Github user vanzin commented on a diff in the pull request: https://github.com/apache/spark/pull/8007#discussion_r37808513 --- Diff: core/src/main/scala/org/apache/spark/scheduler/cluster/YarnSchedulerBackend.scala --- @@ -91,6 +92,52 @@ private[spark] abstract class YarnSchedulerBackend( } /** + * Override the DriverEndpoint to add extra logic for the case when an executor is disconnected. + * We should check the cluster manager and find if the loss of the executor was caused by YARN + * force killing it due to preemption. + */ + private class YarnDriverEndpoint(rpcEnv: RpcEnv, sparkProperties: ArrayBuffer[(String, String)]) + extends DriverEndpoint(rpcEnv, sparkProperties) { + + private val handleDisconnectedExecutorThreadPool = + ThreadUtils.newDaemonCachedThreadPool("yarn-driver-handle-lost-executor-thread-pool") + implicit val askSchedulerExecutor = ExecutionContext.fromExecutor(handleDisconnectedExecutorThreadPool) + + /** + * When onDisconnected is received at the driver endpoint, the superclass DriverEndpoint + * handles it by assuming the Executor was lost for a bad reason and removes the executor + * immediately. + * + * In YARN's case however it is crucial to talk to the application master and ask why the + * executor had exited. In particular, the executor may have exited due to the executor + * having been preempted. If the executor "exited normally" according to the application + * master then we pass that information down to the TaskSetManager to inform the + * TaskSetManager that tasks on that lost executor should not count towards a job failure. + */ + override def onDisconnected(rpcAddress: RpcAddress): Unit = { + addressToExecutorId.get(rpcAddress).foreach({ executorId => + val future = yarnSchedulerEndpoint.ask[ExecutorLossReason](GetExecutorLossReason(executorId), askTimeout) + future onSuccess { + case reason: ExecutorLossReason => + driverEndpoint.askWithRetry[Boolean](RemoveExecutor(executorId, reason)) --- End diff -- (BTW, my comment should have been on L119, where you send `GetExecutorLossReason`.) Following the code in `ApplicationMaster`, it doesn't look like it would be possible for that situation to happen. The AM endpoint is started (and registers with the driver endpoint) before the AM starts allocating executors for the application. Maybe adding an assert (that `amEndpoint` is set) would suffice. Also, just want to point out that, if the race really exists, going through the RPC layer wouldn't fix it, just make it less likely. Ignoring the initialization order in `ApplicationMaster` for a second, the message could still reach the other endpoint before the AM has registered.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org