[
https://issues.apache.org/jira/browse/HBASE-17093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15668983#comment-15668983
]
Enis Soztutar commented on HBASE-17093:
---------------------------------------
It is worth noting that this is for HBASE-14417. The idea seems to be that we
do a 2-PC like operation for saving the bulk load entries to the backup table
to make sure that next incremental backup will save those.
For the patch, I thought the plan was to save the BL paths from the server side
itself. Is there a reason why are these propagated all the way to the client?
Is it because we do not want to block the BL event in case backup table is not
available?
> Enhance SecureBulkLoadClient#secureBulkLoadHFiles() to return the family
> paths of the final hfiles
> --------------------------------------------------------------------------------------------------
>
> Key: HBASE-17093
> URL: https://issues.apache.org/jira/browse/HBASE-17093
> Project: HBase
> Issue Type: Improvement
> Reporter: Ted Yu
> Attachments: 17093.v3.txt, 17093.v4.txt
>
>
> Currently SecureBulkLoadClient#secureBulkLoadHFiles() returns boolean value
> to indicate success / failure.
> Since SecureBulkLoadClient.java is new to master branch, we can change the
> return type to be the family paths of the final hfiles.
> LoadQueueItem would be moved to hbase-client module.
> LoadQueueItem in hbase-server module would delegate to the class in
> hbase-client module.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
