Yeah, the problem was work around by adding config spark.sql.codegen.wholeStage=false.
Thanks and Regards, Tony 发件人: kathleen li <kathleenli...@gmail.com> 发送时间: 2018年10月8日 12:15 收件人: 阎志涛 <tony....@tendcloud.com> 抄送: user@spark.apache.org 主题: Re: 答复: Executor hang https://jaceklaskowski.gitbooks.io/mastering-spark-sql/spark-sql-whole-stage-codegen.html#spark.sql.codegen.wholeStage Sent from my iPhone On Oct 7, 2018, at 3:21 PM, 阎志涛 <tony....@tendcloud.com<mailto:tony....@tendcloud.com>> wrote: It is not data skewed issue. The executor running more than 8 hours and I have to kill the executor. I doubt the problem is caused by Spark codegen but it sound spark.sql.codegen.wholeStage=false can not work. I met another problem with codegen which cause my spark job failed with following stack trace: 18/10/08 01:58:36 WARN TaskSetManager: Lost task 1.0 in stage 7.0 (TID 1912, data1.aidata360.com<http://data1.aidata360.com>, executor 4): java.util.NoSuchElementException at java.util.LinkedList.removeFirst(LinkedList.java:270) at org.apache.spark.unsafe.map.BytesToBytesMap$MapIterator.hasNext(BytesToBytesMap.java:310) at org.apache.spark.sql.execution.UnsafeFixedWidthAggregationMap$1.next(UnsafeFixedWidthAggregationMap.java:177) at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIterator.processNext(Unknown Source) at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43) at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$8$$anon$1.hasNext(WholeStageCodegenExec.scala:377) at org.apache.spark.sql.Dataset$$anonfun$56$$anon$1.hasNext(Dataset.scala:2712) at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at scala.collection.Iterator$$anon$13.hasNext(Iterator.scala:461) at org.apache.spark.util.collection.ExternalSorter.insertAll(ExternalSorter.scala:191) at org.apache.spark.shuffle.sort.SortShuffleWriter.write(SortShuffleWriter.scala:63) at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96) at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53) at org.apache.spark.scheduler.Task.run(Task.scala:99) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:322) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) at java.lang.Thread.run(Thread.java:748) How can I disable wholestage code gen? Thanks and Regards, Tony 发件人: kathleen li <kathleenli...@gmail.com<mailto:kathleenli...@gmail.com>> 发送时间: 2018年10月8日 1:04 收件人: 阎志涛 <tony....@tendcloud.com<mailto:tony....@tendcloud.com>> 抄送: user@spark.apache.org<mailto:user@spark.apache.org> 主题: Re: Executor hang It seemed you had data skewed issue since shuffle read size for executor 4 is almost 2 times than other executors and GC time 11s almost 15 to 20 times than others. Kathleen Sent from my iPhone On Oct 7, 2018, at 5:24 AM, 阎志涛 <tony....@tendcloud.com<mailto:tony....@tendcloud.com>> wrote: Hi, All, I am running Spark 2.1 on Hadoop 2.7.2 with yarn. While executing spark tasks, some executor keep running forever without success. From the following screenshot: <image002.jpg> We can see that executor 4 keep running for 26 minutes and the shuffle read size/records keep unchanged for 26mins too. Threaddump for the thread is as following: <image004.jpg> <image009.jpg> The linux version is: Linux version 4.14.62-70.117.amzn2.x86_64 (mockbuild@ip-10-0-1-79) and jdk version is Oracle JDK 1.8.0_181. With jstack on the machine, I can see following thread dump: "Executor task launch worker for task 3806" #54 daemon prio=5 os_prio=0 tid=0x0000000001230800 nid=0x1fc runnable [0x00007fba0e600000] java.lang.Thread.State: RUNNABLE at java.lang.StringCoding.encode(StringCoding.java:364) at java.lang.String.getBytes(String.java:941) at org.apache.spark.unsafe.types.UTF8String.fromString(UTF8String.java:109) at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIterator.processNext(Unknown Source) at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43) at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$8$$anon$1.hasNext(WholeStageCodegenExec.scala:377) at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.execute(FileFormatWriter.scala:243) at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:190) at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:188) at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1341) at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:193) at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$3.apply(FileFormatWriter.scala:129) at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$3.apply(FileFormatWriter.scala:128) at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87) at org.apache.spark.scheduler.Task.run(Task.scala:99) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:322) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) at java.lang.Thread.run(Thread.java:748) I wonder why this happened? Is it related to my environment of a bug of Spark? Thanks and Regards, Tony 阎志涛 研发副总裁 M + 86－139 1181 5695 Wechat zhitao_yan 北京腾云天下科技有限公司 北京市东直门外大街39号院2号楼608室，100027 TalkingData.com<http://TalkingData.com>
