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

Posted to issues@hive.apache.org by "Sankar Hariappan (JIRA)" <ji...@apache.org> on 2018/06/11 12:13:00 UTC

[jira] [Comment Edited] (HIVE-19739) Bootstrap REPL LOAD to use checkpoints to validate and skip the loaded data/metadata.

    [ https://issues.apache.org/jira/browse/HIVE-19739?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16507927#comment-16507927 ] 

Sankar Hariappan edited comment on HIVE-19739 at 6/11/18 12:12 PM:
-------------------------------------------------------------------

Attached 03.patch with fix for ptest failure. Also, supported clean bootstrap if DB properties for ckpt and last repl ID is reset to empty values.


was (Author: sankarh):
Attached 03.patch with fix for ptest failure.

> Bootstrap REPL LOAD to use checkpoints to validate and skip the loaded data/metadata.
> -------------------------------------------------------------------------------------
>
>                 Key: HIVE-19739
>                 URL: https://issues.apache.org/jira/browse/HIVE-19739
>             Project: Hive
>          Issue Type: Sub-task
>          Components: HiveServer2, repl
>    Affects Versions: 3.0.0
>            Reporter: Sankar Hariappan
>            Assignee: Sankar Hariappan
>            Priority: Major
>              Labels: DR, pull-request-available, replication
>             Fix For: 4.0.0
>
>         Attachments: HIVE-19739.01.patch, HIVE-19739.02.patch, HIVE-19739.03.patch
>
>
> Currently. bootstrap REPL LOAD have added checkpoint identifiers in DB/table/partition object properties once the data/metadata related to the object is successfully loaded.
> If the Db exist and is not empty, then currently we are throwing exception. But need to support it for the retry scenario after a failure.
> If there is a retry of bootstrap load using the same dump, then instead of throwing error, we should check if any of the tables/partitions are completely loaded using the checkpoint identifiers. If yes, then skip it or else drop/create them again.
> If the bootstrap load is performed using different dump, then it should throw exception.
> Allow bootstrap on empty Db only if ckpt property is not set. Also, if bootstrap load is completed on the target Db, then shouldn't allow bootstrap retry at all.



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