[jira] [Commented] (ACCUMULO-1973) Log better message when Hadoop configuration files are not found

["ASF subversion and git services (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ACCUMULO-1973?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13861934#comment-13861934 ] 

ASF subversion and git services commented on ACCUMULO-1973:
-----------------------------------------------------------

Commit 79ff4f1bbee3d9603b1b5007e7041ef6f1d1a8de in branch refs/heads/1.4.5-SNAPSHOT from [~elserj]
[ https://git-wip-us.apache.org/repos/asf?p=accumulo.git;h=79ff4f1 ]

ACCUMULO-1973 Add a warning if we don't find non-default values for namenode configuration

This is one of biggest confusions that new users seem to run into when
trying to initialize Accumulo (as it's a problem with us not finding
Hadoop configurations). Hadoop will work properly, as its configuration
is likely correct, but Accumulo can't find core-site.xml, tries to write
to the root of the local fs, and fails. Try to be a bit more explicit
about what happened when we initialized and the Configuration from the FS.


> Log better message when Hadoop configuration files are not found
> ----------------------------------------------------------------
>
>                 Key: ACCUMULO-1973
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-1973
>             Project: Accumulo
>          Issue Type: Improvement
>    Affects Versions: 1.5.0
>            Reporter: Josh Elser
>            Assignee: Josh Elser
>             Fix For: 1.4.5, 1.5.1, 1.6.0
>
>
> I keep running into this -- with the introduction of Hadoop-2 into the configuration, if you don't properly set HADOOP_CONF_DIR, you'll get the "default" instance.dfs.uri (I believe) which will (almost) always not work out for you.
> You get a strange stack trace around ChecksumFileSystem or FilterFileSystem, but the underlying problem is that your configuration is likely pointing at the wrong configuration directory for Hadoop ($HADOOP_PREFIX/conf instead of $HADOOP_PREFIX/etc/hadoop, or vice versa).
> We should be able to do something that makes this more intuitive to new-comers with printing error messages.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)