Mridul Muralidharan created SPARK-36772:
-------------------------------------------
Summary: FinalizeShuffleMerge fails with an exception due to
attempt id not matching
Key: SPARK-36772
URL: https://issues.apache.org/jira/browse/SPARK-36772
Project: Spark
Issue Type: Bug
Components: Shuffle
Affects Versions: 3.2.0
Reporter: Mridul Muralidharan
As part of driver request to external shuffle services (ESS) to finalize the
merge, it also passes its [application attempt
id|https://github.com/apache/spark/blob/3f09093a21306b0fbcb132d4c9f285e56ac6b43c/common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/ExternalBlockStoreClient.java#L180]
so that ESS can validate the request is from the correct attempt.
This attempt id is fetched from the TransportConf passed in when creating the
[ExternalBlockStoreClient|https://github.com/apache/spark/blob/67421d80b8935d91b86e8cd3becb211fa2abd54f/core/src/main/scala/org/apache/spark/SparkEnv.scala#L352]
- and the transport conf leverages a [cloned
copy|https://github.com/apache/spark/blob/0494dc90af48ce7da0625485a4dc6917a244d580/core/src/main/scala/org/apache/spark/network/netty/SparkTransportConf.scala#L47]
of the SparkConf passed to it.
Application attempt id is set as part of SparkContext
[initialization|https://github.com/apache/spark/blob/67421d80b8935d91b86e8cd3becb211fa2abd54f/core/src/main/scala/org/apache/spark/SparkContext.scala#L586].
But this happens after driver SparkEnv has [already been
created|https://github.com/apache/spark/blob/67421d80b8935d91b86e8cd3becb211fa2abd54f/core/src/main/scala/org/apache/spark/SparkContext.scala#L460].
Hence the attempt id that ExternalBlockStoreClient uses will always end up
being -1 : which will not match the attempt id at ESS (which is based on
spark.app.attempt.id) : resulting in merge finalization to always fail ("
java.lang.IllegalArgumentException: The attempt id -1 in this
FinalizeShuffleMerge message does not match with the current attempt id 1
stored in shuffle service for application ...")
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
