You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Joydeep Sen Sarma (JIRA)" <ji...@apache.org> on 2008/09/08 08:00:45 UTC
[jira] Issue Comment Edited: (HADOOP-4090) The configuration file
lists two paths to hadoop directories (bin and conf). Startup should check
that these are valid directories and give appropriate messages.
[ https://issues.apache.org/jira/browse/HADOOP-4090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12629075#action_12629075 ]
jsensarma edited comment on HADOOP-4090 at 9/7/08 10:59 PM:
--------------------------------------------------------------------
here's the background: lot of regular hadoop users have a 'hadoop' in their path. they have already setup their environment for hadoop. hive is a utility on top of hadoop. as such using the 'hadoop' found in the user's path is meant to help the user. otherwise they have to setup another env variable. that's a negative for user experience.
secondly - HADOOP_HOME is not user defined today. From hadoop-config.sh:
# the root of the Hadoop installation
export HADOOP_HOME=`dirname "$this"`/..
so user's are not setting this today (just in case there is a misconception)
I would agree with renaming 'HADOOP' in hive shell scripts with 'HADOOP_HOME' - that would certainly seem more consistent. i feel that we should default to hadoop found in path - since that is the common case. we should not force users to set more things than are necessary to make hadoop run. would also agree to have HADOOP_HOME override hadoop in path (for advanced users).
was (Author: jsensarma):
hmmm - i don't quite agree with the thesis here.
here's the background: lot of regular hadoop users have a 'hadoop' in their path. they have already setup their environment for hadoop. hive is a utility on top of hadoop. as such using the 'hadoop' found in the user's path is meant to help the user. otherwise they have to setup another env variable. that's a negative for user experience. so how's this a bug?
secondly - HADOOP_HOME is not user defined today. From hadoop-config.sh:
# the root of the Hadoop installation
export HADOOP_HOME=`dirname "$this"`/..
so user's are not setting this today (just in case there is a misconception)
I would agree with renaming 'HADOOP' in hive shell scripts with 'HADOOP_HOME' - that would certainly seem more consistent. would not agree with giving it precedence over path setting since no one sets this env variable today and everyone sets hadoop into their path.
> The configuration file lists two paths to hadoop directories (bin and conf). Startup should check that these are valid directories and give appropriate messages.
> ------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: HADOOP-4090
> URL: https://issues.apache.org/jira/browse/HADOOP-4090
> Project: Hadoop Core
> Issue Type: Improvement
> Components: contrib/hive
> Affects Versions: 0.19.0
> Reporter: Ashish Thusoo
> Priority: Minor
> Attachments: hadoop-4090.patch
>
>
> The configuration file lists two paths to hadoop directories (bin and conf). Startup should check that these are valid directories and give appropriate messages.
> Reported by Chad.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.