Github user Ru-Xiang commented on a diff in the pull request:
https://github.com/apache/spark/pull/16033#discussion_r89775912
--- Diff:
core/src/main/scala/org/apache/spark/partial/ApproximateActionListener.scala ---
@@ -34,11 +34,13 @@ private[spark] class ApproximateActionListener[T, U, R](
rdd: RDD[T],
func: (TaskContext, Iterator[T]) => U,
evaluator: ApproximateEvaluator[U, R],
- timeout: Long)
+ timeout: Long = 1000*60*60*24*30*12,
--- End diff --
This submission stems from our needs that drop out the tasks which are
delayed by the hardware, scheduler or the networks but not the skew data. In a
production environment, different machines may be congested at different
times. While in some machine learning algorithms like SGD, we do not need to
wait all the tasks completion in each iteration for we can guarantee the
convergence as long as the slow nodes are not always slow nodes.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
