[ https://issues.apache.org/jira/browse/HIVE-13759?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15308991#comment-15308991 ]
Siddharth Seth commented on HIVE-13759: --------------------------------------- LlapRowInputFormat.close() - who invokes this ? It's not part of the InputFormat contract. Closeable is probably the right interface to use. However, I've found this to be a little painful while debugging - since IDEs are not able to find targeted usage. I tend to add a stop() / close() method without implementing Closeable to get around this. Your call on whether to leave 'implements Closeable' or remove it. {code} + if (caughtException != null) { + throw new IOException("Exception during close: " + caughtException.getMessage(), caughtException); + } {code} Is this wrapping required, especially if caughtException is an IOException itself. LlapBaseInputFormat - the connection (con) is created each time getSplits is invoked. Should the close happen in a finally block in getSplits itself. HS2 does not handle dangling connections very well afaik. > LlapTaskUmbilicalExternalClient should be closed by the record reader > --------------------------------------------------------------------- > > Key: HIVE-13759 > URL: https://issues.apache.org/jira/browse/HIVE-13759 > Project: Hive > Issue Type: Sub-task > Components: llap > Reporter: Jason Dere > Assignee: Jason Dere > Attachments: HIVE-13759.1.patch > > > The umbilical external client (and the server socket it creates) doesn't look > like it's getting closed. -- This message was sent by Atlassian JIRA (v6.3.4#6332)