I'd like to run a websocket server and send messages from another module to it.
So far I have only managed to pass a channel to the module which starts the server. But I'd like to have as globally as writeFile which can be called from any module at any time.
Also I'd like to have multiple clients with sendMessage. Once a connection closes I assume the thread still stays in the forever loop.
Server.hs
import Network.WebSockets
import Control.Concurrent
import Control.Monad
import Data.ByteString
createServer :: IO (Chan ByteString)
createServer = do
chan <- newChan
forkIO $ runServer "127.0.0.1" 8080 (serverApp chan)
ready <- readChan chan -- wait for client
return chan
serverApp :: Chan ByteString -> PendingConnection -> IO ()
serverApp chan pending =
do
print "Client connected"
connection <- acceptRequest pending
writeChan chan "ready"
forever $ do
msg <- readChan chan
sendTextData connection msg
sendMessage :: Chan ByteString -> ByteString -> IO ()
sendMessage = writeChan
Main.hs
main :: IO ()
main = do
client <- createServer
sendMessage client ("hello" :: ByteString)
I ended up using an MVar with unsafePerformIO. While not exactly recommended the the code is neat and simple.
createServer :: IO ()
createServer = do
_ <- forkIO $ runServer "127.0.0.1" 8080 serverApp
return ()
serverApp :: PendingConnection -> IO ()
serverApp pending =
do
connection <- acceptRequest pending
forever $ do
msg <- takeMVar channel
sendTextData connection msg
channel :: MVar ByteString
{-# NOINLINE channel #-}
channel = unsafePerformIO newEmptyMVar
sendMessage :: ByteString -> IO ()
sendMessage = putMVar channel
The code is still missing exception handling and it only works for 1 connected client.
Related
I have the following code for a simple web server in Haskell.
module Main where
import Control.Exception
import Control.Monad
import qualified Data.ByteString as B
import qualified Data.ByteString.Char8
import Data.Char
import Network.Socket
import Network.Socket.ByteString (recv, sendAll)
import System.Mem (performGC)
packStr = B.pack . map (fromIntegral . ord)
msg :: String
msg =
"HTTP/1.1 200 OK\n"
++ "Content-Type: text/html\n\n"
++ "<html><body><a href=#>download</a></body></html>\n"
manageConnections sock =
forever
( do
(conn, addr) <- accept sock
putStrLn $ "Connection from" ++ show addr ++ "\n"
r <- recv conn 1024
B.putStr r
sendAll conn $ packStr msg
seq id close conn -- for force close the socket the socket
)
main :: IO ()
main = do
putStrLn "Starting"
sock <- socket AF_INET Stream 0
setSocketOption sock ReuseAddr 1
bind sock (SockAddrInet 1337 $ tupleToHostAddress (127, 0, 0, 1))
listen sock 4
manageConnections sock
As you can see the function manageConnections is where the main thread spends most of the time. My question is how do I handle user interrupts in the manageConnections function ?. What I mean is, if I press Ctrl-C the program would not close the "conn" socket and it would exit. I want to catch the user interrupt and close the socket properly.
if I press Ctrl-C the program would not close the "conn" socket
All file descriptors (including sockets) are closed by the OS when the program finishes its execution.
I want to catch the user interrupt
You might want to try System.Posix.Signals.installHandler.
I have a java jar that generates the tcp stream on particular port.
I cam run the using java command like java -jar runner.jar and this start producing the stream of message on port 8888.
When I do nc -l 8888 I can see the messages.
I want to read this stream using scala and another framework or tool like akka, akka-stream.
Can anyone help me understand the best tool, framework or any other technique to read this tcp stream.
I tried using akka stream with following code :-
implicit val system = ActorSystem()
implicit val mater = ActorMaterializer() val ss = Tcp().outgoingConnection("127.0.0.1", 8888)
.to(Sink.foreach(println(_)))
Source.empty.to(ss).run()
I also tried
Tcp().outgoingConnection(new InetSocketAddress("127.0.0.1", 8888))
.runWith(Source.maybe[ByteString], Sink.foreach(bs => println(bs.utf8String)))
This doesn't work.
I only need to read the messages and process on my own.
Thanks
As I understand you want to setup TCP server, here is example of TCP Echo using akka streams
def server(system: ActorSystem, address: String, port: Int): Unit = {
implicit val sys = system
import system.dispatcher
implicit val materializer = ActorMaterializer()
val handler = Sink.foreach[Tcp.IncomingConnection] { conn =>
println("Client connected from: " + conn.remoteAddress)
conn handleWith Flow[ByteString]
}
val connections = Tcp().bind(address, port)
val binding = connections.to(handler).run()
binding.onComplete {
case Success(b) =>
println("Server started, listening on: " + b.localAddress)
case Failure(e) =>
println(s"Server could not bind to $address:$port: ${e.getMessage}")
system.terminate()
}
}
The current (10.1.3) Akka HTTP docs:
https://doc.akka.io/docs/akka-http/current/server-side/graceful-termination.html
talk about graceful termination, using this code sample:
import akka.actor.ActorSystem
import akka.http.scaladsl.server.Directives._
import akka.http.scaladsl.server.Route
import akka.stream.ActorMaterializer
import scala.concurrent.duration._
implicit val system = ActorSystem()
implicit val dispatcher = system.dispatcher
implicit val materializer = ActorMaterializer()
val routes = get {
complete("Hello world!")
}
val binding: Future[Http.ServerBinding] =
Http().bindAndHandle(routes, "127.0.0.1", 8080)
// ...
// once ready to terminate the server, invoke terminate:
val onceAllConnectionsTerminated: Future[Http.HttpTerminated] =
Await.result(binding, 10.seconds)
.terminate(hardDeadline = 3.seconds)
// once all connections are terminated,
// - you can invoke coordinated shutdown to tear down the rest of the system:
onceAllConnectionsTerminated.flatMap { _ ⇒
system.terminate()
}
I am wondering at what point this get called at, the comment states:
// once ready to terminate the server
What does this mean exactly, i.e. who/what determines the server is ready to terminate?
Do I have to put the shutdown code above in some hook function somewhere so that it is invoked on Akka HTTP receiving a SIGTERM?
I’ve tried putting this into the shutdown hook:
CoordinatedShutdown(system).addCancellableJvmShutdownHook{
// once ready to terminate the server, invoke terminate:
val onceAllConnectionsTerminated: Future[Http.HttpTerminated] =
Await.result(binding, 10.seconds)
.terminate(hardDeadline = 3.seconds)
// once all connections are terminated,
// - you can invoke coordinated shutdown to tear down the rest of the system:
onceAllConnectionsTerminated.flatMap { _ ⇒
system.terminate()
}
}
But requests in progress are ended immediately upon sending a SIGTERM (kill ), rather than completing.
I also found a slightly different way of shutdown from https://github.com/akka/akka-http/issues/1210#issuecomment-338825745:
CoordinatedShutdown(system).addTask(
CoordinatedShutdown.PhaseServiceUnbind, "http_shutdown") { () =>
bind.flatMap(_.unbind).flatMap { _ =>
Http().shutdownAllConnectionPools
}.map { _ =>
Done
}
}
Maybe I should using this to handle SIGTERM? I'm not sure..
Thanks!
Resolution taken from this answer here:
https://discuss.lightbend.com/t/graceful-termination-on-sigterm-using-akka-http/1619
CoordinatedShutdown(system).addTask(
CoordinatedShutdown.PhaseServiceUnbind, "http_shutdown") { () =>
bind.flatMap(_.terminate(hardDeadline = 1.minute)).map { _ =>
Done
}
}
For me the main part was to increase akka.coordinated-shutdown.default-phase-timeout as it took longer to finish the processing of the request than the 5 second default. You can also just increase the timeout for that one phase. I had the following message in my logs:
Coordinated shutdown phase [service-unbind] timed out after 5000 milliseconds
I have a running Akka HTTP application and I want to shut it down.
Pressing Ctrl + C in SBT does not work for me (my shell is currently Git Bash for Windows).
What's the recommended way of shutting down an Akka app gracefully?
Taking inspiration from this thread, I added a route to my application that shuts down the application:
def shutdownRoute: Route = path("shutdown") {
Http().shutdownAllConnectionPools() andThen { case _ => system.terminate() }
complete("Shutting down app")
}
where system is the app's ActorSystem.
Given this route, I can now shut down my application with
curl http://localhost:5000/shutdown
Edit:
Being able to shut down a server remotely is not a good idea for production code. In the comments, Henrik pointed to a different way that shuts down the server by hitting Enter in the SBT console:
StdIn.readLine()
// Unbind from the port and shut down when done
bindingFuture
.flatMap(_.unbind())
.onComplete(_ => system.terminate())
For context, I put the above code at the end of server initialization:
// Gets the host and a port from the configuration
val host = system.settings.config.getString("http.host")
val port = system.settings.config.getInt("http.port")
implicit val materializer = ActorMaterializer()
// bindAndHandle requires an implicit ExecutionContext
implicit val ec = system.dispatcher
import akka.http.scaladsl.server.Directives._
val route = path("hi") {
complete("How's it going?")
}
// Starts the HTTP server
val bindingFuture: Future[ServerBinding] =
Http().bindAndHandle(route, host, port)
val log = Logging(system.eventStream, "my-application")
bindingFuture.onComplete {
case Success(serverBinding) =>
log.info(s"Server bound to ${serverBinding.localAddress}")
case Failure(ex) =>
log.error(ex, "Failed to bind to {}:{}!", host, port)
system.terminate()
}
log.info("Press enter key to stop...")
// Let the application run until we press the enter key
StdIn.readLine()
// Unbind from the port and shut down when done
bindingFuture
.flatMap(_.unbind())
.onComplete(_ => system.terminate())
Here is the setup: I want to be able to stream messages (jsons converted to bytestrings) from a publisher to a remote server subscriber over a tcp connection.
Ideally, the publisher would be an actor that would receive internal messages, queue them and then stream them to the subscriber server if there is outstanding demand of course. I understood that what is necessary for this is to extend ActorPublisher class in order to onNext() the messages when needed.
My problem is that so far I am able just to send (receive and decode properly) one shot messages to the server opening a new connection each time. I did not manage to get my head around the akka doc and be able to set the proper tcp Flow with the ActorPublisher.
Here is the code from the publisher:
def send(message: Message): Unit = {
val system = Akka.system()
implicit val sys = system
import system.dispatcher
implicit val materializer = ActorMaterializer()
val address = Play.current.configuration.getString("eventservice.location").getOrElse("localhost")
val port = Play.current.configuration.getInt("eventservice.port").getOrElse(9000)
/*** Try with actorPublisher ***/
//val result = Source.actorPublisher[Message] (Props[EventActor]).via(Flow[Message].map(Json.toJson(_).toString.map(ByteString(_))))
/*** Try with actorRef ***/
/*val source = Source.actorRef[Message](0, OverflowStrategy.fail).map(
m => {
Logger.info(s"Sending message: ${m.toString}")
ByteString(Json.toJson(m).toString)
}
)
val ref = Flow[ByteString].via(Tcp().outgoingConnection(address, port)).to(Sink.ignore).runWith(source)*/
val result = Source(Json.toJson(message).toString.map(ByteString(_))).
via(Tcp().outgoingConnection(address, port)).
runFold(ByteString.empty) { (acc, in) ⇒ acc ++ in }//Handle the future
}
and the code from the actor which is quite standard in the end:
import akka.actor.Actor
import akka.stream.actor.ActorSubscriberMessage.{OnComplete, OnError}
import akka.stream.actor.{ActorPublisherMessage, ActorPublisher}
import models.events.Message
import play.api.Logger
import scala.collection.mutable
class EventActor extends Actor with ActorPublisher[Message] {
import ActorPublisherMessage._
var queue: mutable.Queue[Message] = mutable.Queue.empty
def receive = {
case m: Message =>
Logger.info(s"EventActor - message received and queued: ${m.toString}")
queue.enqueue(m)
publish()
case Request => publish()
case Cancel =>
Logger.info("EventActor - cancel message received")
context.stop(self)
case OnError(err: Exception) =>
Logger.info("EventActor - error message received")
onError(err)
context.stop(self)
case OnComplete =>
Logger.info("EventActor - onComplete message received")
onComplete()
context.stop(self)
}
def publish() = {
while (queue.nonEmpty && isActive && totalDemand > 0) {
Logger.info("EventActor - message published")
onNext(queue.dequeue())
}
}
I can provide the code from the subscriber if necessary:
def connect(system: ActorSystem, address: String, port: Int): Unit = {
implicit val sys = system
import system.dispatcher
implicit val materializer = ActorMaterializer()
val handler = Sink.foreach[Tcp.IncomingConnection] { conn =>
Logger.info("Event server connected to: " + conn.remoteAddress)
// Get the ByteString flow and reconstruct the msg for handling and then output it back
// that is how handleWith work apparently
conn.handleWith(
Flow[ByteString].fold(ByteString.empty)((acc, b) => acc ++ b).
map(b => handleIncomingMessages(system, b.utf8String)).
map(ByteString(_))
)
}
val connections = Tcp().bind(address, port)
val binding = connections.to(handler).run()
binding.onComplete {
case Success(b) =>
Logger.info("Event server started, listening on: " + b.localAddress)
case Failure(e) =>
Logger.info(s"Event server could not bind to $address:$port: ${e.getMessage}")
system.terminate()
}
}
thanks in advance for the hints.
My first recommendation is to not write your own queue logic. Akka provides this out-of-the-box. You also don't need to write your own Actor, Akka Streams can provide it as well.
First we can create the Flow that will connect your publisher to your subscriber via Tcp. In your publisher code you only need to create the ActorSystem once and connect to the outside server once:
//this code is at top level of your application
implicit val actorSystem = ActorSystem()
implicit val actorMaterializer = ActorMaterializer()
import actorSystem.dispatcher
val host = Play.current.configuration.getString("eventservice.location").getOrElse("localhost")
val port = Play.current.configuration.getInt("eventservice.port").getOrElse(9000)
val publishFlow = Tcp().outgoingConnection(host, port)
publishFlow is a Flow that will input ByteString data that you want to send to the external subscriber and outputs ByteString data that comes from subscriber:
// data to subscriber ----> publishFlow ----> data returned from subscriber
The next step is the publisher Source. Instead of writing your own Actor you can use Source.actorRef to "materialize" the Stream into an ActorRef. Essentially the Stream will become an ActorRef for us to use later:
//these values control the buffer
val bufferSize = 1024
val overflowStrategy = akka.stream.OverflowStrategy.dropHead
val messageSource = Source.actorRef[Message](bufferSize, overflowStrategy)
We also need a Flow to convert Messages into ByteString
val marshalFlow =
Flow[Message].map(message => ByteString(Json.toJson(message).toString))
Finally we can connect all of the pieces. Since you aren't receiving any data back from the external subscriber we'll ignore any data coming in from the connection:
val subscriberRef : ActorRef = messageSource.via(marshalFlow)
.via(publishFlow)
.runWith(Sink.ignore)
We can now treat this stream as if it were an Actor:
val message1 : Message = ???
subscriberRef ! message1
val message2 : Message = ???
subscriberRef ! message2