Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/6361#discussion_r30932075
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/ShuffleMapTask.scala ---
@@ -58,8 +59,18 @@ private[spark] class ShuffleMapTask(
// Deserialize the RDD using the broadcast variable.
val deserializeStartTime = System.currentTimeMillis()
val ser = SparkEnv.get.closureSerializer.newInstance()
+ // Kryo deserialization is not thread-safe w.r.t. underlying buffer
+ // create a copy of the buffer if Kryo is being used
+ val copy = if
(SparkEnv.get.closureSerializer.isInstanceOf[KryoSerializer]) {
--- End diff --
I suppose that there's some code duplication between ShuffleMapTask and
ResultTask, but maybe that's not easily avoidable. We might be able to factor
the deserialization code into a method in `Task` and have it be parametrized by
the type passed to `ser.deserialize`. Maybe you could try doing this to see if
it's possible, since it would be way cleaner?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
