"It's just has to take the actor reference and call something like release(Actor)."
in autofac for example you would not do it that way but instead reslove a lifetime scope or an owned instance and then dispose of that after, there is no release method on the actual container.
@stefansedich The release thing is based on the following post http://blog.ploeh.dk/2014/05/19/di-friendly-framework/ which I think does an excellent job explaining the need plus its an interesting read.
they don't have release in that way instead the request gets wrapped in a lifetime scope which itself is disposed.
Although if using autofac we could resolve an Owned<T> and in the release dispose that
would probably be enough
where T being the actor type
so yeah sorry release is probably enough :)
let actorOfCtor (state : 'a) (fn : 'a -> 'Message -> unit) (mailbox : Actor<'Message>) : Cont<'Message, 'Returned> =
let rec loop() =
actor {
let! msg = mailbox.Receive()
fn state msg
return! loop()
}
loop()
allows the creation of an actor that will have some sort of initial data that is usable by the messageHandler every message
@raymens if you must use some state in F# API, pass it as a rec loop parameter , eg:
type MyState = { X: int; Y: int }
let aref = spawn system "my-actor" <| fun mailbox ->
let rec loop state = actor {
let! msg = mailbox.Receive()
return! loop ({ state with Y = msg })
}
loop ({X = 1; Y = 2})
this way it maintain F# principles - immutable data, function as unit of composition etc.
the only real docs I have for at it at the moment are its specs and samples
@stefansedich submitted a PR a couple of days ago which should enable Mono support in Helios
and should fix the Akka Remoting issues
turns out that
TcpClient and UdpClient call Socket.Bind in their constructors on Mono, which makes it impossible to set socket options on them
@smalldave you might appreciate this, btw https://github.com/petabridge/lighthouse - took an old project I started on with Akka.Cluster a while back and I'm modernizing it now
basically it's a windows service that acts as a dedicated seed node for clusters that want to use roles
runs virtually no code other than what's in Akka.Cluster so it doesn't need to be updated
designed for PaaS environments like Elastic Beanstalk, AppHarbor, Azure Websites
I'm in the process of rewriting it to use Topshelf now
and an NLog adapter that lets it write to the windows event log
Topshelf itself is super neat - never used it until I saw some of our contributors mention it: http://topshelf-project.com/
lighthouse looks interesting
can't believe I hadn't used it until now
the idea behind lighthouse is basically to have a 1-2 lighthouse instances sitting on dedicated IaaS boxes
it's meant to be a scriptable, configurable Akka.Cluster service that you will eventually just be able to deploy via a packaged binary
and some .config modifications
the idea is to just keep a couple of these around as a lightweight, do-it-yourself Akka.Cluster service discovery endpoint
they're not necessary in environments where you can use something like the EC2 SDK to query all of the IPs of other machines in your deployment group