You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flume.apache.org by "Peter Turcsanyi (JIRA)" <ji...@apache.org> on 2018/12/18 08:52:00 UTC

[jira] [Created] (FLUME-3307) Get rid of the reflection calls in AbstractHDFSWriter

Peter Turcsanyi created FLUME-3307:
--------------------------------------

             Summary: Get rid of the reflection calls in AbstractHDFSWriter
                 Key: FLUME-3307
                 URL: https://issues.apache.org/jira/browse/FLUME-3307
             Project: Flume
          Issue Type: Task
            Reporter: Peter Turcsanyi


AbstractHDFSWriter uses some HDFS client calls through reflection in order to retrieve the default / current replication count and also to determine if the client supports hflush() or only sync().
This reflection magic comes from the the early versions of Hadoop and it is quite obsolete now because Hadoop 2+ provides public API for the these calls and supports hflush().
So the reflection calls could be eliminated now.

Furthermore, the whole replication count checking logic seems to be a bit overkill, because it should be handled by HDFS itself and not in the client code.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@flume.apache.org
For additional commands, e-mail: issues-help@flume.apache.org