[
https://issues.apache.org/jira/browse/PHOENIX-1653?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14349380#comment-14349380
]
Gabriel Reid commented on PHOENIX-1653:
---------------------------------------
Thanks for the patch [~gjacoby]!
A few remarks on this:
* I think it would probably be better if
ConnectionUtil#getConnection(org.apache.hadoop.conf.Configuration) were
completely removed, and there were either specific methods for getting the
input and output cluster (or just stick with the methods that take a ZK quorum
as parameter). The problem is that ConnectionUtil.getConnection is used in some
spots without correctly being scoped to the input or output cluster. For
example, in PhoenixConfigurationUtil#getSelectColumnMetadataList. By removing
the possibility to just retrieve a Connection via a configuration without
scoping it to the input or output will prevent bugs like this.
* It would be good to get some javadoc on
{{PhoenixConfigurationUtil.getMapReduceInputCluster}} and
{{PhoenixConfigurationUtil.getMapReduceOutputCluster}} to make it clear that
the parameter and return value of these methods is a ZK quorum string. The
naming of these methods could also maybe be reconsidered -- the "MapReduce"
portion of the name seems superfluous.
* There are some missing @param entries in the new methods in
PhoenixMapReduceUtil -- these are public-facing API methods, so it's especially
important here that the javadoc is complete
* PhoenixConfigurationUtil seems to have some tabs for indentation, which is
against the general code conventions in Phoenix (indentation should always be 4
spaces)
> Allow option to pass peer zookeeper address to load data into a target
> cluster in Map Reduce api
> ------------------------------------------------------------------------------------------------
>
> Key: PHOENIX-1653
> URL: https://issues.apache.org/jira/browse/PHOENIX-1653
> Project: Phoenix
> Issue Type: Improvement
> Affects Versions: 5.0.0
> Reporter: maghamravikiran
> Labels: newbie, patch
> Attachments: PHOENIX-1653.patch
>
>
> Provide an option to pass the peer zookeeper address within a MapReduce job
> where PhoenixInputFormat reads from one HBase cluster, and
> PhoenixOutputFormat writes to a different cluster.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
