[
https://issues.apache.org/jira/browse/SPARK-8498?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Reynold Xin resolved SPARK-8498.
--------------------------------
Resolution: Fixed
Target Version/s: (was: 1.4.2)
Resolving this since we are only fixing it for 1.5, not branch-1.4 since it is
an experimental feature in 1.4.
> Fix NullPointerException in error-handling path in UnsafeShuffleWriter
> ----------------------------------------------------------------------
>
> Key: SPARK-8498
> URL: https://issues.apache.org/jira/browse/SPARK-8498
> Project: Spark
> Issue Type: Bug
> Components: Shuffle
> Affects Versions: 1.4.0
> Reporter: Josh Rosen
> Assignee: holdenk
> Fix For: 1.5.0
>
>
> This bug was reported by [~prudenko] on the dev list. When the
> {{tungsten-sort}} shuffle manager was enabled, an executor died with the
> following exception:
> {code}
> 15/06/19 17:53:35 WARN TaskSetManager: Lost task 38.0 in stage 41.0 (TID
> 3176, ip-10-50-225-214.ec2.internal): java.lang.NullPointerException
> at
> org.apache.spark.shuffle.unsafe.UnsafeShuffleWriter.write(UnsafeShuffleWriter.java:151)
> at
> org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:70)
> at
> org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:41)
> at org.apache.spark.scheduler.Task.run(Task.scala:70)
> at
> org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:213)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
> at java.lang.Thread.run(Thread.java:745)
> {code}
> I think that this is actually due to an error-handling issue. In the stack
> trace, the NPE is being thrown from an error-handling branch of a `finally`
> block:
> {code}
> public void write(scala.collection.Iterator<Product2<K, V>> records) throws
> IOException {
> boolean success = false;
> try {
> while (records.hasNext()) {
> insertRecordIntoSorter(records.next());
> }
> closeAndWriteOutput();
> success = true;
> } finally {
> if (!success) {
> sorter.cleanupAfterError(); // <---- this is the line throwing the
> error
> }
> }
> }
> {code}
> I suspect that what's happening is that an exception is being thrown from
> user / upstream code in the initial call to records.next(), but the
> error-handling block is failing because sorter == null since we haven't
> initialized it yet.
> We should fix this bug with a {{sorter != null}} check and should also add a
> regression test to ShuffleSuite to ensure that exceptions thrown by user code
> at this step of the shuffle write path don't get masked by error-handling
> bugs inside of the shuffle code.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
