We're going to be spending most of our time in Unit 2 working with the ChartingActor
, an actor that is responsible for actually plotting all the data on this chart:
NOTE: If you're following along using the eBook / .ePub, you won't see the animation. Click here to see it.
BUT, if you try to build and run SystemCharting.sln
right now (in the /DoThis/ folder) for Unit 2 right away, you'll see the following output:
Well, that's not very exciting. Aren't we supposed to be building a real-time data visualization application in Unit 2? What gives?
Oh wait, there's an exception in the Debug window. What does it say?
[ERROR][2/24/2015 11-48-34 AM][Thread 0010][akka://ChartActors/user/charting] Cross-thread operation not valid: Control 'sysChart' accessed from a thread other than the thread it was created on. Cause: System.InvalidOperationException: Cross-thread operation not valid: Control 'sysChart' accessed from a thread other than the thread it was created on.
None of the events we want to chart are getting graphed. Hmm... sounds like our events aren't getting to the UI thread. We need to find some way to dispatch and synchronize events with the UI thread so that our chart is updated!
Does this mean we have to rewrite ChartingActor
with some evil code to manually synchronize with the UI thread?
Nope! We can relax.
We can solve this problem using HOCON configuration in Akka.NET without updating any of the code that defines ChartingActor
.
But first, we need to understand Dispatcher
s.
A Dispatcher
is the piece of glue that pushes messages from your actor's mailbox into your actor instances themselves. That is, the Dispatcher
is what pushes messages into the OnReceive()
method of your actors. All actors which share a given Dispatcher
also share that Dispatcher
's threads for parallel execution.
The default dispatcher in Akka.NET is the ThreadPoolDispatcher
. As you can probably guess, this dispatcher runs all of our actors on top of the CLR ThreadPool
.
There are several types of Dispatcher
s we can use with our actors:
This Dispatcher
runs multiple actors on a single thread.
This Dispatcher
runs actors on top of the CLR ThreadPool
for maximum concurrency.
This Dispatcher
schedules all actor messages to be processed in the same synchronization context as the caller. 99% of the time, this is where you're going to run actors that need access to the UI thread, such as in client applications.
The SynchronizedDispatcher
uses the current SynchronizationContext to schedule executions.
Note: As a general rule, actors running in the
SynchronizedDispatcher
shouldn't do much work. Avoid doing any extra work that may be done by actors running in other pools.
In this lesson, we're going to use the SynchronizedDispatcher
to ensure that the ChartingActor
runs on the UI thread of our WinForms application. That way, the ChartingActor
can update any UI element it wants without having to do any cross-thread marshalling - the actor's Dispatcher
can automatically take care of that for us!
This Dispatcher
runs actors on top of a dedicated group of threads, for tunable concurrency.
This is meant for actors that need their own dedicated threads in order to run (that need isolation guarantees). This is primarily used by System
actors so you won't touch it much.
The short answer is "no".
Running actors on the UI thread is fine, as long as those actors don't perform any long-running operations such as disk or network I/O. In fact, running actors on the UI thread is a smart thing to do for handling UI events and updates.
Why? Because running actors on the UI thread eliminates all of the normal synchronization worries you'd otherwise have to do in a multi-threaded WPF or WinForms app.
Remember: Akka.NET actors are lazy. They don't do any work when they're not receiving messages. They don't consume resources when they're inactive.
As we realized before, our chart isn't updating because the actor doing the graphing (ChartingActor
) is not synchronizing its events with the UI thread.
To solve this problem, all we have to do is change the ChartingActor
to use the CurrentSynchronizationContextDispatcher
, and it will automatically run on the UI thread for us!
BUT: we want to do this without touching our actual actor code. How can we deploy the ChartingActor
so that it uses the CurrentSynchronizationContextDispatcher
without modifying the actor itself?
Time to meet HOCON.
Akka.NET leverages a configuration format, called HOCON, to allow you to configure your Akka.NET applications with whatever level of granularity you want.
HOCON (Human-Optimized Config Object Notation) is a flexible and extensible configuration format. It will allow you to configure everything from Akka.NET's IActorRefProvider
implementation, logging, network transports, and more commonly - how individual actors are deployed.
Values returned by HOCON are strongly typed (i.e. you can fetch out an int
, a Timespan
, etc).
HOCON allows you to embed easily-readable configuration inside of the otherwise hard-to-read XML in App.config and Web.config. HOCON also lets you query configs by their section paths, and those sections are exposed strongly typed and parsed values you can use inside your applications.
HOCON also lets you nest and/or chain sections of configuration, creating layers of granularity and providing you a semantically namespaced config.
HOCON is commonly used for tuning logging settings, enabling special modules (such as Akka.Remote
), or configuring deployments such as the Dispatcher
for our ChartingActor
in this lesson.
For example, let's configure an ActorSystem
with HOCON:
var config = ConfigurationFactory.ParseString(@"
akka.remote.helios.tcp {
transport-class =
""Akka.Remote.Transport.Helios.HeliosTcpTransport, Akka.Remote""
transport-protocol = tcp
port = 8091
hostname = ""127.0.0.1""
}");
var system = ActorSystem.Create("MyActorSystem", config);
As you can see in that example, a HOCON Config
object can be parsed from a string
using the ConfigurationFactory.ParseString
method. Once you have a Config
object, you can then pass this to your ActorSystem
inside the ActorSystem.Create
method.
NOTE: In this example we configured a specific network transport for use with
Akka.Remote
, a concept that goes well beyond what's covered in Unit 2. Don't worry about the specifics for now.
Deployment is a vague concept, but it's closely tied to HOCON. An actor is "deployed" when it is instantiated and put into service within the ActorSystem
somewhere.
When an actor is instantiated within the ActorSystem
it can be deployed in one of two places: inside the local process or in another process (this is what Akka.Remote
does.)
When an actor is deployed by the ActorSystem
, it has a range of configuration settings. These settings control a wide range of behavior options for the actor, such as: is this actor going to be a router? What Dispatcher
will it use? What type of mailbox will it have? (More on these concepts in later lessons.)
We haven't gone over what all these options mean, but the key thing to know for now is that the settings used by the ActorSystem
to deploy an actor into service can be set within HOCON.
This also means that you can change the behavior of actors dramatically (by changing these settings) without having to actually touch the actor code itself.
Flexible config FTW!
NOTE We know that App.config and Web.config are deprecated in dotnet core, we will update this sample with a new way to declare HOCON configuration once the standalone HOCON 3.0 are released.
Parsing HOCON from a string
is handy for small configuration sections, but what if you want to be able to take advantage of Configuration Transforms for App.config
and Web.config
and all of the other nice tools we have in the System.Configuration
namespace?
As it turns out, you can use HOCON inside these configuration files too!
Here's an example of using HOCON inside App.config
:
<?xml version="1.0" encoding="utf-8" ?>
<configuration>
<configSections>
<section name="akka"
type="Akka.Configuration.Hocon.AkkaConfigurationSection, Akka" />
</configSections>
<akka>
<hocon>
<![CDATA[
akka {
# here we are configuring log levels
log-config-on-start = off
stdout-loglevel = INFO
loglevel = ERROR
# this config section will be referenced as akka.actor
actor {
provider = "Akka.Remote.RemoteActorRefProvider, Akka.Remote"
debug {
receive = on
autoreceive = on
lifecycle = on
event-stream = on
unhandled = on
}
}
# here we're configuring the Akka.Remote module
remote {
helios.tcp {
transport-class =
"Akka.Remote.Transport.Helios.HeliosTcpTransport, Akka.Remote"
#applied-adapters = []
transport-protocol = tcp
port = 8091
hostname = "127.0.0.1"
}
log-remote-lifecycle-events = INFO
}
]]>
</hocon>
</akka>
</configuration>
And then we can load this configuration section into our ActorSystem
via the following code:
var system = ActorSystem.Create("Mysystem");
// Loads section.AkkaConfig from App or Web.config automatically
// FYI, section.AkkaConfig is built into Akka.NET for you
Although this isn't a concept we leverage explicitly in Unit 2, it's a powerful trait of the Config
class that comes in handy in lots of production use cases.
HOCON supports the concept of "fallback" configurations - it's easiest to explain this concept visually.
NOTE: If you're following along using the eBook / .ePub, you won't see the animation. Click here to see it.
To create something that looks like the diagram above, we have to create a Config
object that has three fallbacks chained behind it using syntax like this:
var f0 = ConfigurationFactory.ParseString("a = bar");
var f1 = ConfigurationFactory.ParseString("b = biz");
var f2 = ConfigurationFactory.ParseString("c = baz");
var f3 = ConfigurationFactory.ParseString("a = foo");
var yourConfig = f0.WithFallback(f1)
.WithFallback(f2)
.WithFallback(f3);
If we request a value for a HOCON object with key "a", using the following code:
var a = yourConfig.GetString("a");
Then the internal HOCON engine will match the first HOCON file that contains a definition for key a
. In this case, that is f0
, which returns the value "bar".
The reason is because HOCON only searches through fallback Config
objects if a match is NOT found earlier in the Config
chain. If the top-level Config
object has a match for a
, then the fallbacks won't be searched. In this case, a match for a
was found in f0
so the a=foo
in f3
was never reached.
What happens if we run the following code, given that c
isn't defined in f0
or f1
?
var c = yourConfig.GetString("c");
NOTE: If you're following along using the eBook / .ePub, you won't see the animation. Click here to see it.
In this case yourConfig
will fallback twice to f2
and return "baz" as the value for key c
.
Now that we understand HOCON, let's use it to fix the Dispatcher
for ChartingActor
!
We need to configure ChartingActor
to use the SynchronizedDispatcher
in order to make our charting work correctly on the UI thread.
The first thing you need to do is declare the AkkaConfigurationSection
at the top of your App.config
:
<!-- in App.config file -->
<!-- add this right after the opening <configuration> tag -->
<configSections>
<section name="akka" type="Akka.Configuration.Hocon.AkkaConfigurationSection, Akka" />
</configSections>
Next, add the content of the AkkaConfigurationSection
to App.config
:
<!-- in App.config file -->
<!-- add this anywhere after <configSections> -->
<akka>
<hocon>
<![CDATA[
akka {
actor {
deployment {
# this nested section will be accessed by akka.actor.deployment
# used to configure our ChartingActor
/charting {
# causes ChartingActor to run on the UI thread for WinForms
dispatcher = akka.actor.synchronized-dispatcher
}
}
}
}
]]>
</hocon>
</akka>
We should point out that akka.actor.synchronized-dispatcher
is the shorthand name built into Akka.NET's default configuration for the CurrentSynchronizationContextDispatcher
. So you don't need to use a fully-qualified type name.
You might have also noticed that the configuration section that pertains to the ChartingActor
was declared as /charting
- this is because actor deployment is done by the path and name of the actor, not the actor's type.
Here's how we create the ChartingActor
inside Main.cs
:
_chartActor = Program.ChartActors.ActorOf(Props.Create(() =>
new ChartingActor(sysChart)), "charting");
When we call ActorSystem.ActorOf
the ActorOf
method will automatically look for any deployments declared in the akka.actor.deployment
configuration section that correspond to the path of this actor. In this case, the path of this actor is /user/charting
, which corresponds to the akka.actor.deployment
values for /charting
in the config section above.
As the Akka.NET end-user, you can only specify deployment settings for actors created inside the
/user/
hierarchy. Because of this, you don't need to specify/user
when you declare your deployment settings - it's implicit.By extension, you also cannot specify how to deploy the
/system
actors. This is up to theActorSystem
.
And... we're finished!
Build and run SystemCharting.sln
and you should see the following:
Compare your code to the code in the /Completed/ folder to compare your final output to what the instructors produced.
Nice work on completing your first lesson in Unit 2! We covered a lot of concepts and hopefully you're going to walk away from this with an appreciation for just how powerful Akka.NET's configuration model truly is.
Let's move onto Lesson 2 - Using ReceiveActor
for Smarter Message Handling.
As you probably guessed while reading the HOCON configs above, any line with #
at the front of it is treated as a comment in HOCON. Learn more about HOCON syntax here.
Come ask any questions you have, big or small, in this ongoing Bootcamp chat with the Petabridge & Akka.NET teams.
If there is a problem with the code running, or something else that needs to be fixed in this lesson, please create an issue and we'll get right on it. This will benefit everyone going through Bootcamp.