31 Spark资源调度

最新推荐文章于 2023-04-12 23:45:59 发布

喜欢较真的小金

最新推荐文章于 2023-04-12 23:45:59 发布

阅读量982

点赞数

分类专栏：大数据文章标签： spark 大数据 scala Spark笔记-IFM课程集群

本文链接：https://blog.csdn.net/sinat_25306771/article/details/51474572

版权

大数据专栏收录该内容

22 篇文章 2 订阅

订阅专栏

内容：

1.分配Driver（Cluster模式）

2.为Application分配资源

3.两种不同的资源分配方式

4.Spark资源分配的思考

一、任务调度与资源调度的区别

1.任务调度

任务调度是（指Task set怎么）通过DAGScheduler、TaskScheduler、SchedulerBackend等进行（运行）的。

2.资源调度

资源调度是指应用程序如何获得资源的。

3.关系

任务调度是在资源调度的基础上进行的，没有资源调度，那么任务调度就成了无源之水。

二、资源调度

1.Master负责资源调度

因为Master负责资源管理与调度，所以资源调度的方法Scheduler位于Master.scala这个类中，当注册过的程序或资源发生改变时会导致Schedule的调用。例如注册程序时：

case RegisterApplication(description, driver) => {
  // TODO Prevent repeated registrations from some driver
  if (state == RecoveryState.STANDBY) {
    // ignore, don't send response
  } else {
    logInfo("Registering app " + description.name)
    val app = createApplication(description, driver)
    registerApplication(app)
    logInfo("Registered app " + description.name + " with ID " + app.id)
    persistenceEngine.addApplication(app)
    driver.send(RegisteredApplication(app.id, self))
    schedule()
  }

2.scheduler调用的时机

Scheduler调用的时机，每次有新的应用程序提交或者集群的资源状况发生改变时（包括Executor增加或减少、worker增加或减少等）。

3.Master可调度的条件

当前Master必须是Alive的方式才能进行资源的调度，若不是alive状态就会直接返回，也就是说standby Master是不会进行Application的资源调度。

private def schedule(): Unit = {
  if (state != RecoveryState.ALIVE) { return }
  // Drivers take strict precedence over executors
  val shuffledWorkers = Random.shuffle(workers) // Randomization helps balance drivers
  for (worker <- shuffledWorkers if worker.state == WorkerState.ALIVE) {
    for (driver <- waitingDrivers) {
      if (worker.memoryFree >= driver.desc.mem && worker.coresFree >= driver.desc.cores) {
        launchDriver(worker, driver)
        waitingDrivers -= driver
      }
    }
  }
  startExecutorsOnWorkers()
}

4.Worker信息的处理

使用Random.shuffle把Master中保存的集群中所有worker的信息随机打乱。

private def schedule(): Unit = {
  if (state != RecoveryState.ALIVE) { return }
  // Drivers take strict precedence over executors
  val shuffledWorkers = Random.shuffle(workers) // Randomization helps balance drivers
  for (worker <- shuffledWorkers if worker.state == WorkerState.ALIVE) {
    for (driver <- waitingDrivers) {
      if (worker.memoryFree >= driver.desc.mem && worker.coresFree >= driver.desc.cores) {
        launchDriver(worker, driver)
        waitingDrivers -= driver
      }
    }
  }
  startExecutorsOnWorkers()
}

其算法内部是循环随机交换所有的worker在Master的缓存数据结构中的位置。

/** Returns a new collection of the same type in a randomly chosen order.
 *
 *  @return         the shuffled collection
 */
def shuffle[T, CC[X] <: TraversableOnce[X]](xs: CC[T])(implicit bf: CanBuildFrom[CC[T], T, CC[T]]): CC[T] = {
  val buf = new ArrayBuffer[T] ++= xs

  def swap(i1: Int, i2: Int) {
    val tmp = buf(i1)
    buf(i1) = buf(i2)
    buf(i2) = tmp
  }

  for (n <- buf.length to 2 by -1) {
    val k = nextInt(n)
    swap(n - 1, k)
  }

  (bf(xs) ++= buf).result
}

5.判断worker的状态

接着，会判断所有worker中哪些是Alive级别的worker，Alive级别的worker才能参与资源的分配工作。

private def schedule(): Unit = {
  if (state != RecoveryState.ALIVE) { return }
  // Drivers take strict precedence over executors
  val shuffledWorkers = Random.shuffle(workers) // Randomization helps balance drivers
  for (worker <- shuffledWorkers if worker.state == WorkerState.ALIVE) {
    for (driver <- waitingDrivers) {
      if (worker.memoryFree >= driver.desc.mem && worker.coresFree >= driver.desc.cores) {
        launchDriver(worker, driver)
        waitingDrivers -= driver
      }
    }
  }
  startExecutorsOnWorkers()
}

6.Cluster模式下的提交应用程序

当SparkSubmit时，指定Driver是Cluster模式情况下，此时Driver会加入waitingDrivers等待列表中，在每个DriverInfo数据结构中的DriverDescription里有保存着启动Driver时对Worker的内存和Cores的要求等内容。

private[deploy] class DriverInfo(
    val startTime: Long,
    val id: String,
    val desc: DriverDescription,
    val submitDate: Date)
  extends Serializable {


private[deploy] case class DriverDescription(
    jarUrl: String,
    mem: Int,
    cores: Int,
    supervise: Boolean,
    command: Command) {

  override def toString: String = s"DriverDescription (${command.mainClass})"
}

在符合资源要求的情况下，会选取随机打乱后的Worker集合中的一个worker来启动Driver。

private def schedule(): Unit = {
  if (state != RecoveryState.ALIVE) { return }
  // Drivers take strict precedence over executors
  val shuffledWorkers = Random.shuffle(workers) // Randomization helps balance drivers
  for (worker <- shuffledWorkers if worker.state == WorkerState.ALIVE) {
    for (driver <- waitingDrivers) {
      if (worker.memoryFree >= driver.desc.mem && worker.coresFree >= driver.desc.cores) {
        launchDriver(worker, driver)
        waitingDrivers -= driver
      }
    }
  }
  startExecutorsOnWorkers()
}

Master发指令给worker让远程的worker启动Driver；

launchDriver(worker, driver)

 private def launchDriver(worker: WorkerInfo, driver: DriverInfo) {
  logInfo("Launching driver " + driver.id + " on worker " + worker.id)
  worker.addDriver(driver)
  driver.worker = Some(worker)
  worker.endpoint.send(LaunchDriver(driver.id, driver.desc))
  driver.state = DriverState.RUNNING
}

7.资源调度开始

先启动Driver才会发生后续的一切资源调度模式。

8.资源调度模式（StandAlone）

Spark默认为应用程序启动Executor的方式是FIFO的方式，也就是说所有的提交的应用程序都是放在调度队列中的，先进先出，只有满足了前面应用程序的资源分配的基础上，才能够为下一个应用程序进行资源分配。

9.Executor及Cores分配过程

(1)判断应用程序资源是否以满足

首先，在位应用程序分配Executor之前需要进行判断，应用程序是否还需要分配Cores，如果不需要则不会为应用程序分配Executor；

(2)分配条件

在具体分配Executor之前要求Worker的状态必须是ALIVE且必须满足Application对每个Executor的内存和Cores的要求，并且在此基础上进行排序产生计算资源有小到大的useableWorkers数据结构；（Master.scala）

private def startExecutorsOnWorkers(): Unit = {
  // Right now this is a very simple FIFO scheduler. We keep trying to fit in the first app
  // in the queue, then the second app, etc.
  for (app <- waitingApps if app.coresLeft > 0) {
    val coresPerExecutor: Option[Int] = app.desc.coresPerExecutor
    
    val usableWorkers = workers.toArray.filter(_.state == WorkerState.ALIVE)
      .filter(worker => worker.memoryFree >= app.desc.memoryPerExecutorMB &&
        worker.coresFree >= coresPerExecutor.getOrElse(1))
      .sortBy(_.coresFree).reverse
    val assignedCores = scheduleExecutorsOnWorkers(app, usableWorkers, spreadOutApps)

    // Now that we've decided how many cores to allocate on each worker, let's allocate them
    for (pos <- 0 until usableWorkers.length if assignedCores(pos) > 0) {
      allocateWorkerResourceToExecutors(
        app, assignedCores(pos), coresPerExecutor, usableWorkers(pos))
    }
  }

在FIFO的情况下默认是spreadOutApps来让应用程序尽可能多的运行在所有的Node上;

// As a temporary workaround before better ways of configuring memory, we allow users to set
// a flag that will perform round-robin scheduling across the nodes (spreading out each app
// among all the nodes) instead of trying to consolidate each app onto a small # of nodes.
private val spreadOutApps = conf.getBoolean("spark.deploy.spreadOut", true)

(3)分配方式

为应用程序分配Executors有两种方式，第一种方式是尽可能在集群的所有worker上分配Executor，这种方式往往会带来潜在的更好的数据本地性（这种情况下，有利于应用的并发计算）。

具体在集群上分配cores的时候会尽可能的满足我们的要求，所以下面求了一个最小值：

var coresToAssign = math.min(app.coresLeft, usableWorkers.map(_.coresFree).sum)

如果是每个worker下面只能为当前的应用程序分配一个Executor的话，每次是分配一个core。

/**
 * Schedule executors to be launched on the workers.
 * Returns an array containing number of cores assigned to each worker…….
*/
private def scheduleExecutorsOnWorkers(
    app: ApplicationInfo,
    usableWorkers: Array[WorkerInfo],
    spreadOutApps: Boolean): Array[Int] = {
……
// If we are launching one executor per worker, then every iteration assigns 1 core
// to the executor. Otherwise, every iteration assigns cores to a new executor.
if (oneExecutorPerWorker) {
  assignedExecutors(pos) = 1
} else {
  assignedExecutors(pos) += 1
}
……}

(4)Master通知worker

准备好要为当前的应用程序分配的Executor信息后，Master要通过远程通信发指令给Worker来具体启动ExecutorBackend进程，来具体负责启动Executor。

private def startExecutorsOnWorkers(): Unit = { ……
   // Now that we've decided how many cores to allocate on each worker, let's allocate them
    for (pos <- 0 until usableWorkers.length if assignedCores(pos) > 0) {
      allocateWorkerResourceToExecutors(
        app, assignedCores(pos), coresPerExecutor, usableWorkers(pos))
    }
  }

private def allocateWorkerResourceToExecutors(……){
……
for (i <- 1 to numExecutors) {
    val exec = app.addExecutor(worker, coresToAssign)
    launchExecutor(worker, exec)
    app.state = ApplicationState.RUNNING
  }
}

private def launchExecutor(worker: WorkerInfo, exec: ExecutorDesc): Unit = {
  logInfo("Launching executor " + exec.fullId + " on worker " + worker.id)
  worker.addExecutor(exec)
  worker.endpoint.send(LaunchExecutor(masterUrl,
    exec.application.id, exec.id, exec.application.desc, exec.cores, exec.memory))
  exec.application.driver.send(
    ExecutorAdded(exec.id, worker.id, worker.hostPort, exec.cores, exec.memory))
}

(5)通知Driver

如上代码，最后Master会给我们应用程序的Driver发送一个ExecutorAdded的信息。

------------------------------------------EOF--------------------------------------------------------------------------