flume-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "zhenzhao wang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLUME-3268) Introducing micro batch processing to HDFSEventSink
Date Wed, 15 Aug 2018 00:51:00 GMT

    [ https://issues.apache.org/jira/browse/FLUME-3268?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16580590#comment-16580590

zhenzhao wang commented on FLUME-3268:

[~fszabo] hdfs.batchSize is used for flush frequency. It will call append rpc for each event
. What we are going to do is batching events for HDFS append. There're some problem with current
pull request while I try to refactor the code, will let you known it's good for reviewing.

> Introducing micro batch processing to HDFSEventSink
> ---------------------------------------------------
>                 Key: FLUME-3268
>                 URL: https://issues.apache.org/jira/browse/FLUME-3268
>             Project: Flume
>          Issue Type: New Feature
>            Reporter: zhenzhao wang
>            Priority: Major
>         Attachments: FLUME-3268-0.patch
> In our test with HDFSEvent sink, we found that we could increase the draining speed of
HDFSSink up to 4x by introducing micro batch processing. With the micro batch processing feature,
we will batch the events written to HDFS instead of one by one.

This message was sent by Atlassian JIRA

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

View raw message