[
https://issues.apache.org/jira/browse/SPARK-6675?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Michael Armbrust updated SPARK-6675:
------------------------------------
Target Version/s: 1.4.0 (was: 1.3.0)
> HiveContext setConf is not stable
> ---------------------------------
>
> Key: SPARK-6675
> URL: https://issues.apache.org/jira/browse/SPARK-6675
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 1.3.0
> Environment: AWS ec2 xlarge2 cluster launched by spark's script
> Reporter: Julien
>
> I find HiveContext.setConf does not work correctly. Here are some code
> snippets showing the problem:
> snippet 1:
> {code}
> import org.apache.spark.sql.hive.HiveContext
> import org.apache.spark.{SparkConf, SparkContext}
> object Main extends App {
> val conf = new SparkConf()
> .setAppName("context-test")
> .setMaster("local[8]")
> val sc = new SparkContext(conf)
> val hc = new HiveContext(sc)
> hc.setConf("spark.sql.shuffle.partitions", "10")
> hc.setConf("hive.metastore.warehouse.dir",
> "/home/spark/hive/warehouse_test")
> hc.getAllConfs filter(_._1.contains("warehouse.dir")) foreach println
> hc.getAllConfs filter(_._1.contains("shuffle.partitions")) foreach println
> }
> {code}
> Results:
> (hive.metastore.warehouse.dir,/home/spark/hive/warehouse_test)
> (spark.sql.shuffle.partitions,10)
> snippet 2:
> {code}
> ...
> hc.setConf("hive.metastore.warehouse.dir",
> "/home/spark/hive/warehouse_test")
> hc.setConf("spark.sql.shuffle.partitions", "10")
> hc.getAllConfs filter(_._1.contains("warehouse.dir")) foreach println
> hc.getAllConfs filter(_._1.contains("shuffle.partitions")) foreach println
> ...
> {code}
> Results:
> (hive.metastore.warehouse.dir,/user/hive/warehouse)
> (spark.sql.shuffle.partitions,10)
> You can see that I just permuted the two setConf call, then that leads to two
> different Hive configuration.
> It seems that HiveContext can not set a new value on
> "hive.metastore.warehouse.dir" key in one or the first "setConf" call.
> You need another "setConf" call before changing
> "hive.metastore.warehouse.dir". For example, set
> "hive.metastore.warehouse.dir" twice and the snippet 1
> snippet 3:
> {code}
> ...
> hc.setConf("hive.metastore.warehouse.dir",
> "/home/spark/hive/warehouse_test")
> hc.setConf("hive.metastore.warehouse.dir",
> "/home/spark/hive/warehouse_test")
> hc.getAllConfs filter(_._1.contains("warehouse.dir")) foreach println
> ...
> {code}
> Results:
> (hive.metastore.warehouse.dir,/home/spark/hive/warehouse_test)
> You can reproduce this if you move to the latest branch-1.3 (1.3.1-snapshot,
> htag = 7d029cb1eb6f1df1bce1a3f5784fb7ce2f981a33)
> I have also tested the released 1.3.0 (htag =
> 4aaf48d46d13129f0f9bdafd771dd80fe568a7dc). It has the same problem.
--
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
