入口
// 最后,针对stage的task,创建TaskSet对象,调用taskScheduler的submitTasks()方法,提交taskSet
// 默认情况下,我们的standalone模式,是使用的TaskSchedulerImpl,TaskScheduler只是一个trait
taskScheduler.submitTasks(
new TaskSet(tasks.toArray, stage.id, stage.newAttemptId(), stage.jobId, properties))
看看taskScheduler.submitTasks()方法,TaskSchedulerImpl的submitTasks()方法
/**
* TaskScheduler提交任务的入口
* @param taskSet
*/
override def submitTasks(taskSet: TaskSet) {
val tasks = taskSet.tasks
logInfo("Adding task set " + taskSet.id + " with " + tasks.length + " tasks")
this.synchronized {
// 给每一个taskSet,都会创建一个TaskSetManager
// TaskSetManager实际上,在后面,会负责他的那个TaskSet的任务执行状况的监视和管理
val manager = createTaskSetManager(taskSet, maxTaskFailures)
// 加入内存缓存中
activeTaskSets(taskSet.id) = manager
schedulableBuilder.addTaskSetManager(manager, manager.taskSet.properties)
if (!isLocal && !hasReceivedTask) {
starvationTimer.scheduleAtFixedRate(new TimerTask() {
override def run() {
if (!hasLaunchedTask) {
logWarning("Initial job has not accepted any resources; " +
"check your cluster UI to ensure that workers are registered " +
"and have sufficient resources")
} else {
this.cancel()
}
}
}, STARVATION_TIMEOUT, STARVATION_TIMEOUT)
}
hasReceivedTask = true
}
// sparkContext原理剖析的时候,创建TaskScheduler的时候,一件非常重要的事情,就是为TaskSchedulerImpl创建
// 一个SparkDeploySchedulerBackend,这里的backend,指的就是之前创建好的SparkDeploySchedulerBackend,而且这个
// backend是负责创建AppClient,向Master注册Application的
backend.reviveOffers()
}
看看TaskSetManager这个类
/**
* 在TaskSchedulerImpl中,对一个单独的TaskSet的任务进行调度,这个类负责追踪每一个task,如果task失败的话,
* 会负责重试task,直到超过重试的次数限制,并且会通过延迟调度,为这个TaskSet处理本地化调度机制。它的主要接口是resourceOffer,
* 在这个接口中,TaskSet会希望在一个节点上运行一个任务,并且接受任务的状态改变消息,来知道它负责的task的状态改变了
*/
private[spark] class TaskSetManager(
sched: TaskSchedulerImpl,
val taskSet: TaskSet,
val maxTaskFailures: Int,
clock: Clock = new SystemClock())
extends Schedulable with Logging {
看看backend.reviveOffers()方法,CoarseGrainedSchedulerBackend的reviveOffers()方法
override def reviveOffers() {
driverActor ! ReviveOffers
}
CoarseGrainedSchedulerBackend这个类的,DriverActor这个类的ReviveOffers
case ReviveOffers =>
makeOffers()
看makeOffers()方法
// Make fake resource offers on all executors
def makeOffers() {
// 第一步,调用TaskSchedulerImpl的resourceOffers()方法,执行任务分配算法,将各个task分配到executor上去
// 第二步,分配好task到Executor之后,执行自己的的launchTasks()方法,将分配的task发送launchTask消息到对应的Executor上去,由Executor启动并执行task
// 给resourceOffers方法传入的是这个Application所有可用的Executor,并且将其封装成了WorkerOffer,每个WorkerOffer代表了每个Executor可用的cpu资源数量
launchTasks(scheduler.resourceOffers(executorDataMap.map { case (id, executorData) =>
new WorkerOffer(id, executorData.executorHost, executorData.freeCores)
}.toSeq))
}
首先看scheduler.resourceOffers()
def resourceOffers(offers: Seq[WorkerOffer]): Seq[Seq[TaskDescription]] = synchronized {
// Mark each slave as alive and remember its hostname
// Also track if new executor is added
var newExecAvail = false
for (o <- offers) {
executorIdToHost(o.executorId) = o.host
activeExecutorIds += o.executorId
if (!executorsByHost.contains(o.host)) {
executorsByHost(o.host) = new HashSet[String]()
executorAdded(o.executorId, o.host)
newExecAvail = true
}
for (rack <- getRackForHost(o.host)) {
hostsByRack.getOrElseUpdate(rack, new HashSet[String]()) += o.host
}
}
// 首先,将可用的executor进行shuffle,也就是说,进行打散,从而做到,尽可能可以进行负载均衡
// Randomly shuffle offers to avoid always placing tasks on the same set of workers.
val shuffledOffers = Random.shuffle(offers)
// Build a list of tasks to assign to each worker.
// 然后针对WorkerOffer,创建一堆需要用的东西
// 比如tasks,它可以理解为一个二维数组,即ArrayBuffer的元素又是一个ArrayBuffer,并且每个子ArrayBuffer的数量是固定的,也就是这个Executor可用的cpu数量
val tasks = shuffledOffers.map(o => new ArrayBuffer[TaskDescription](o.cores))
val availableCpus = shuffledOffers.map(o => o.cores).toArray
// 这个很重要,从rootPool中取出了排序的TaskSet,之前讲解TaskScheduler初始化的时候,创建完TaskSchedulerImpl、SparkDeploySchedulerBackend之后,执行一个initialize()
// 方法,在这个方法中,其实会创建一个调度池,这里,相当于是说,所有提交的taskSet,首先呢,会放入这个调度池,然后再执行task分配算法的时候,会从这个调度池中,取出排好队的TaskSet
val sortedTaskSets = rootPool.getSortedTaskSetQueue
for (taskSet <- sortedTaskSets) {
logDebug("parentName: %s, name: %s, runningTasks: %s".format(
taskSet.parent.name, taskSet.name, taskSet.runningTasks))
if (newExecAvail) {
taskSet.executorAdded()
}
}
// Take each TaskSet in our scheduling order, and then offer it each node in increasing order
// of locality levels so that it gets a chance to launch local tasks on all of them.
// NOTE: the preferredLocality order: PROCESS_LOCAL, NODE_LOCAL, NO_PREF, RACK_LOCAL, ANY
// 这里,是任务分配算法的核心,双重for循环,遍历所有的taskSet,以及每一种本地化级别
// 本地化级别有
// PROCESS_LOCAL,进程本地化,rdd的partition和task,进入一个Executor内,速度当然快
// NODE_LOCAL,dd的partition和task,不在一个Executor重,不在一个进程,但是在一个worker节点上
// NO_PREF,无,没有所谓的本地化级别
// RACK_LOCAL,机架本地化,至少rdd的partition和task,在一个机架上
// ANY,任意的本地化级别
// 这几种本地化级别 是从小到大排列的
var launchedTask = false
// 对每一个taskSet,从最好的一种本地化级别,开始遍历
for (taskSet <- sortedTaskSets; maxLocality <- taskSet.myLocalityLevels) {
do {
// 对当前taskSet,尝试优先使用最小的本地化级别,将taskset的task,在Executor上进行启动
// 如果启动不了,那么就跳出这个do while循环,进入下一种本地化级别,也就是放大本地化级别
// 以此类推,直到尝试将taskset在某些本地化级别下,在task在Executor上全部启动
launchedTask = resourceOfferSingleTaskSet(
taskSet, maxLocality, shuffledOffers, availableCpus, tasks)
} while (launchedTask)
}
if (tasks.size > 0) {
hasLaunchedTask = true
}
return tasks
}
继续看resourceOfferSingleTaskSet()方法
private def resourceOfferSingleTaskSet(
taskSet: TaskSetManager,
maxLocality: TaskLocality,
shuffledOffers: Seq[WorkerOffer],
availableCpus: Array[Int],
tasks: Seq[ArrayBuffer[TaskDescription]]) : Boolean = {
var launchedTask = false
// 遍历所有Executor
for (i <- 0 until shuffledOffers.size) {
val execId = shuffledOffers(i).executorId
val host = shuffledOffers(i).host
// 如果当前Executor的cpu数量大于每个task要使用的cpu数量,默认是1
if (availableCpus(i) >= CPUS_PER_TASK) {
try {
// 调用taskSetManager的resourceOffer方法,去找到,在这个Executor,用这种本地化级别,taskset的哪些task可以启动
// resourceOffer()方法,就是说,会去判断这个task在这个这个本地化级别,之前的等待时间是多少,如果说,本地化级别的等待时间在一定范围内
// 那么就认为task使用本地化级别可以在executor上启动
for (task <- taskSet.resourceOffer(execId, host, maxLocality)) {
tasks(i) += task
val tid = task.taskId
taskIdToTaskSetId(tid) = taskSet.taskSet.id
taskIdToExecutorId(tid) = execId
executorsByHost(host) += execId
availableCpus(i) -= CPUS_PER_TASK
assert(availableCpus(i) >= 0)
launchedTask = true
}
} catch {
case e: TaskNotSerializableException =>
logError(s"Resource offer failed, task set ${taskSet.name} was not serializable")
// Do not offer resources for this task, but don't throw an error to allow other
// task sets to be submitted.
return launchedTask
}
}
}
return launchedTask
}
接下来看launchTasks()方法
// Launch tasks returned by a set of resource offers
// 根据分配好的情况,去Executor上启动相应的task
def launchTasks(tasks: Seq[Seq[TaskDescription]]) {
for (task <- tasks.flatten) {
// 首先将每个Executor要执行的task信息,统一进行序列化操作
val ser = SparkEnv.get.closureSerializer.newInstance()
val serializedTask = ser.serialize(task)
if (serializedTask.limit >= akkaFrameSize - AkkaUtils.reservedSizeBytes) {
val taskSetId = scheduler.taskIdToTaskSetId(task.taskId)
scheduler.activeTaskSets.get(taskSetId).foreach { taskSet =>
try {
var msg = "Serialized task %s:%d was %d bytes, which exceeds max allowed: " +
"spark.akka.frameSize (%d bytes) - reserved (%d bytes). Consider increasing " +
"spark.akka.frameSize or using broadcast variables for large values."
msg = msg.format(task.taskId, task.index, serializedTask.limit, akkaFrameSize,
AkkaUtils.reservedSizeBytes)
taskSet.abort(msg)
} catch {
case e: Exception => logError("Exception in error callback", e)
}
}
}
else {
// 找到对应的executor
val executorData = executorDataMap(task.executorId)
// 给executor上的资源,减去要使用的cpu资源
executorData.freeCores -= scheduler.CPUS_PER_TASK
// 向executor发送LaunchTask消息,来在executor上启动task
executorData.executorActor ! LaunchTask(new SerializableBuffer(serializedTask))
}
}
}