[
https://issues.apache.org/jira/browse/PHOENIX-1980?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14556626#comment-14556626
]
Nick Dimiduk commented on PHOENIX-1980:
---------------------------------------
[~enis]
bq. Previously -z meant that you can only pass the zookeeper quorum (not port
and znode parent from my understanding). If that is not the case, this logic
seems correct. If that is the case, then why don't we still do the same thing
by just setting the conf value and let the QueryUtil.getConnectionUrl() deal
with constructing the JDBC string?
Thsi patch does simply set the conf value and let the
QueryUtil.getConnectionUrl() deal with constructing the JDBC string. I'm using
PhoenixDriver.ConnectionInfo object to parse the parameter provided by the user
and using it's results to populate the conf object. The user-provided string is
prepended with the jdbc bits because PhoenixDriver.ConnectionInfo.create()
assumes it's working with a jdbc connection string. I can change it to handle
either case, then there will be no need to prepend the user input with the
"jdbc:...".
> CsvBulkLoad cannot load hbase-site.xml from classpath
> -----------------------------------------------------
>
> Key: PHOENIX-1980
> URL: https://issues.apache.org/jira/browse/PHOENIX-1980
> Project: Phoenix
> Issue Type: Bug
> Reporter: Nick Dimiduk
> Assignee: Nick Dimiduk
> Fix For: 5.0.0, 4.5.0
>
> Attachments: 1980.patch, PHOENIX-1980.addendum.00.patch
>
>
> When I launch a job using a distributed cluster where hbase-site.xml is
> provided in {{HADOOP_CLASSPATH}} instead of providing --zookeeper, I see
> errors showing "localhost:2181/hbase" is the target connection, where I would
> expect "<hostname>:2181/hbase-unsecure".
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
