falcon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Venkatesh Seetharam (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FALCON-90) Propagate hive table through Process mapper
Date Mon, 30 Sep 2013 22:58:24 GMT

    [ https://issues.apache.org/jira/browse/FALCON-90?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13782385#comment-13782385
] 

Venkatesh Seetharam commented on FALCON-90:
-------------------------------------------

bq. Should we add a "." between prefix and property name ? Also should we rename this method
to propogateCatalogProperties 
both makes sense. I'll also remove camel case then.
{code}
input.database
input.table
input.pig.filter
{code}

bq. LateRerunConsumer::detectLate assumes falconInputFeeds & falconInPaths are populated.

I'm not sure about late data handling for Hive. It does not make sense in its current implementation.
Whenever data needs to be added to a already committed partition, that particular partition
needs to be removed and re-added which should ideally trigger an oozie coordinator action
and workflow re-executed. Makes sense?

Also, in this particular case, both are populated but does not make sense. Pls look at: org.apache.falcon.converter.OozieProcessMapper#createDefaultCoordinator
What do you suggest to disable this behavior? Send storage-type in the message?

> Propagate hive table through Process mapper
> -------------------------------------------
>
>                 Key: FALCON-90
>                 URL: https://issues.apache.org/jira/browse/FALCON-90
>             Project: Falcon
>          Issue Type: Sub-task
>    Affects Versions: 0.3
>            Reporter: Venkatesh Seetharam
>            Assignee: Venkatesh Seetharam
>         Attachments: FALCON-90.patch
>
>




--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message