[
https://issues.apache.org/jira/browse/SPARK-5917?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14328769#comment-14328769
]
Sean Owen commented on SPARK-5917:
----------------------------------
Isn't this just another symptom of the known issues with S3 and inconsistent
reads?
https://issues.apache.org/jira/browse/SPARK-2579
Did you try the items in that JIRA?
> Distinct is broken
> ------------------
>
> Key: SPARK-5917
> URL: https://issues.apache.org/jira/browse/SPARK-5917
> Project: Spark
> Issue Type: Bug
> Components: MLlib
> Affects Versions: 1.1.1
> Environment: Spark 1.1.1 running on YARN 2.4 via Amazon EMR.
> Reporter: Derrick Burns
> Priority: Critical
>
> I hate to file bugs that are hard to reproduce (by other people), but after
> spending a full week trying to debug my code, I constructed a scenario where
> the following assertion FAILS.
> val x : RDD[T] = ....
> val y = x.distinct()
> assert( y.count() <= x.count() )
> I am at a complete loss as to how this can occur under ANY definition of
> equality/order unless the RDD underlying x is mutable. Since none of my RDD
> transforms mutate any existing RDD data and I am reading from immutable
> sources (data on S3), I conclude that there must be a bug in Spark or I am
> mutating my data unknowingly.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
