szetszwo edited a comment on pull request #2090: URL: https://github.com/apache/hadoop/pull/2090#issuecomment-650316791
As stated in the JIRA comment (https://issues.apache.org/jira/browse/HADOOP-17048?focusedCommentId=17139883&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17139883), there are two problems for reverting the JIRAs 1) For HADOOP-16126, we do observe a 10x improvement for some Hive jobs as stated in the JIRA – > In the code above, the sleep time is 100ms. We found that simply changing the sleep time to 10ms could improve a Hive job running time by 10x. I guess we may get a better performance for one kind of job but degrade the performance for some other jobs. 2) For HADOOP-16127, the bug is > getConnection(..), running can be initially true but becomes false before putIfAbsent. Reverting it (i.e. removing the lock) has a better performance since it just ignores the synchronization bug. Why not fix the synchronization bug in a better way? Making the value configurable could address Problem 1. How about Problem 2? It is a more serious bug on correctness. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org
