[
https://issues.apache.org/jira/browse/SQOOP-1751?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14221185#comment-14221185
]
Ryan Blue commented on SQOOP-1751:
----------------------------------
What effect does this have on the user? Are they all just config options, or
does an administrator need to set up the dataset URI if it is part of the link
data?
It seems like the link data should be used for anything that can be used across
jobs or needs to be hidden. While the dataset URI is like a JDBC URI, it is
also much more like a table. In fact, that's exactly what the Hive URI pattern
encodes: {{dataset:hive:<db>/<table>}}. That doesn't need to be hidden and
could be different between jobs. The link information *can* be encoded in the
URI, but it doesn't need to be. For example, we can add the metastore host and
port: {{dataset:hive://ms-host:9083/db/table}}, but we prefer if that comes
from the environment, which looks like a perfect use of the link-level data. I
think we should plan on this mode of operation in the future and keep link
information out of Kite URIs when we define new URIs in Kite (the current ones
don't have sensitive link info).
In summary, from [~stanleyxu2005]'s examples, I currently think that the Kite
URI is a job option.
> Rearrange LinkConfig and ToJobConfig
> ------------------------------------
>
> Key: SQOOP-1751
> URL: https://issues.apache.org/jira/browse/SQOOP-1751
> Project: Sqoop
> Issue Type: Sub-task
> Components: connectors
> Reporter: Qian Xu
> Assignee: Qian Xu
> Priority: Minor
> Fix For: 1.99.5
>
>
> Regarding Abe's review comments, here are thoughts to rearrange configuration
> items.
> 1. File Output Format should be specified in ToJobConfig.
> 2. Corresponding validation should happen on time.
> 3. Split protocol, host and port part to link and leave destination to
> toJobConfig
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
