[
https://issues.apache.org/jira/browse/HBASE-20334?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16502135#comment-16502135
]
Josh Elser commented on HBASE-20334:
------------------------------------
{code:java}
/path/to/component/bin-install{code}
You mean the HBase installation by this? (the unpacked HBase binary tarball)
{code:java}
+ <property>
+ <name>hbase.rootdir</name>
+ <!-- We rely on the defaultFS being set in our hadoop confs -->
+ <value>/hbase</value>
+ </property>{code}
Double-checking, the script is always starting up a minicluster – never trying
to use an existing HDFS installation?
{code:java}
+yarn_server_tests_test_jar="$3"
+mapred_jobclient_test_jar="$4"{code}
Could add some bash "is file" conditional checks on these two.
I'm sure you would've done this on your own already if there was an easy way,
but is there a way we could avoid generated the TSV data and the Java file at
the end of the script?
> add a test that expressly uses both our shaded client and the one from hadoop
> 3
> -------------------------------------------------------------------------------
>
> Key: HBASE-20334
> URL: https://issues.apache.org/jira/browse/HBASE-20334
> Project: HBase
> Issue Type: Sub-task
> Components: hadoop3, shading
> Affects Versions: 2.0.0
> Reporter: Sean Busbey
> Assignee: Sean Busbey
> Priority: Major
> Attachments: HBASE-20334.0.patch
>
>
> Since we're making a shaded client that bleed out of our namespace and into
> Hadoop's, we should ensure that we can show our clients coexisting. Even if
> it's just an IT that successfully talks to both us and HDFS via our
> respective shaded clients, that'd be a big help in keeping us proactive.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
