You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Stephen Kestle (Jira)" <ji...@apache.org> on 2022/06/02 11:53:00 UTC
[jira] [Commented] (SPARK-33121) Spark Streaming 3.1.1 hangs on shutdown
[ https://issues.apache.org/jira/browse/SPARK-33121?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17545437#comment-17545437 ]
Stephen Kestle commented on SPARK-33121:
----------------------------------------
I seem to have encountered this problem without streaming in version 3.2.1. My large session has many complex stages and writes unique data. I've enabled re-runs, and so do an anti-join on already written data in the (jdbc) database.
At this stage, it seems to be applying the anti-join right at the end after all the complex computations and reduces millions of rows to 0. Because I have many executors, I coalesce down to 30 partitions for writing to jdbc.
I'm getting hundreds of {{RejectedExecutionExceptions}} - it seems to me that the coalescing starts, and simultaneously it determines 0 rows and finishes the write and exits, resulting in non-graceful shutdown.
Calling {{sc.stop()}} does nothing, but {{df.cache()}} before coalescing and writing does.
Should this be reported as a separate ticket? I asked on [gitter|https://gitter.im/spark-scala/Lobby?at=6298a15306a77e1e18684826] too, and thought this actually did seem similar enough to comment.
> Spark Streaming 3.1.1 hangs on shutdown
> ---------------------------------------
>
> Key: SPARK-33121
> URL: https://issues.apache.org/jira/browse/SPARK-33121
> Project: Spark
> Issue Type: Bug
> Components: DStreams
> Affects Versions: 3.1.1
> Reporter: Dmitry Tverdokhleb
> Priority: Major
> Labels: Streaming, hang, shutdown
>
> Hi. I am trying to migrate from spark 2.4.5 to 3.1.1 and there is a problem in graceful shutdown.
> Config parameter "spark.streaming.stopGracefullyOnShutdown" is set as "true".
> Here is the code:
> {code:java}
> inputStream.foreachRDD {
> rdd =>
> rdd.foreachPartition {
> Thread.sleep(5000)
> }
> }
> {code}
> I send a SIGTERM signal to stop the spark streaming and after sleeping an exception arises:
> {noformat}
> streaming-agg-tds-data_1 | java.util.concurrent.RejectedExecutionException: Task org.apache.spark.executor.Executor$TaskRunner@7ca7f0b8 rejected from java.util.concurrent.ThreadPoolExecutor@2474219c[Terminated, pool size = 0, active threads = 0, queued tasks = 0, completed tasks = 1]
> streaming-agg-tds-data_1 | at java.util.concurrent.ThreadPoolExecutor$AbortPolicy.rejectedExecution(ThreadPoolExecutor.java:2063)
> streaming-agg-tds-data_1 | at java.util.concurrent.ThreadPoolExecutor.reject(ThreadPoolExecutor.java:830)
> streaming-agg-tds-data_1 | at java.util.concurrent.ThreadPoolExecutor.execute(ThreadPoolExecutor.java:1379)
> streaming-agg-tds-data_1 | at org.apache.spark.executor.Executor.launchTask(Executor.scala:270)
> streaming-agg-tds-data_1 | at org.apache.spark.scheduler.local.LocalEndpoint.$anonfun$reviveOffers$1(LocalSchedulerBackend.scala:93)
> streaming-agg-tds-data_1 | at org.apache.spark.scheduler.local.LocalEndpoint.$anonfun$reviveOffers$1$adapted(LocalSchedulerBackend.scala:91)
> streaming-agg-tds-data_1 | at scala.collection.Iterator.foreach(Iterator.scala:941)
> streaming-agg-tds-data_1 | at scala.collection.Iterator.foreach$(Iterator.scala:941)
> streaming-agg-tds-data_1 | at scala.collection.AbstractIterator.foreach(Iterator.scala:1429)
> streaming-agg-tds-data_1 | at scala.collection.IterableLike.foreach(IterableLike.scala:74)
> streaming-agg-tds-data_1 | at scala.collection.IterableLike.foreach$(IterableLike.scala:73)
> streaming-agg-tds-data_1 | at scala.collection.AbstractIterable.foreach(Iterable.scala:56)
> streaming-agg-tds-data_1 | at org.apache.spark.scheduler.local.LocalEndpoint.reviveOffers(LocalSchedulerBackend.scala:91)
> streaming-agg-tds-data_1 | at org.apache.spark.scheduler.local.LocalEndpoint$$anonfun$receive$1.applyOrElse(LocalSchedulerBackend.scala:68)
> streaming-agg-tds-data_1 | at org.apache.spark.rpc.netty.Inbox.$anonfun$process$1(Inbox.scala:115)
> streaming-agg-tds-data_1 | at org.apache.spark.rpc.netty.Inbox.safelyCall(Inbox.scala:213)
> streaming-agg-tds-data_1 | at org.apache.spark.rpc.netty.Inbox.process(Inbox.scala:100)
> streaming-agg-tds-data_1 | at org.apache.spark.rpc.netty.MessageLoop.org$apache$spark$rpc$netty$MessageLoop$$receiveLoop(MessageLoop.scala:75)
> streaming-agg-tds-data_1 | at org.apache.spark.rpc.netty.MessageLoop$$anon$1.run(MessageLoop.scala:41)
> streaming-agg-tds-data_1 | at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> streaming-agg-tds-data_1 | at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> streaming-agg-tds-data_1 | at java.lang.Thread.run(Thread.java:748)
> streaming-agg-tds-data_1 | 2021-04-22 13:33:41 WARN JobGenerator - Timed out while stopping the job generator (timeout = 10000)
> streaming-agg-tds-data_1 | 2021-04-22 13:33:41 INFO JobGenerator - Waited for jobs to be processed and checkpoints to be written
> streaming-agg-tds-data_1 | 2021-04-22 13:33:41 INFO JobGenerator - Stopped JobGenerator{noformat}
> After this exception and "JobGenerator - Stopped JobGenerator" log, streaming freezes, and halts by timeout (Config parameter "hadoop.service.shutdown.timeout").
> Besides, there is no problem with the graceful shutdown in spark 2.4.5.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org