[
https://issues.apache.org/jira/browse/HIVE-20291?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16567304#comment-16567304
]
Jaume M commented on HIVE-20291:
--------------------------------
This intent of using this is in a similar manner to the added test. The
HiveStreamingConnection would receive a writeId instead of figuring it out by
itself. Several HiveStreamingConnection may receive the same writeId and this
seems to result in an error for all except the first HiveStreamingConnection,
therefore the statementId was added.
We can probably remove the StatementIdChooser part and directly pass a the
statementId to the Writer. What do you think of this [~ekoifman]?
bq. If statement id is ever reused this will lead to a data loss.
Wouldn't this lead to an exception in the {{OrcRecordUpdater}} saying that the
file already exists?
> Allow HiveStreamingConnection to receive a WriteId
> --------------------------------------------------
>
> Key: HIVE-20291
> URL: https://issues.apache.org/jira/browse/HIVE-20291
> Project: Hive
> Issue Type: Improvement
> Reporter: Jaume M
> Assignee: Jaume M
> Priority: Major
> Labels: pull-request-available
> Attachments: HIVE-20291.1.patch
>
>
> If the writeId is received externally it won't need to open connections to
> the metastore. It won't be able to the commit in this case as well so it must
> be done by the entity passing the writeId.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
