You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@storm.apache.org by "Sriharsha Chintalapani (JIRA)" <ji...@apache.org> on 2015/11/10 01:18:10 UTC
[jira] [Resolved] (STORM-1014) Use Hive Streaming API bucket info
to bucket correctly
[ https://issues.apache.org/jira/browse/STORM-1014?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sriharsha Chintalapani resolved STORM-1014.
-------------------------------------------
Resolution: Implemented
fixed on hive side.
> Use Hive Streaming API bucket info to bucket correctly
> ------------------------------------------------------
>
> Key: STORM-1014
> URL: https://issues.apache.org/jira/browse/STORM-1014
> Project: Apache Storm
> Issue Type: Improvement
> Components: storm-hive
> Reporter: Raj Bains
> Assignee: Sriharsha Chintalapani
> Priority: Critical
>
> The Storm bolt get a random bucket and writes data to it. Hive has expectation that rows (tuples for storm) are distributed across buckets using Hive's hash distribution. Writing to a random bucket by Storm leads to Hive optimizations that rely on bucketing to return incorrect results.
> The solution is for Storm Hive Bolt to use Hive bucket distribution information and put the rows/tuples in the correct buckets. This relies on Hive-11672.
> This might require a shuffle within Storm.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)