如何在一个进程上运行一个actor,将消息发送给在另一个进程上运行的另一个actor?

时间:2017-03-15 21:22:05

标签: f# akka.net akka.net-cluster akka.fsharp

我想让各个进程(或节点)上运行的actor向其他运行不同进程(或节点)的actor发送消息,同时保持容错和负载平衡。我目前正在尝试使用Akka.Cluster的Sharding功能来实现这一目标。

但是,我不知道如何做到这一点......

我有以下代码反映了我的种子节点:

let configurePort port =
    let config = Configuration.parse ("""
        akka {
            actor {
              provider = "Akka.Cluster.ClusterActorRefProvider, Akka.Cluster"
              serializers {
                hyperion = "Akka.Serialization.HyperionSerializer, Akka.Serialization.Hyperion"
              }
              serialization-bindings {
                "System.Object" = hyperion
              }
            }
          remote {
            helios.tcp {
              public-hostname = "localhost"
              hostname = "localhost"
              port = """ + port.ToString() + """
            }
          }
          cluster {
            auto-down-unreachable-after = 5s
            seed-nodes = [ "akka.tcp://cluster-system@localhost:2551/" ]
          }
          persistence {
            journal.plugin = "akka.persistence.journal.inmem"
            snapshot-store.plugin = "akka.persistence.snapshot-store.local"
          }
        }
        """)
    config.WithFallback(ClusterSingletonManager.DefaultConfig())

let consumer (actor:Actor<_>) msg = printfn "\n%A received %s" (actor.Self.Path.ToStringWithAddress()) msg |> ignored

// spawn two separate systems with shard regions on each of them
let system1 = System.create "cluster-system" (configurePort 2551)
let shardRegion1 = spawnSharded id system1 "shardRegion1" <| props (actorOf2 consumer)
System.Threading.Thread.Sleep(1000)

let system2 = System.create "cluster-system" (configurePort 2552)
let shardRegion2 = spawnSharded id system2 "shardRegion2" <| props (actorOf2 consumer)
System.Threading.Thread.Sleep(1000)

let system3 = System.create "cluster-system" (configurePort 2553)
let shardRegion3 = spawnSharded id system3 "shardRegion3" <| props (actorOf2 consumer)
System.Threading.Thread.Sleep(3000)


// NOTE: Even thou we sent all messages through single shard region,
//       some of them will be executed on the second and third one thanks to shard balancing
System.Threading.Thread.Sleep(3000)
shardRegion1 <! ("shard-1", "entity-1", "hello world 1")
shardRegion1 <! ("shard-1", "entity-2", "hello world 2")
shardRegion1 <! ("shard-2", "entity-3", "hello world 3")
shardRegion1 <! ("shard-2", "entity-4", "hello world 4")

System.Threading.Thread.Sleep(1000)

let printShards shardRegion =
    async {
        let! (reply:AskResult<ShardRegionStats>) = (retype shardRegion) <? GetShardRegionStats.Instance
        let (stats: ShardRegionStats) = reply.Value
        for kv in stats.Stats do
            printfn "\tShard '%s' has %d entities on it" kv.Key kv.Value
    } |> Async.RunSynchronously

let printNodes() =
    printfn "\nShards active on node 'localhost:2551':"
    printShards shardRegion1
    printfn "\nShards active on node 'localhost:2552':"
    printShards shardRegion2
    printfn "\nShards active on node 'localhost:2553':"
    printShards shardRegion3

printNodes()

输出如下所示:

Shards active on node 'localhost:2551':
    Shard 'shard-1' has 2 entities on it
    Shard 'shard-2' has 2 entities on it

在节点'localhost:2552'上激活的碎片:

然后我有一个执行以下代码的单独进程:

let configurePort port =
    let config = Configuration.parse ("""
        akka {
            actor {
              provider = "Akka.Cluster.ClusterActorRefProvider, Akka.Cluster"
              serializers {
                hyperion = "Akka.Serialization.HyperionSerializer, Akka.Serialization.Hyperion"
              }
              serialization-bindings {
                "System.Object" = hyperion
              }
            }
          remote {
            helios.tcp {
              public-hostname = "localhost"
              hostname = "localhost"
              port = "0"
            }
          }
          cluster {
            auto-down-unreachable-after = 5s
            seed-nodes = [ "akka.tcp://cluster-system@localhost:2551/" ]
          }
          persistence {
            journal.plugin = "akka.persistence.journal.inmem"
            snapshot-store.plugin = "akka.persistence.snapshot-store.local"
          }
        }
        """)
    config.WithFallback(ClusterSingletonManager.DefaultConfig())

let consumer (actor:Actor<_>) msg = printfn "\n%A received %s" (actor.Self.Path.ToStringWithAddress()) msg |> ignored

// spawn two separate systems with shard regions on each of them
let system1 = System.create "cluster-system" (configurePort 2554)
let shardRegion1 = spawnSharded id system1 "printer" <| props (actorOf2 consumer)
System.Threading.Thread.Sleep(1000)

let system2 = System.create "cluster-system" (configurePort 2555)
let shardRegion2 = spawnSharded id system2 "printer" <| props (actorOf2 consumer)
System.Threading.Thread.Sleep(1000)

let system3 = System.create "cluster-system" (configurePort 2556)
let shardRegion3 = spawnSharded id system3 "printer" <| props (actorOf2 consumer)

我的群集系统(在单独的进程上运行)识别正在加入的新节点:

> [INFO][3/15/2017 9:12:13 PM][Thread 0054][[akka://cluster-system/system/cluster/core/daemon#2086121649]] Node [akka.tcp://cluster-system@localhost:52953] is JOINING, roles []
[INFO][3/15/2017 9:12:14 PM][Thread 0006][[akka://cluster-system/system/cluster/core/daemon#2086121649]] Node [akka.tcp://cluster-system@localhost:52956] is JOINING, roles []
[INFO][3/15/2017 9:12:15 PM][Thread 0054][[akka://cluster-system/system/cluster/core/daemon#2086121649]] Node [akka.tcp://cluster-system@localhost:52961] is JOINING, roles []
[INFO][3/15/2017 9:12:18 PM][Thread 0055][[akka://cluster-system/system/cluster/core/daemon#2086121649]] Leader is moving node [akka.tcp://cluster-system@localhost:52953] to [Up]
[INFO][3/15/2017 9:12:18 PM][Thread 0055][[akka://cluster-system/system/cluster/core/daemon#2086121649]] Leader is moving node [akka.tcp://cluster-system@localhost:52956] to [Up]
[INFO][3/15/2017 9:12:18 PM][Thread 0055][[akka://cluster-system/system/cluster/core/daemon#2086121649]] Leader is moving node [akka.tcp://cluster-system@localhost:52961] to [Up]

结论:

总之,我希望让各个进程(或节点)上运行的actor向其他运行不同进程(或节点)的actor发送消息,同时保持容错和负载平衡。我目前正在尝试使用Akka.Cluster的Sharding功能来实现这一目标。

附录

open System
open System.IO
#if INTERACTIVE
let cd = Path.Combine(__SOURCE_DIRECTORY__, "../src/Akkling.Cluster.Sharding/bin/Debug")
System.IO.Directory.SetCurrentDirectory(cd)
#endif

#r "../src/Akkling.Cluster.Sharding/bin/Debug/System.Collections.Immutable.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Hyperion.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Newtonsoft.Json.dll"
#r @"C:\Users\Snimrod\Documents\Visual Studio 2015\Projects\Temp\packages\Akka.FSharp.1.1.3\lib\net45\Akka.FSharp.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/FSharp.PowerPack.Linq.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Helios.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/FsPickler.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Google.ProtocolBuffers.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Google.ProtocolBuffers.Serialization.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.Remote.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Google.ProtocolBuffers.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.Persistence.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.Cluster.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.Cluster.Tools.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.Cluster.Sharding.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akka.Serialization.Hyperion.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akkling.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akkling.Persistence.dll"
#r "../src/Akkling.Cluster.Sharding/bin/Debug/Akkling.Cluster.Sharding.dll"


open Akka.Actor
open Akka.Configuration
open Akka.Cluster
open Akka.Cluster.Tools.Singleton
open Akka.Cluster.Sharding
open Akka.Persistence

open Akkling
open Akkling.Persistence
open Akkling.Cluster
open Akkling.Cluster.Sharding
open Hyperion

1 个答案:

答案 0 :(得分:4)

为了保持分片及其位置的一致视图,Akka.Cluster.Sharding持久后端必须指向所有进程都可见的数据库。在您的配置中,您使用的是akka.persistence.journal.inmem内存数据存储(仅用于测试和开发)。它不会从其他进程中看到。

您需要配置持久后端,以便在生活在不同计算机/进程上的节点之间显示分片。你可以这样做,即使用Akka.Persistence.SqlServer或任何其他插件。这是仅通过分片使用的持久性后端的最基本配置:

akka.persistence {
    journal {
        plugin = "akka.persistence.journal.sql-server"
        sql-server {
            connection-string = "<connection-string>"
            auto-initialize = on
        }
    }
    snapshot-store {
        plugin = "akka.persistence.snapshot-store.sql-server"
        sql-server {
            connection-string = "<connection-string>"
            auto-initialize = on
        }
    }
}

有关更实用的内容,请参阅this article

另请注意,Akka.Cluster.Sharding和Akka.Persistence插件仅在预发布模式下可用(因此您需要使用-pre标志安装包)。