You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@phoenix.apache.org by "Lars Hofhansl (JIRA)" <ji...@apache.org> on 2019/06/28 20:57:00 UTC

[jira] [Commented] (PHOENIX-5376) Bulk loading data tables with indexes leads to index rows with incorrect timestamps

    [ https://issues.apache.org/jira/browse/PHOENIX-5376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16875220#comment-16875220 ] 

Lars Hofhansl commented on PHOENIX-5376:
----------------------------------------

Do we have to disable this functionality or is there an "easy" fix to Phoenix bulkloading (i.e. the way in which the different HFiles are built)?

> Bulk loading data tables with indexes leads to index rows with incorrect timestamps
> -----------------------------------------------------------------------------------
>
>                 Key: PHOENIX-5376
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-5376
>             Project: Phoenix
>          Issue Type: Bug
>    Affects Versions: 5.0.0, 4.14.2
>            Reporter: Kadir OZDEMIR
>            Priority: Major
>
> Bulk loading data on a table is done directly on HFiles by bypassing Phoenix. A table and its indexes are loaded independently. This results in timestamp mismatch between data table rows and their corresponding index rows. It is known that this mismatch causes consistency issues. Consistent global indexes introduced by PHOENIX-5156 and PHOENIX-5211 require a data table row and the corresponding index rows should have the same timestamp all the time. Thus, it is required that data and index table rows should get the same timestamp during bulk loading too.



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