hive-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jay Vyas <jayunit...@gmail.com>
Subject Re: Hive, datanucleus, jdbc, localmode.
Date Sun, 29 Dec 2013 01:09:54 GMT
-Local mode should have clear instructions on how to run fully local hive jobs, with no hadoop
installation.

-I like the hive_test repo but I'm not yet sure hive_test is 100% up to date with the simplest
strategy for testing hive workflows on the JVM.


> On Dec 28, 2013, at 4:19 PM, Lefty Leverenz <leftyleverenz@gmail.com> wrote:
> 
> This sounds like something the documentation should cover. 
> What information should be added to the Local Mode section? 
> Should the wiki have a link to hive_test (for example, in Hive Developer FAQ)?
> 
> -- Lefty
> 
> 
>> On Sat, Dec 28, 2013 at 8:02 AM, Edward Capriolo <edlinuxguru@gmail.com> wrote:
>> I do not think so. Local mode, just implies the job tracker is local (and some of
the temp storage directories) it does not imply hive will use hadoop without forking. 
>> 
>> 
>>> On Sat, Dec 28, 2013 at 10:43 AM, Jay Vyas <jayunit100@gmail.com> wrote:
>>> Thanks... But are you sure this is the only way? Or is there some magic way to
run hive in local mode that we both are missing out on ?:)...
>>> 
>>> - isn't hive in local mode supposed to be run simply via the jdbc://hive URL
which runs local mode... Or maybe by the fork config parameter?
>>> 
>>> - For example see the parameters in this file:
>>> 
>>> https://github.com/riptano/brisk/blob/master/resources/hive/conf/hive-site.xml
>>> 
>>> 
>>>> On Dec 28, 2013, at 10:22 AM, Edward Capriolo <edlinuxguru@gmail.com>
wrote:
>>>> 
>>>> You can follow along to what I do here.
>>>> 
>>>> https://github.com/edwardcapriolo/hive_test
>>>> 
>>>> Essentially hive requires a HADOOP_HOME because it always wants to fork a
bin/hadoop process. Hive-test helps you unpack hadoop inside target and change your hadoop_home
to some other directory. 
>>>> 
>>>> It would be nice if there was some other way to do this.
>>>> 
>>>> 
>>>>> On Fri, Dec 27, 2013 at 10:27 PM, Jay Vyas <jayunit100@gmail.com>
wrote:
>>>>> Hi Hive:
>>>>> 
>>>>> I'm attempting to create a robust eclipse based dev environment for testing
my hive jobs in localmode however I run into classnotfound errors depending on which maven
dependencies I use. Also, it seems when I change these dependencies from hive 0.12 to hive
0.11, I get other errors related to hive trying to launch jobs via calling /usr/bin/hadoop.
>>>>> 
>>>>> This I am stuck: I can't run hive 12 in local java mode because of subtle
datanucleus class and API inconsistencies which are tough to resolve, and when going to hive
11, it seems local mode is not natively detected via the jdbc URL...
>>>>> 
>>>>> So I have 2 questions:
>>>>> 
>>>>> 0) how does hive 12 versus 11 implement local mode differently ?
>>>>> 
>>>>> And
>>>>> 
>>>>> 1) What is the right way to in hive in pure java/ local environments?
>>>>> 
>>>>> The hive book suggests modifying configuration properties, for local
mode..
>>>>> 
>>>>> but I also have found  that in hive 0.12 , using the jdbc://hive connection
URL automagically launches jobs in local mode..
>>>>> 
>>>>> However in 0.11 , I see calls to /usr/bin/hadoop when running java classes
in local eclipse environment.
>>>>> 
>>>>> Thanks!
>>>>> 
>>>>> FYI to see an example of my pom.xml, you can checkout the github://jayunit100/bigpetstore
pom.xml file.
> 

Mime
View raw message