spark-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "EdisonWang (JIRA)" <>
Subject [jira] [Resolved] (SPARK-26500) Add conf to support ignore hdfs data locality
Date Mon, 31 Dec 2018 08:46:00 GMT


EdisonWang resolved SPARK-26500.
    Resolution: Not A Problem

> Add conf to support ignore hdfs data locality
> ---------------------------------------------
>                 Key: SPARK-26500
>                 URL:
>             Project: Spark
>          Issue Type: Improvement
>          Components: Spark Core
>    Affects Versions: 2.4.0
>            Reporter: EdisonWang
>            Priority: Trivial
> When reading a large hive table/directory with thousands of files, it will cost up
to several minutes or even hours to calculate data locality for each split in driver, while
executors are in idle status.
> This situation is even worth when running in SparkThriftServer mode, because handleJobSubmitted(it
will call getPreferedLocation) is handled in a single thread. One big sql will block all the
following sqls.
> At the same time, most companies's internal networks are all gigabit network cards, so
it is ok to read a data not locality.

This message was sent by Atlassian JIRA

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message