D'Agostino
@dandago2_twitter
Oct 10 2016 20:08
I deleted the obsolete note in the clustering page, and it has been merged. I see the note is still on the webpage. How does it actually get updated?
Arjen Smits
@Danthar
Oct 10 2016 20:26
The merge should trigger an website update
Damian Reeves
@DamianReeves
Oct 10 2016 21:24
Just an FYI... the ClusterSharding.Node example throws an exception on running

What are the minimal settings I need to get cluster sharding working?

This does not seem to be enough:

      akka {
        # here we are configuring log levels
        log-config-on-start = off
        stdout-loglevel = DEBUG
        loglevel = DEBUG
        # this config section will be referenced as akka.actor
        actor {
          debug {
              receive = on
              autoreceive = on
              lifecycle = on
              event-stream = on
              unhandled = on
          }
          serializers {
            wire = "Akka.Serialization.WireSerializer, Akka.Serialization.Wire"
          }
          serialization-bindings {
            "System.Object" = wire
          }
        }
        remote {
          helios.tcp {
            public-hostname = "localhost"
            hostname = "localhost"
            port = 0
          }
        }
        cluster {
          auto-down-unreachable-after = 5s
          sharding {
            least-shard-allocation-strategy.rebalance-threshold = 3
          }
        }
        persistence {
          # publish-plugin-commands = on              
          journal {
              plugin = "akka.persistence.journal.sqlite"
              sqlite {
                  class = "Akka.Persistence.Sqlite.Journal.SqliteJournal, Akka.Persistence.Sqlite"
                  plugin-dispatcher = "akka.actor.default-dispatcher"
                  table-name = event_journal
                  metadata-table-name = journal_metadata
                  auto-initialize = on
                  connection-string-name = "SqliteDb"
              }
          }

          snapshot-store {
            plugin = "akka.persistence.snapshot-store.sqlite"
            sqlite {
                # Class name of the plugin.
                class = "Akka.Persistence.Sqlite.Snapshot.SqliteSnapshotStore, Akka.Persistence.Sqlite"

                # Dispatcher for the plugin actor.
                plugin-dispatcher = "akka.persistence.dispatchers.default-plugin-dispatcher"

                # Dispatcher for streaming snapshot IO.
                stream-dispatcher = "akka.persistence.dispatchers.default-stream-dispatcher"

                table-name = snapshot
                auto-initialize = on
                connection-string-name = "SqliteDb"
            }
          }
        }
      }
Damian Reeves
@DamianReeves
Oct 10 2016 21:39
Ahh, figured it out was missing:
provider = "Akka.Cluster.ClusterActorRefProvider, Akka.Cluster"
Maciek Misztal
@mmisztal1980
Oct 10 2016 22:39
blob
Confirmed - I'm able to get lighthouse started on Mono :) with a minor burp inside TopShelf
@Aaronontheweb I'd like to take this a bit further and have some fun with Docker on Linux on the way, I'm curious about one thing though - how do I determine the addresses of seed nodes running as containers?