[SPARK-25318] Add exception handling when wrapping the input stream during the the fetch or stage retry in response to a corrupted block

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.1.3, 2.2.2, 2.3.1, 2.4.0
• Fix Version/s: 2.4.0
• Component/s: Spark Core
• Assignee: Reza Safi
• Reporter: Reza Safi
• Created: 04/Sep/18 03:05
• Updated: 26/Sep/18 16:30
• Resolved: 26/Sep/18 16:30

Description

SPARK-4105 provided a solution to block corruption issue by retrying the fetch or the stage. In the solution there is a step that wraps the input stream with compression and/or encryption. This step is prone to exceptions, but in the current code there is no exception handling for this step and this has caused confusion for the user. In fact we have customers who reported an exception like the following when SPARK-4105 is available to them:

2018-08-28 22:35:54,361 ERROR [Driver] org.apache.spark.deploy.yarn.ApplicationMaster:95 User class threw exception: java.lang.RuntimeException: org.apache.spark.SparkException: Job aborted due to        stage failure: Task 452 in stage 209.0 failed 4 times, most recent failure: Lost task 452.3 in stage y.0 (TID z, xxxxx, executor xx): java.io.IOException: FAILED_TO_UNCOMPRESS(5)
  3976         at org.xerial.snappy.SnappyNative.throw_error(SnappyNative.java:78)
  3977         at org.xerial.snappy.SnappyNative.rawUncompress(Native Method)
  3978         at org.xerial.snappy.Snappy.rawUncompress(Snappy.java:395)
  3979         at org.xerial.snappy.Snappy.uncompress(Snappy.java:431)
  3980         at org.xerial.snappy.SnappyInputStream.readFully(SnappyInputStream.java:127)
  3981         at org.xerial.snappy.SnappyInputStream.readHeader(SnappyInputStream.java:88)
  3982         at org.xerial.snappy.SnappyInputStream.<init>(SnappyInputStream.java:58)
  3983         at org.apache.spark.io.SnappyCompressionCodec.compressedInputStream(CompressionCodec.scala:159)
  3984         at org.apache.spark.storage.BlockManager.wrapForCompression(BlockManager.scala:1219)
  3985         at org.apache.spark.shuffle.BlockStoreShuffleReader$$anonfun$2.apply(BlockStoreShuffleReader.scala:48)
  3986         at org.apache.spark.shuffle.BlockStoreShuffleReader$$anonfun$2.apply(BlockStoreShuffleReader.scala:47)
  3987         at org.apache.spark.storage.ShuffleBlockFetcherIterator.next(ShuffleBlockFetcherIterator.scala:328)
  3988         at org.apache.spark.storage.ShuffleBlockFetcherIterator.next(ShuffleBlockFetcherIterator.scala:55)
  3989         at scala.collection.Iterator$$anon$13.hasNext(Iterator.scala:371)
  3990         a

In this customer's version of spark, line 328 of ShuffleBlockFetcherIterator.scala is the line that the following occurs:

input = streamWrapper(blockId, in)

It would be nice to add exception handling around this line to avoid confusions.

Issue Links

• links to: [Github] Pull Request #22325 (rezasafi)