1. 启动脚本
sbin/start-slaves.sh
# Launch the slaves
if [ "$SPARK_WORKER_INSTANCES" = "" ]; then
exec "$sbin/slaves.sh" cd "$SPARK_HOME" \; "$sbin/start-slave.sh" 1 "spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT"
else
if [ "$SPARK_WORKER_WEBUI_PORT" = "" ]; then
SPARK_WORKER_WEBUI_PORT=8081
fi
for ((i=0; i<$SPARK_WORKER_INSTANCES; i++)); do
"$sbin/slaves.sh" cd "$SPARK_HOME" \; "$sbin/start-slave.sh" $(( $i + 1 )) "spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT" --webui-port $(( $SPARK_WORKER_WEBUI_PORT + $i ))
done
fi
假设每个节点启动一个Worker。
具体执行:
exec "$sbin/slaves.sh" cd "$SPARK_HOME" \; "$sbin/start-slave.sh" 1 "spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT"
该语句分为两部分:
(1)
exec "$sbin/slaves.sh" cd "$SPARK_HOME"
登录到worker服务器并cd到SPARK_HOME目录。
(2)
"$sbin/start-slave.sh" 1 "spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT"
在worker服务器执行sbin/start-slave.sh脚本。
参数“1”代码worker的编号,用来区分不同worker实例的日志文件。如:
spark-xxx-org.apache.spark.deploy.worker.Worker-1-CentOS-02.out
spark-xxx-org.apache.spark.deploy.worker.Worker-1.pid
其中“Worker-1”中的“1”就代表worker编号。
这个参数并不会传入Worker类。传入Worker类的参数为:
spark://$SPARK_MASTER_IP:$SPARK_MASTER_PORT。
2. Worker.main
def main(argStrings: Array[String]) {
SignalLogger.register(log)
val conf = new SparkConf
val args = new WorkerArguments(argStrings, conf)
val (actorSystem, _) = startSystemAndActor(args.host, args.port, args.webUiPort, args.cores,
args.memory, args.masters, args.workDir)
actorSystem.awaitTermination()
}
main函数的职责:
(1)创建WorkerArguments对象并初始化其成员;
(2)调用startSystemAndActor方法,创建ActorSystem对象并启动Worker actor;
2.1. WorkerArguments
var cores = inferDefaultCores()
var memory = inferDefaultMemory()
(1)计算默认核数
(2)计算默认内存大小
parse(args.toList)
// This mutates the SparkConf, so all accesses to it must be made after this line
propertiesFile = Utils.loadDefaultSparkProperties(conf, propertiesFile)
(1)parse方法负责解析启动脚本所带的命令行参数;
(2)loadDefaultSparkProperties负责从配置文件中加载spark运行属性,默认而配置文件为spark-defaults.conf;
2.2. startSystemAndActor
val (actorSystem, boundPort) = AkkaUtils.createActorSystem(systemName, host, port,
conf = conf, securityManager = securityMgr)
val masterAkkaUrls = masterUrls.map(Master.toAkkaUrl(_, AkkaUtils.protocol(actorSystem)))
actorSystem.actorOf(Props(classOf[Worker], host, boundPort, webUiPort, cores, memory,
masterAkkaUrls, systemName, actorName, workDir, conf, securityMgr), name = actorName)
(1)通过AkkaUtils.createActorSystem创建ActorSystem对象
(2)创建Worker actor并启动
3. Worker Actor
3.1. 重要数据成员
val executors = new HashMap[String, ExecutorRunner]
val finishedExecutors = new HashMap[String, ExecutorRunner]
val drivers = new HashMap[String, DriverRunner]
val finishedDrivers = new HashMap[String, DriverRunner]
val appDirectories = new HashMap[String, Seq[String]]
val finishedApps = new HashSet[String]
3.2. Worker.preStart
createWorkDir()
context.system.eventStream.subscribe(self, classOf[RemotingLifecycleEvent])
shuffleService.startIfEnabled()
webUi = new WorkerWebUI(this, workDir, webUiPort)
webUi.bind()
registerWithMaster()
(1)创建Worker节点工作目录;
(2)监听RemotingLifecycleEvent事件,它一个trait:
sealed trait RemotingLifecycleEvent extends Serializable {
def logLevel: Logging.LogLevel
}
Worker只处理了DisassociatedEvent消息。
(3)创建并启动WorkerWebUI
(4)向Master进行注册,registerWithMaster将调用tryRegisterAllMasters方法向Master节点发送注册消息
3.3. Worker.registerWithMaster
registrationRetryTimer match {
case None =>
registered = false
tryRegisterAllMasters()
connectionAttemptCount = 0
registrationRetryTimer = Some {
context.system.scheduler.schedule(INITIAL_REGISTRATION_RETRY_INTERVAL,
INITIAL_REGISTRATION_RETRY_INTERVAL, self, ReregisterWithMaster)
}
case Some(_) =>
logInfo("Not spawning another attempt to register with the master, since there is an" +
" attempt scheduled already.")
}
(1)调用tryRegisterAllMasters方法向Master发起注册消息;
(2)创建注册重试定时器,通过向自己(Worker Actor)发送ReregisterWithMaster消息;
3.3.1. Worker.tryRegisterAllMasters
for (masterAkkaUrl <- masterAkkaUrls) {
logInfo("Connecting to master " + masterAkkaUrl + "...")
val actor = context.actorSelection(masterAkkaUrl)
actor ! RegisterWorker(workerId, host, port, cores, memory, webUi.boundPort, publicAddress)
}
(1)创建Master Actor远程引用;
(2)向Master发送RegisterWorker消息;如果注册成功,Master将向Worker发送RegisteredWorker消息。
workerId是一个字符串,定义:
val workerId = generateWorkerId()
...
def generateWorkerId(): String = {
"worker-%s-%s-%d".format(createDateFormat.format(new Date), host, port)
}
格式:worker-时间-主机名-端口
3.4. Worker消息处理
3.4.1. RegisteredWorker消息
此消息表示Worker向Master注册成功消息;该消息处理的主要目的是启动心跳发送定时器。
case RegisteredWorker(masterUrl, masterWebUiUrl) =>
logInfo("Successfully registered with master " + masterUrl)
registered = true
changeMaster(masterUrl, masterWebUiUrl)
context.system.scheduler.schedule(0 millis, HEARTBEAT_MILLIS millis, self, SendHeartbeat)
if (CLEANUP_ENABLED) {
logInfo(s"Worker cleanup enabled; old application directories will be deleted in: $workDir")
context.system.scheduler.schedule(CLEANUP_INTERVAL_MILLIS millis,
CLEANUP_INTERVAL_MILLIS millis, self, WorkDirCleanup)
}
(1)设置注册状态;
(2)调用changeMaster方法
(3)创建心跳发送定时器,向自己(Worker Actor)发送SendHeartbeat消息;
3.4.1.1. Worker.changeMaster
// activeMasterUrl it‘s a valid Spark url since we receive it from master.
activeMasterUrl = url
activeMasterWebUiUrl = uiUrl
master = context.actorSelection(
Master.toAkkaUrl(activeMasterUrl, AkkaUtils.protocol(context.system)))
masterAddress = Master.toAkkaAddress(activeMasterUrl, AkkaUtils.protocol(context.system))
connected = true
// Cancel any outstanding re-registration attempts because we found a new master
registrationRetryTimer.foreach(_.cancel())
registrationRetryTimer = None
职责:
(1)创建Master远程引用并赋值给master;
(2)将连接状态设置为true;
(3)取消registrationRetryTimer定时器;
3.4.2. SendHeartbeat消息
case SendHeartbeat =>
if (connected) { master ! Heartbeat(workerId) }
向master发送Heartbeat消息。
3.4.3. ReregisterWithMaster消息
case ReregisterWithMaster =>
reregisterWithMaster()
reregisterWithMaster方法职责:
(1)如果已经注册成功,取消registrationRetryTimer定时器;
(2)如果注册失败,从新向master发送RegisterWorker消息;初始默认重连次数为6,最大重连次数为16。
// The first six attempts to reconnect are in shorter intervals (between 5 and 15 seconds)
// Afterwards, the next 10 attempts are between 30 and 90 seconds.
// A bit of randomness is introduced so that not all of the workers attempt to reconnect at
// the same time.
val INITIAL_REGISTRATION_RETRIES = 6
val TOTAL_REGISTRATION_RETRIES = INITIAL_REGISTRATION_RETRIES + 10
前6次和后10次采用不同的周期。
4. 启动结束
到此,Worker节点就启动完成,它定时向Master节点发送心跳。在SparkSubmit提交Application时,将接收Master发送的启动Executor消息,由Executor和Driver进行消息通信。