[jira] [Commented] (SQOOP-1341) Sqoop Export Upsert for MySQL lacks batch support

["Gwen Shapira (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SQOOP-1341?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14043860#comment-14043860 ] 

Gwen Shapira commented on SQOOP-1341:
-------------------------------------

Thanks for contributing this correction! Row by row updates are indeed super slow. 

Just one question:
Your comment says: "This assumes that the update key column is the last column serialized in by the underlying record. "
Can you show where does the code depends on this assumption? It looked very generic to me.



> Sqoop Export Upsert for MySQL lacks batch support
> -------------------------------------------------
>
>                 Key: SQOOP-1341
>                 URL: https://issues.apache.org/jira/browse/SQOOP-1341
>             Project: Sqoop
>          Issue Type: Improvement
>          Components: connectors/mysql
>    Affects Versions: 1.4.4
>            Reporter: Andy Skelton
>              Labels: easyfix, newbie, patch
>             Fix For: 1.4.5
>
>         Attachments: SQOOP-1341.patch
>
>   Original Estimate: 1h
>  Remaining Estimate: 1h
>
> MySQL export upserts are limited to one row per statement. This wastes bandwidth and makes the exports unusably slow.
> I wrote a patch to support multiple rows per statement at (https://github.com/cloudera/sqoop/pull/22/files). Now the records.per.statement parameter actually works, and I get fast exports (60k rows/sec) with 1000 records per statement, 10 statements per transactions, and 5 mappers.



--
This message was sent by Atlassian JIRA
(v6.2#6252)