【发布时间】：2011-08-16 15:15:29
【问题描述】：
我正在尝试在 akka Actors 中获得容错行为。我正在编写一些代码，这些代码依赖于系统中的 Actor 可用于长期处理。我发现我的处理在几个小时后停止（大约需要 10 小时）并且没有发生太多事情。我相信我的 Actor 没有从异常中恢复。
我需要做什么才能让 Actor 以一对一的方式永久重启？我希望这可以从这个文档中完成http://akka.io/docs/akka/1.1.3/scala/fault-tolerance
我正在使用 akka 1.1.3 和 scala 2.9
import akka.actor.Actor
import akka.actor.Actor._
import akka.actor.ActorRef
import akka.actor.MaximumNumberOfRestartsWithinTimeRangeReached
import akka.dispatch.Dispatchers
import akka.routing.CyclicIterator
import akka.routing.LoadBalancer
import akka.config.Supervision._
object TestActor {
val dispatcher = Dispatchers.newExecutorBasedEventDrivenWorkStealingDispatcher("pool")
.setCorePoolSize(100)
.setMaxPoolSize(100)
.build
}
class TestActor(val name: Integer) extends Actor {
self.lifeCycle = Permanent
self.dispatcher = TestActor.dispatcher
def receive = {
case num: Integer => {
if( num % 2 == 0 )
throw new Exception("This is a simulated failure")
println("Actor: " + name + " Received: " + num)
//Thread.sleep(100)
}
}
override def postStop(){
println("TestActor post Stop ")
}
//callback method for restart handling
override def preRestart(reason: Throwable){
println("TestActor "+ name + " restaring after shutdown because of " + reason)
}
//callback method for restart handling
override def postRestart(reason: Throwable){
println("Restaring TestActor "+name+"after shutdown because of " + reason)
}
}
trait CyclicLoadBalancing extends LoadBalancer { this: Actor =>
val testActors: List[ActorRef]
val seq = new CyclicIterator[ActorRef](testActors)
}
trait TestActorManager extends Actor {
self.lifeCycle = Permanent
self.faultHandler = OneForOneStrategy(List(classOf[Exception]), 1000, 5000)
val testActors: List[ActorRef]
override def preStart = testActors foreach { self.startLink(_) }
override def postStop = { System.out.println("postStop") }
}
object FaultTest {
def main(args : Array[String]) : Unit = {
println("starting FaultTest.main()")
val numOfActors = 5
val supervisor = actorOf(
new TestActorManager with CyclicLoadBalancing {
val testActors = (0 until numOfActors toList) map (i => actorOf(new TestActor(i)));
}
)
supervisor.start();
println("Number of Actors: " + Actor.registry.actorsFor(classOf[TestActor]).length)
val testActor = Actor.registry.actorsFor(classOf[TestActor]).head
(1 until 200 toList) foreach { testActor ! _ }
}
}
这段代码在 LoadBalancer 后面设置了 5 个 Actor，它们只打印发送给它们的整数，除了它们在偶数上抛出异常以模拟错误。整数 0 到 200 被发送到这些 Actor。我希望奇数会得到输出，但在偶数出现几次错误后，一切似乎都关闭了。使用 sbt 运行此代码会产生以下输出：
[info] Running FaultTest
starting FaultTest.main()
Loading config [akka.conf] from the application classpath.
Number of Actors: 5
Actor: 2 Received: 1
Actor: 2 Received: 9
Actor: 1 Received: 3
Actor: 3 Received: 7
[info] == run ==
[success] Successful.
[info]
[info] Total time: 13 s, completed Aug 16, 2011 11:00:23 AM
我认为这里发生的是 5 个演员开始，前 5 个偶数使他们停业，他们没有重新开始。
如何更改此代码以使 Actor 从异常中恢复？
我希望这实际上会打印出从 1 到 200 的所有奇数。我认为每个参与者都会在偶数上失败，但在异常时会以完整的邮箱重新启动。我希望从 preRestart 和 postRestart 中看到 println。需要在此代码示例中进行哪些配置才能使这些事情发生？
这里有一些关于 akka 和 Actors 的额外假设可能会导致我的误解。我假设一个 Actor 可以配置一个 Supervisor 或一个 faultHandler ，以便在接收期间抛出异常时它将重新启动并继续可用。我假设如果在接收期间抛出异常，发送给参与者的消息将丢失。我假设将调用引发异常的actor上的preRestart()和postRestart()。
代码示例代表我正在尝试做的事情，并且基于Why is my Dispatching on Actors scaled down in Akka?
** 另一个代码示例 **
这是另一个更简单的代码示例。我正在启动一个在偶数上抛出异常的演员。没有负载均衡器或其他东西。我正在尝试打印有关该演员的信息。在将消息发送到 Actor 并监视正在发生的事情之后，我正在等待退出程序一分钟。
我希望这会打印出奇数，但看起来 Actor 坐在那里，邮箱里有消息。
我的 OneForOneStrategy 设置有误吗？我需要将 Actor 链接到某些东西吗？这种配置是否从根本上误导了我？调度程序是否需要以某种方式设置容错？我会弄乱 Dispatcher 中的线程吗？
import akka.actor.Actor
import akka.actor.Actor._
import akka.actor.ActorRef
import akka.actor.ActorRegistry
import akka.config.Supervision._
class SingleActor(val name: Integer) extends Actor {
self.lifeCycle = Permanent
self.faultHandler = OneForOneStrategy(List(classOf[Exception]), 30, 1000)
def receive = {
case num: Integer => {
if( num % 2 == 0 )
throw new Exception("This is a simulated failure, where does this get logged?")
println("Actor: " + name + " Received: " + num)
}
}
override def postStop(){
println("TestActor post Stop ")
}
override def preRestart(reason: Throwable){
println("TestActor "+ name + " restaring after shutdown because of " + reason)
}
override def postRestart(reason: Throwable){
println("Restaring TestActor "+name+"after shutdown because of " + reason)
}
}
object TestSingleActor{
def main(args : Array[String]) : Unit = {
println("starting TestSingleActor.main()")
val testActor = Actor.actorOf( new SingleActor(1) ).start()
println("number of actors: " + registry.actors.size)
printAllActorsInfo
(1 until 20 toList) foreach { testActor ! _ }
for( i <- 1 until 120 ){
Thread.sleep(500)
printAllActorsInfo
}
}
def printAllActorsInfo() ={
registry.actors.foreach( (a) =>
println("Actor hash: %d has mailbox %d isRunning: %b isShutdown: %b isBeingRestarted: %b "
.format(a.hashCode(),a.mailboxSize,a.isRunning,a.isShutdown,a.isBeingRestarted)))
}
}
我得到如下输出：
[info] Running TestSingleActor
starting TestSingleActor.main()
Loading config [akka.conf] from the application classpath.
number of actors: 1
Actor hash: -1537745664 has mailbox 0 isRunning: true isShutdown: false isBeingRestarted: false
Actor: 1 Received: 1
Actor hash: -1537745664 has mailbox 17 isRunning: true isShutdown: false isBeingRestarted: false
... 117 more of these lines repeted ...
Actor hash: -1537745664 has mailbox 17 isRunning: true isShutdown: false isBeingRestarted: false
[info] == run ==
[success] Successful.
[info]
[info] Total time: 70 s, completed Aug 17, 2011 2:24:49 PM
【问题讨论】：