[
https://issues.apache.org/jira/browse/SPARK-7837?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14563977#comment-14563977
]
Yin Huai commented on SPARK-7837:
---------------------------------
We have made the parquet reader side robust to files left in _temporary. So,
this problem should have a much smaller impact.
I am re-targeting it to 1.5. Will keep an eye on it and investigate the root
cause.
> NPE when save as parquet in speculative tasks
> ---------------------------------------------
>
> Key: SPARK-7837
> URL: https://issues.apache.org/jira/browse/SPARK-7837
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 1.4.0
> Reporter: Yin Huai
> Priority: Critical
>
> The query is like {{df.orderBy(...).saveAsTable(...)}}.
> When there is no partitioning columns and there is a skewed key, I found the
> following exception in speculative tasks. After these failures, seems we
> could not call {{SparkHadoopMapRedUtil.commitTask}} correctly.
> {code}
> java.lang.NullPointerException
> at
> parquet.hadoop.InternalParquetRecordWriter.flushRowGroupToStore(InternalParquetRecordWriter.java:146)
> at
> parquet.hadoop.InternalParquetRecordWriter.close(InternalParquetRecordWriter.java:112)
> at parquet.hadoop.ParquetRecordWriter.close(ParquetRecordWriter.java:73)
> at
> org.apache.spark.sql.parquet.ParquetOutputWriter.close(newParquet.scala:115)
> at
> org.apache.spark.sql.sources.DefaultWriterContainer.abortTask(commands.scala:385)
> at
> org.apache.spark.sql.sources.InsertIntoHadoopFsRelation.org$apache$spark$sql$sources$InsertIntoHadoopFsRelation$$writeRows$1(commands.scala:150)
> at
> org.apache.spark.sql.sources.InsertIntoHadoopFsRelation$$anonfun$insert$1.apply(commands.scala:122)
> at
> org.apache.spark.sql.sources.InsertIntoHadoopFsRelation$$anonfun$insert$1.apply(commands.scala:122)
> at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:63)
> at org.apache.spark.scheduler.Task.run(Task.scala:70)
> at org.apache.spark.executor.Executor$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}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
