[ https://issues.apache.org/jira/browse/HIVE-12615?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15052743#comment-15052743 ]
Xuefu Zhang commented on HIVE-12615: ------------------------------------ [~nemon], Thanks for working on this. I noticed: 1. A lot of test output needs to be updated. 2. I'm not sure showing -1 as the number of reducers makes sense. SetSparkReducerParallelism has logic of getting parallelism w/o knowing the execturor/memory. Given these additional changes required, do you think the problem is worth a fix? In another word, doesn't it really bothers your use case? > Do not start spark session when only explain > --------------------------------------------- > > Key: HIVE-12615 > URL: https://issues.apache.org/jira/browse/HIVE-12615 > Project: Hive > Issue Type: Improvement > Components: Spark > Affects Versions: 1.3.0, 2.1.0 > Reporter: Nemon Lou > Assignee: Nemon Lou > Priority: Minor > Attachments: HIVE-12615.patch > > > When using beeline -e "set hive.execution.engine=spark;explain select > count(*) from sometable",it's very slow due to starting of spark session on > yarn. -- This message was sent by Atlassian JIRA (v6.3.4#6332)