[ https://issues.apache.org/jira/browse/HADOOP-16126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16773424#comment-16773424 ]
Tsz Wo Nicholas Sze commented on HADOOP-16126: ---------------------------------------------- > Why the choice of 10ms? It is because 10ms works well in the test. 10ms is a long time in modern computers. So, it is still far away from busy waiting. As you already aware, we have HADOOP-16127 for a better (but more complicated) fix. The patch here is a safe, short term fix for the clusters which do not welcome big changes. > Can you tease this out as a private constant in the IPC file, just so it's > less hidden deep in the code. Will do. > ipc.Client.stop() may sleep too long to wait for all connections > ---------------------------------------------------------------- > > Key: HADOOP-16126 > URL: https://issues.apache.org/jira/browse/HADOOP-16126 > Project: Hadoop Common > Issue Type: Improvement > Components: ipc > Reporter: Tsz Wo Nicholas Sze > Assignee: Tsz Wo Nicholas Sze > Priority: Major > Attachments: c16126_20190219.patch > > > {code} > //Client.java > public void stop() { > ... > // wait until all connections are closed > while (!connections.isEmpty()) { > try { > Thread.sleep(100); > } catch (InterruptedException e) { > } > } > ... > } > {code} > 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. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-issues-h...@hadoop.apache.org