Grokbase › Groups › Hadoop › scm-users › March 2013
FAQ
Hi,

The answer here rather depends on what you're running. eg: If you're trying
to submit a job you'd use 'hadoop jar'
but I assume that's not your situation here.

If you're trying to run a program that respects the classpath, you would
append the output of 'hadoop classpath'
to the classpath of your program. This command automatically calculates the
classpath that's correct for the
currently active parcel.

--phil

On 13 March 2013 18:04, Rainman wrote:

Can anyone please tell me for this version what am I supposed to see under
/usr/lib and what should be my env. variables to run java client programs?
In the older version after install, I had links to
/usr/lib/hadoop-mapreduce etc etc..I see none after upgrade.

It seems to have installed under /opt/cloudera and I see parcels folder
(under CDH-4.2.0-1.cdh4.2.0.p0.10). I used auto install using the manager.
I added hdfs/mr1 services.

Thanks


--

Search Discussions

  • Philip Langdale at Mar 18, 2013 at 11:53 pm ⇧
    Hi,

    Parcels are self-contained and, by default, live under
    /opt/cloudera/parcels (but this is configurable in the
    Agent configuration). You can look inside this directory and you'll see one
    directory for each distributed
    parcel. Your manager is showing all green because everything is working
    correctly - which is what I'd
    hope you'd see :-)

    We do not create symlinks into /usr/lib - that's deliberate - to avoid
    situations where someone tries to
    place a jar in those locations expecting them to work properly - they'll
    work great until they switch to
    a newer parcel and the jar is now 'lost' inside the old parcel.

    If you're trying to run an example jar, then use the /opt/ path rather than
    a /usr/lib path and everything
    should work the same.

    --phil

    On 15 March 2013 13:42, Rainman wrote:

    Hi,

    Thanks but let me clarify my question. I did a clean install of CDH using
    the manager (v4.5). I need to know what gets installed where to validate
    that my install directories are all in place along with any symbolic links.
    Cluster status itself in the manager is showing all green with hdfs and
    mapreduce (v1).

    My suspicion is that the installation might not have created all the links
    in /usr/lib needed to set classpaths etc for running examples on command
    line. How does your install tree look under /usr/lib. (mine is installed
    under opt as in /opt/cloudera/parcels/CDH-4.2.0-1.cdh4.2.0.p0.10). Does
    install create any links outside of the /opt folder?

    Thanks
    On Friday, March 15, 2013 11:29:46 AM UTC-7, Philip Langdale wrote:

    Hi,

    The answer here rather depends on what you're running. eg: If you're
    trying to submit a job you'd use 'hadoop jar'
    but I assume that's not your situation here.

    If you're trying to run a program that respects the classpath, you would
    append the output of 'hadoop classpath'
    to the classpath of your program. This command automatically calculates
    the classpath that's correct for the
    currently active parcel.

    --phil

    On 13 March 2013 18:04, Rainman wrote:

    Can anyone please tell me for this version what am I supposed to see
    under /usr/lib and what should be my env. variables to run java client
    programs?
    In the older version after install, I had links to
    /usr/lib/hadoop-mapreduce etc etc..I see none after upgrade.

    It seems to have installed under /opt/cloudera and I see parcels folder
    (under CDH-4.2.0-1.cdh4.2.0.p0.10). I used auto install using the manager.
    I added hdfs/mr1 services.

    Thanks


    --


    --

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupscm-users @
categorieshadoop
postedMar 15, '13 at 6:29p
activeMar 18, '13 at 11:53p
posts2
users1
websitecloudera.com
irc#hadoop

1 user in discussion

Philip Langdale: 2 posts

Content

People

Support

Translate

site design / logo © 2022 Grokbase