Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Play-slick with SecureSocial: Running DB IO in a separate thread pool

I have a Play 2.2.1 app that uses play-slick 0.5.0.8 to persist data to a Postgresql backend and SecureSocial 2.1.2 to handle user authorisation.

Since play-slick transactions are blocking, I have created a separate slick-context execution context in my /conf/application.conf file, as per the instructions found in the plugin's Wiki:

play {
  akka {
    actor {
      slick-context = {
        fork-join-executor {
          parallelism-min = 300
          parallelism-max = 300
        }
      }
    }
  }
}

This allows me to create a controller Action that runs in a separate execution context and does not block threads in the default thread pool. Eg. /app/controllers/Application.scala:

Example One - Using play-slick's DBAction:

import play.api.db.slick._
object Application extends Controller{ 

  // this controller Action won't block threads in the default pool since DBAction uses my separate slick-context execution context
  def recipes = DBAction { implicit rs =>
    val recipes  = Query(Recipes).list
    Ok(recipes.mkString)
  }

}

For certain controller actions I want to be able to utilise SecureSocial's Actions (SecuredAction, UserAwareAction etc) in conjunction with play-slick's DBAction. What is the best way to combine the two?

I realise I can do something like the below, but my understanding is that the DB call won't use my separate slick-context and will therefore block the default thread pool:

Example Two - Using SecureSocial's Action:

import play.api.db.slick._
import securesocial.core._
object Application extends Controller{ 

  // changing from a DBAction to a SecuredAction so that I can use SS's goodies
  def recipes = SecuredAction { implicit request =>
    val recipes  =  DB.withSession { implicit session:Session => Query(Recipes).list } // i'm guessing this WILL BLOCK the default thread pool since it isn't using my separate slick-context execution context??
    Ok(recipes.mkString)
  }

}

Am I correct in assuming that Example Two will use/block the default thread pool instead of my separate slick-context thread pool? If so, is there a way to change this?

I could obviously get around this by bumping up Play's default thread pool (default-dispatcher), but ideally I want to keep the default thread pool quite lean, and run all blocking DB calls in a separate pool.

Assistance appreciated!

like image 243
dbau Avatar asked Nov 05 '13 01:11

dbau


1 Answers

To answer your question,

Am I correct in assuming that Example Two will use/block the default thread pool instead of my separate slick-context thread pool? If so,

Yes, that would use up/block the default pool.

If you want to use the separate slick-context thread pool, then you could try something like this?

  import scala.concurrent.Future

  // Note the use of '.async' |
  //                          V
  def recipes = SecuredAction.async { implicit request =>
    Future { // your code that may block
      val recipes  =  DB.withSession { implicit s:Session => 
        Query(Recipes).list 
      } 
      Ok(recipes.mkString)
    } 
  }

Future expects an ExecutionContext (an implicit will do); all you need to to pass in the one that play-slick uses (implicitly):

import play.api._
implicit val slickExecutionContext = 
  Akka.system.dispatchers.lookup("play.akka.actor.slick-context")
like image 131
Faiz Avatar answered Nov 10 '22 01:11

Faiz