[SPARK-34645] [K8S] Driver pod stuck in Running state after job completes

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.0.2
• Component/s: Kubernetes

Environment

Kubernetes:

Client Version: version.Info{Major:"1", Minor:"19", GitVersion:"v1.19.2", GitCommit:"f5743093fd1c663cb0cbc89748f730662345d44d", GitTreeState:"clean", BuildDate:"2020-09-16T13:41:02Z", GoVersion:"go1.15", Compiler:"gc", Platform:"linux/amd64"}
Server Version: version.Info{Major:"1", Minor:"13", GitVersion:"v1.13.5", GitCommit:"2166946f41b36dea2c4626f90a77706f426cdea2", GitTreeState:"clean", BuildDate:"2019-03-25T15:19:22Z", GoVersion:"go1.11.5", Compiler:"gc", Platform:"linux/amd64"}
 

• Assignee: Unassigned
• Reporter: Andy Grove
• Created: 05/Mar/21 20:43
• Updated: 07/Sep/23 10:50

Description

I am running automated benchmarks in k8s, using spark-submit in cluster mode, so the driver runs in a pod.

When running with Spark 3.0.1 and 3.1.1 everything works as expected and I see the Spark context being shut down after the job completes.

However, when running with Spark 3.0.2 I do not see the context get shut down and the driver pod is stuck in the Running state indefinitely.

This is the output I see after job completion with 3.0.1 and 3.1.1 and this output does not appear with 3.0.2. With 3.0.2 there is no output at all after the job completes.

2021-03-05 20:09:24,576 INFO spark.SparkContext: Invoking stop() from shutdown hook
2021-03-05 20:09:24,592 INFO server.AbstractConnector: Stopped Spark@784499d0{HTTP/1.1, (http/1.1)}{0.0.0.0:4040}
2021-03-05 20:09:24,594 INFO ui.SparkUI: Stopped Spark web UI at http://benchmark-runner-3e8a38780400e0d1-driver-svc.default.svc:4040
2021-03-05 20:09:24,599 INFO k8s.KubernetesClusterSchedulerBackend: Shutting down all executors
2021-03-05 20:09:24,600 INFO k8s.KubernetesClusterSchedulerBackend$KubernetesDriverEndpoint: Asking each executor to shut down
2021-03-05 20:09:24,609 WARN k8s.ExecutorPodsWatchSnapshotSource: Kubernetes client has been closed (this is expected if the application is shutting down.)
2021-03-05 20:09:24,719 INFO spark.MapOutputTrackerMasterEndpoint: MapOutputTrackerMasterEndpoint stopped!
2021-03-05 20:09:24,736 INFO memory.MemoryStore: MemoryStore cleared
2021-03-05 20:09:24,738 INFO storage.BlockManager: BlockManager stopped
2021-03-05 20:09:24,744 INFO storage.BlockManagerMaster: BlockManagerMaster stopped
2021-03-05 20:09:24,752 INFO scheduler.OutputCommitCoordinator$OutputCommitCoordinatorEndpoint: OutputCommitCoordinator stopped!
2021-03-05 20:09:24,768 INFO spark.SparkContext: Successfully stopped SparkContext
2021-03-05 20:09:24,768 INFO util.ShutdownHookManager: Shutdown hook called
2021-03-05 20:09:24,769 INFO util.ShutdownHookManager: Deleting directory /var/data/spark-67fa44df-e86c-463a-a149-25d95817ff8e/spark-a5476c14-c103-4108-b733-961400485d8a
2021-03-05 20:09:24,772 INFO util.ShutdownHookManager: Deleting directory /tmp/spark-9d6261f5-4394-472b-9c9a-e22bde877814
2021-03-05 20:09:24,778 INFO impl.MetricsSystemImpl: Stopping s3a-file-system metrics system...
2021-03-05 20:09:24,779 INFO impl.MetricsSystemImpl: s3a-file-system metrics system stopped.
2021-03-05 20:09:24,779 INFO impl.MetricsSystemImpl: s3a-file-system metrics system shutdown complete.
 

Issue Links

• relates to: SPARK-35304 [k8s] Though finishing a job, the driver pod is running infinitely