forked from apache/openwhisk
-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
mesos-specific changes for invoker-backpressure
- Loading branch information
1 parent
a2a1844
commit dc127ab
Showing
9 changed files
with
848 additions
and
132 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
293 changes: 293 additions & 0 deletions
293
common/scala/src/main/scala/org/apache/openwhisk/core/mesos/MesosClusterData.scala
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,293 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one or more | ||
| * contributor license agreements. See the NOTICE file distributed with | ||
| * this work for additional information regarding copyright ownership. | ||
| * The ASF licenses this file to You under the Apache License, Version 2.0 | ||
| * (the "License"); you may not use this file except in compliance with | ||
| * the License. You may obtain a copy of the License at | ||
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * Unless required by applicable law or agreed to in writing, software | ||
| * distributed under the License is distributed on an "AS IS" BASIS, | ||
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| * See the License for the specific language governing permissions and | ||
| * limitations under the License. | ||
| */ | ||
|
|
||
| package org.apache.openwhisk.core.mesos | ||
| import akka.actor.Actor | ||
| import akka.actor.ActorRef | ||
| import akka.actor.ActorSystem | ||
| import akka.actor.PoisonPill | ||
| import akka.actor.Props | ||
| import akka.cluster.Cluster | ||
| import akka.cluster.ClusterEvent.MemberEvent | ||
| import akka.cluster.ClusterEvent.MemberRemoved | ||
| import akka.cluster.ClusterEvent.MemberUp | ||
| import akka.cluster.ClusterEvent.UnreachableMember | ||
| import akka.cluster.MemberStatus | ||
| import akka.cluster.UniqueAddress | ||
| import akka.cluster.ddata.DistributedData | ||
| import akka.cluster.ddata.LWWRegister | ||
| import akka.cluster.ddata.LWWRegisterKey | ||
| import akka.cluster.ddata.ORSet | ||
| import akka.cluster.ddata.ORSetKey | ||
| import akka.cluster.ddata.Replicator.Changed | ||
| import akka.cluster.ddata.Replicator.Get | ||
| import akka.cluster.ddata.Replicator.GetFailure | ||
| import akka.cluster.ddata.Replicator.GetSuccess | ||
| import akka.cluster.ddata.Replicator.NotFound | ||
| import akka.cluster.ddata.Replicator.ReadLocal | ||
| import akka.cluster.ddata.Replicator.Subscribe | ||
| import akka.cluster.ddata.Replicator.Update | ||
| import akka.cluster.ddata.Replicator.WriteLocal | ||
| import akka.cluster.singleton.ClusterSingletonManager | ||
| import akka.cluster.singleton.ClusterSingletonManagerSettings | ||
| import akka.cluster.singleton.ClusterSingletonProxy | ||
| import akka.cluster.singleton.ClusterSingletonProxySettings | ||
| import akka.management.AkkaManagement | ||
| import akka.management.cluster.bootstrap.ClusterBootstrap | ||
| import akka.pattern.ask | ||
| import akka.util.Timeout | ||
| import com.adobe.api.platform.runtime.mesos.DistributedDataTaskStore | ||
| import com.adobe.api.platform.runtime.mesos.MesosAgentStats | ||
| import com.adobe.api.platform.runtime.mesos.MesosClient | ||
| import com.adobe.api.platform.runtime.mesos.SubscribeComplete | ||
| import java.time.Instant | ||
| import org.apache.openwhisk.common.Logging | ||
| import org.apache.openwhisk.common.TransactionId | ||
| import org.apache.openwhisk.core.entity.UUID | ||
| import scala.concurrent.Await | ||
| import scala.concurrent.Promise | ||
| import scala.concurrent.TimeoutException | ||
| import scala.concurrent.duration._ | ||
| import scala.util.Try | ||
|
|
||
| //events | ||
| case class SetClusterFrameworkId(frameworkId: String) | ||
| case class SetClusterStats(stats: MesosAgentStats) | ||
| case class AddTask(taskId: String) | ||
| case class RemoveTask(taskId: String) | ||
| case class GetTasks(address: UniqueAddress) | ||
| case class CleanTasks(address: UniqueAddress) | ||
|
|
||
| //data | ||
| case class MemberTask(address: UniqueAddress, taskId: String) | ||
|
|
||
| class MesosClusterData(actorSystem: ActorSystem, mesosConfig: MesosConfig, logging: Logging) extends MesosData { | ||
| override val autoSubscribe: Boolean = true | ||
| var frameworkId: Option[String] = None | ||
| implicit val cluster = Cluster(actorSystem) | ||
| implicit val ec = actorSystem.dispatcher | ||
| implicit val log = logging | ||
| val statsTopic = "agentStats" | ||
| val dataManager = actorSystem.actorOf(Props(new MesosClusterListener(this))) | ||
| var mesosClientActor: Option[ActorRef] = None | ||
| implicit val tid = TransactionId("mesoscluster") | ||
| private val initPromise: Promise[ActorRef] = Promise() | ||
|
|
||
| if (mesosConfig.useClusterBootstrap) { | ||
| AkkaManagement(actorSystem).start() | ||
| ClusterBootstrap(actorSystem).start() | ||
| } | ||
|
|
||
| if (cluster.selfMember.status == MemberStatus.Up) { | ||
| completeInit() | ||
| } else { | ||
| logging.info(this, "waiting for cluster join...") | ||
| } | ||
|
|
||
| //singleton for cleaning tasks | ||
| actorSystem.actorOf( | ||
| ClusterSingletonManager | ||
| .props( | ||
| Props(new Actor { | ||
| var cleanedAddresses | ||
| : Map[UniqueAddress, Instant] = Map.empty //track addresses and expiration time (will prune periodically to avoid growing the map) | ||
| case object PruneCleaned | ||
| //periodically prune cleaned addresses | ||
| context.system.scheduler.schedule(10.seconds, 300.seconds, self, PruneCleaned) | ||
|
|
||
| override def receive: Receive = { | ||
| case CleanTasks(address) => | ||
| //only clean once; but all nodes must notify, so we track which ones have been cleaned | ||
| if (!cleanedAddresses.contains(address)) { | ||
| cleanedAddresses = cleanedAddresses + (address -> Instant.now().plusSeconds(300)) | ||
| removeTasks(address) | ||
| } | ||
| case PruneCleaned => | ||
| val now = Instant.now() | ||
| cleanedAddresses = cleanedAddresses.filter(_._2.isBefore(now)) | ||
| } | ||
| }), | ||
| terminationMessage = PoisonPill, | ||
| settings = ClusterSingletonManagerSettings(actorSystem)), | ||
| name = "invokerCleanup") | ||
| val cleaner = actorSystem.actorOf( | ||
| ClusterSingletonProxy | ||
| .props(singletonManagerPath = "/user/invokerCleanup", settings = ClusterSingletonProxySettings(actorSystem)), | ||
| name = "invokerCleanupProxy") | ||
|
|
||
| //every member listens for cluster events and sends cleanup messages to the singleton cleaner (delete only once) | ||
| actorSystem.actorOf(Props(new Actor { | ||
| override def preStart(): Unit = { | ||
| cluster.subscribe(self, classOf[MemberEvent]) | ||
| } | ||
| override def receive: Receive = { | ||
| case MemberUp(member) => | ||
| if (cluster.selfUniqueAddress == member.uniqueAddress) { | ||
| completeInit() | ||
| } | ||
| case MemberRemoved(member, _) => | ||
| cleaner ! CleanTasks(member.uniqueAddress) | ||
| case UnreachableMember(member) => | ||
| cleaner ! CleanTasks(member.uniqueAddress) | ||
| case m => | ||
| logging.debug(this, s"unknown message $m") | ||
| } | ||
| })) | ||
|
|
||
| private def completeInit() = { | ||
| val client = createClient() | ||
| mesosClientActor = Some(client) | ||
| initPromise.success(client) | ||
| } | ||
|
|
||
| private def removeTasks(addr: UniqueAddress) = { | ||
| //if client was not initialized, we will not even try to delete tasks | ||
| mesosClientActor.foreach { mesosClient => | ||
| dataManager | ||
| .ask(GetTasks(addr))(Timeout(5.minutes)) | ||
| .mapTo[Set[MemberTask]] | ||
| .map { tasks => | ||
| if (tasks.nonEmpty) { | ||
| logging.info(this, s"deleting ${tasks.size} orphaned meesos tasks abandoned by invoker at ${addr}") | ||
| tasks.foreach { t => | ||
| MesosContainerFactory | ||
| .destroy(mesosClient, mesosConfig, this, t.taskId)(tid, logging, ec) | ||
| .map(_ => dataManager ! RemoveTask(t.taskId)) | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
| } | ||
| def createClient(): ActorRef = { | ||
| //this listener handles propagating MesosAgentState to NodeStatsUpdateEvent | ||
| val statsListener = actorSystem.actorOf(Props(new MesosClusterStatsListener())) | ||
|
|
||
| val clientProps = MesosClient.props( | ||
| getOrInitFrameworkId _, | ||
| "whisk-containerfactory-framework", | ||
| mesosConfig.masterUrl, | ||
| mesosConfig.role, | ||
| mesosConfig.timeouts.failover, | ||
| taskStore = new DistributedDataTaskStore(actorSystem), //must use DistributedTaskStore for cluster failover reconciliation | ||
| refuseSeconds = mesosConfig.offerRefuseDuration.toSeconds.toDouble, | ||
| heartbeatMaxFailures = mesosConfig.heartbeatMaxFailures, | ||
| autoSubscribe = true, //must be true for cluster usage so when singleton client fails, new one will re-subscribe automatically | ||
| listener = Some(statsListener)) | ||
|
|
||
| //create the MesosClient actor as a cluster singleton | ||
| actorSystem.actorOf( | ||
| ClusterSingletonManager | ||
| .props(clientProps, terminationMessage = PoisonPill, settings = ClusterSingletonManagerSettings(actorSystem)), | ||
| name = "mesosClientMaster") | ||
| actorSystem.actorOf( | ||
| ClusterSingletonProxy | ||
| .props(singletonManagerPath = "/user/mesosClientMaster", settings = ClusterSingletonProxySettings(actorSystem)), | ||
| name = "mesosClientProxy") | ||
| } | ||
|
|
||
| private def getOrInitFrameworkId(): String = { | ||
| frameworkId.getOrElse { | ||
| //create the framework id | ||
| val newFrameworkId = "whisk-containerfactory-" + UUID() | ||
| //save the framework id | ||
| frameworkId = Some(newFrameworkId) | ||
| logging.info(this, s"created new framework id $newFrameworkId") | ||
| //replicate the frameworkd id | ||
| dataManager ! SetClusterFrameworkId(newFrameworkId) | ||
| newFrameworkId | ||
| } | ||
| } | ||
| override def addTask(taskId: String): Unit = { | ||
| super.addTask(taskId) | ||
| dataManager ! AddTask(taskId) | ||
| } | ||
| override def removeTask(taskId: String): Unit = { | ||
| super.removeTask(taskId) | ||
| dataManager ! RemoveTask(taskId) | ||
| } | ||
|
|
||
| class MesosClusterStatsListener extends Actor { | ||
| override def receive: Receive = { | ||
| case a: MesosAgentStats => | ||
| //publish to cluster | ||
| dataManager ! SetClusterStats(a) | ||
| case SubscribeComplete(f) => | ||
| logging.info(this, s"received framework id $f") | ||
| } | ||
| } | ||
| override def getMesosClient(): ActorRef = { | ||
| mesosClientActor.getOrElse { | ||
| val c = Try(Await.result(initPromise.future, mesosConfig.timeouts.subscribe)) | ||
| .map(client => client) | ||
| .recover { | ||
| case _: TimeoutException => | ||
| logging.error(this, "Mesos akka cluster init took too long. Will retry...") | ||
| getMesosClient() | ||
| } | ||
| c.get | ||
| } | ||
| } | ||
| } | ||
|
|
||
| class MesosClusterListener(clusterData: MesosClusterData)(implicit logging: Logging) extends Actor { | ||
|
|
||
| implicit val node = Cluster(context.system) | ||
| val addr = node.selfUniqueAddress | ||
| val replicator = DistributedData(context.system).replicator | ||
| implicit val selfAddress = DistributedData(context.system).selfUniqueAddress | ||
|
|
||
| val FrameworkIdKey = LWWRegisterKey[String]("mesosFrameworkId") | ||
| val MemberTasksKey = ORSetKey[MemberTask]("mesosTasks") | ||
| //cluster stats are only published by the singleton MesosClient, so we could use DistributedPubSub, | ||
| //but we are already using DistributedData, so we will just use that | ||
| val ClusterStatsKey = LWWRegisterKey[MesosAgentStats]("mesosNodeStats") | ||
|
|
||
| //subscribe to changes on frameworkid and cluster stats | ||
| replicator ! Subscribe(FrameworkIdKey, self) | ||
| replicator ! Subscribe(ClusterStatsKey, self) | ||
| override def receive: Receive = { | ||
| //Replication events | ||
| case SetClusterFrameworkId(frameworkId) => | ||
| replicator ! Update(FrameworkIdKey, LWWRegister[String](selfAddress, null), WriteLocal)(reg => | ||
| reg.withValueOf(frameworkId)) | ||
| case SetClusterStats(stats) => | ||
| replicator ! Update( | ||
| ClusterStatsKey, | ||
| LWWRegister[MesosAgentStats](selfAddress, MesosAgentStats(Map.empty)), | ||
| WriteLocal)(reg => reg.withValueOf(stats)) | ||
| case AddTask(task) => | ||
| replicator ! Update(MemberTasksKey, ORSet.empty[MemberTask], WriteLocal)(_ :+ MemberTask(addr, task)) | ||
| case RemoveTask(task) => | ||
| replicator ! Update(MemberTasksKey, ORSet.empty[MemberTask], WriteLocal)(_.remove(MemberTask(addr, task))) | ||
| //Get Handlers | ||
| case GetTasks(address: UniqueAddress) => | ||
| replicator ! Get(MemberTasksKey, ReadLocal, Some((sender(), address))) | ||
| case g @ GetSuccess(MemberTasksKey, Some((replyTo: ActorRef, address: UniqueAddress))) => | ||
| val value = g.get(MemberTasksKey).elements.filter(_.address == address) | ||
| replyTo ! value | ||
| case NotFound(key, _) => | ||
| logging.warn(this, s"$key not found") | ||
| case f @ GetFailure(key, _) => | ||
| logging.warn(this, s"$key failure $f") | ||
| //Change Handlers | ||
| case c @ Changed(FrameworkIdKey) => | ||
| clusterData.frameworkId = Some(c.get(FrameworkIdKey).value) | ||
| case c @ Changed(ClusterStatsKey) => | ||
| clusterData.publishStats(context.system, c.get(ClusterStatsKey).value) | ||
| } | ||
| } |
Oops, something went wrong.