FAQ
Hi,

I'm trying to move move the namenode to a different machine (move it from
machine B to machine A), since the manager didn't automatically install
Hadoop and assign it as namenode on it's own machine which I though it
would.

I've been following this documentation:
https://ccp.cloudera.com/display/FREE374/Moving+the+NameNode+to+A+Different+Host

The instructions say:
"Using the command line, make a backup copy of the dfs.name.dir directories
on the existing NameNode host. Make sure you backup the fsimage and edits
files. They should be the same across all of the directories specified by
the dfs.name.dir property. You can view the setting for this property in
the HDFS Service > Configuration tab. "

However, where are dfs.name.dir directories? Is this the actual name or a
template name? I found a dfs directory, and it contains directories nn, dn,
and snn subsdirectories, each with a 'current' directory. Is that what the
docs are talking about? Copy over all the subdirectories under dfs?

Currently machine B is everything (NN,SN,DN,Balancer), and I added machine
A as a datanode. Do I need to have both machines as datanodes if I want to
store data on them, even if it's already the namenode?

Thanks!

Gene

Search Discussions

  • Joey Echeverria at Sep 26, 2012 at 2:19 pm
    If you look at the current configuration for HDFS in the CM GUI and
    use the filter box to look up "dfs.name.dir", you should see how your
    system is configured.

    In all likelihood, it's in the dfs/nn directory that you've already
    found, but it's better to double check.

    -Joey
    On Tue, Sep 25, 2012 at 6:21 PM, Gene wrote:
    Hi,

    I'm trying to move move the namenode to a different machine (move it from
    machine B to machine A), since the manager didn't automatically install
    Hadoop and assign it as namenode on it's own machine which I though it
    would.

    I've been following this documentation:
    https://ccp.cloudera.com/display/FREE374/Moving+the+NameNode+to+A+Different+Host

    The instructions say:
    "Using the command line, make a backup copy of the dfs.name.dir directories
    on the existing NameNode host. Make sure you backup the fsimage and edits
    files. They should be the same across all of the directories specified by
    the dfs.name.dir property. You can view the setting for this property in the
    HDFS Service > Configuration tab. "

    However, where are dfs.name.dir directories? Is this the actual name or a
    template name? I found a dfs directory, and it contains directories nn, dn,
    and snn subsdirectories, each with a 'current' directory. Is that what the
    docs are talking about? Copy over all the subdirectories under dfs?

    Currently machine B is everything (NN,SN,DN,Balancer), and I added machine A
    as a datanode. Do I need to have both machines as datanodes if I want to
    store data on them, even if it's already the namenode?

    Thanks!

    Gene





    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.
  • Joey Echeverria at Sep 26, 2012 at 7:18 pm
    Sorry, I missed that question.
    " Do I need to have both machines as datanodes if I want to store data on them, even if it's already the namenode?"
    Yes, if you want to store data on the nodes, then you need to add the
    datanode role to them. The namenode only stores metadata.

    -Joey
    On Wed, Sep 26, 2012 at 3:14 PM, Beidl, Gene wrote:
    Thank you, I'll try that.

    What about:

    " Do I need to have both machines as datanodes if I want to store data on them, even if it's already the namenode?"

    Thanks,

    Gene

    -----Original Message-----
    From: Joey Echeverria
    Sent: Wednesday, September 26, 2012 7:19 AM
    To: Beidl, Gene
    Cc: scm-users@cloudera.org
    Subject: Re: clarification on namenodes

    If you look at the current configuration for HDFS in the CM GUI and use the filter box to look up "dfs.name.dir", you should see how your system is configured.

    In all likelihood, it's in the dfs/nn directory that you've already found, but it's better to double check.

    -Joey
    On Tue, Sep 25, 2012 at 6:21 PM, Gene wrote:
    Hi,

    I'm trying to move move the namenode to a different machine (move it
    from machine B to machine A), since the manager didn't automatically
    install Hadoop and assign it as namenode on it's own machine which I
    though it would.

    I've been following this documentation:
    https://ccp.cloudera.com/display/FREE374/Moving+the+NameNode+to+A+Diff
    erent+Host

    The instructions say:
    "Using the command line, make a backup copy of the dfs.name.dir
    directories on the existing NameNode host. Make sure you backup the
    fsimage and edits files. They should be the same across all of the
    directories specified by the dfs.name.dir property. You can view the
    setting for this property in the HDFS Service > Configuration tab. "

    However, where are dfs.name.dir directories? Is this the actual name
    or a template name? I found a dfs directory, and it contains
    directories nn, dn, and snn subsdirectories, each with a 'current'
    directory. Is that what the docs are talking about? Copy over all the subdirectories under dfs?

    Currently machine B is everything (NN,SN,DN,Balancer), and I added
    machine A as a datanode. Do I need to have both machines as datanodes
    if I want to store data on them, even if it's already the namenode?

    Thanks!

    Gene





    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.


    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.
  • Gene at Sep 26, 2012 at 11:59 pm
    Thank you again, I'm keeping the nanmenode as a datanode too (it's a small
    cluster).

    I still haven't been able to find documentation on the secondary name node.
    Do I have to go through the same procedure as changing the name node or can
    I simply delete the existing secondary node and configure one of my other
    machines as the secondary (through the manager)?

    Thank you again for all the help.
    On Wednesday, September 26, 2012 12:18:42 PM UTC-7, Joey Echeverria wrote:

    Sorry, I missed that question.
    " Do I need to have both machines as datanodes if I want to store data
    on them, even if it's already the namenode?"

    Yes, if you want to store data on the nodes, then you need to add the
    datanode role to them. The namenode only stores metadata.

    -Joey
    On Wed, Sep 26, 2012 at 3:14 PM, Beidl, Gene wrote:
    Thank you, I'll try that.

    What about:

    " Do I need to have both machines as datanodes if I want to store data
    on them, even if it's already the namenode?"
    Thanks,

    Gene

    -----Original Message-----
    From: Joey Echeverria [mailto:jo...@cloudera.com <javascript:>]
    Sent: Wednesday, September 26, 2012 7:19 AM
    To: Beidl, Gene
    Cc: scm-...@cloudera.org <javascript:>
    Subject: Re: clarification on namenodes

    If you look at the current configuration for HDFS in the CM GUI and use
    the filter box to look up "dfs.name.dir", you should see how your system is
    configured.
    In all likelihood, it's in the dfs/nn directory that you've already
    found, but it's better to double check.
    -Joey

    On Tue, Sep 25, 2012 at 6:21 PM, Gene <gene....@mtvn.com <javascript:>>
    wrote:
    Hi,

    I'm trying to move move the namenode to a different machine (move it
    from machine B to machine A), since the manager didn't automatically
    install Hadoop and assign it as namenode on it's own machine which I
    though it would.

    I've been following this documentation:
    https://ccp.cloudera.com/display/FREE374/Moving+the+NameNode+to+A+Diff
    erent+Host

    The instructions say:
    "Using the command line, make a backup copy of the dfs.name.dir
    directories on the existing NameNode host. Make sure you backup the
    fsimage and edits files. They should be the same across all of the
    directories specified by the dfs.name.dir property. You can view the
    setting for this property in the HDFS Service > Configuration tab. "

    However, where are dfs.name.dir directories? Is this the actual name
    or a template name? I found a dfs directory, and it contains
    directories nn, dn, and snn subsdirectories, each with a 'current'
    directory. Is that what the docs are talking about? Copy over all the
    subdirectories under dfs?
    Currently machine B is everything (NN,SN,DN,Balancer), and I added
    machine A as a datanode. Do I need to have both machines as datanodes
    if I want to store data on them, even if it's already the namenode?

    Thanks!

    Gene





    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.


    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.
  • Joey Echeverria at Sep 27, 2012 at 12:01 am
    The secondary namenode doesn't need to keep state, so you can safely
    move it to another node.

    -Joey
    On Wed, Sep 26, 2012 at 7:51 PM, Gene wrote:
    Thank you again, I'm keeping the nanmenode as a datanode too (it's a small
    cluster).

    I still haven't been able to find documentation on the secondary name node.
    Do I have to go through the same procedure as changing the name node or can
    I simply delete the existing secondary node and configure one of my other
    machines as the secondary (through the manager)?

    Thank you again for all the help.

    On Wednesday, September 26, 2012 12:18:42 PM UTC-7, Joey Echeverria wrote:

    Sorry, I missed that question.
    " Do I need to have both machines as datanodes if I want to store data
    on them, even if it's already the namenode?"
    Yes, if you want to store data on the nodes, then you need to add the
    datanode role to them. The namenode only stores metadata.

    -Joey
    On Wed, Sep 26, 2012 at 3:14 PM, Beidl, Gene wrote:
    Thank you, I'll try that.

    What about:

    " Do I need to have both machines as datanodes if I want to store data
    on them, even if it's already the namenode?"

    Thanks,

    Gene

    -----Original Message-----
    From: Joey Echeverria
    Sent: Wednesday, September 26, 2012 7:19 AM
    To: Beidl, Gene
    Cc: scm-...@cloudera.org
    Subject: Re: clarification on namenodes

    If you look at the current configuration for HDFS in the CM GUI and use
    the filter box to look up "dfs.name.dir", you should see how your system is
    configured.

    In all likelihood, it's in the dfs/nn directory that you've already
    found, but it's better to double check.

    -Joey
    On Tue, Sep 25, 2012 at 6:21 PM, Gene wrote:
    Hi,

    I'm trying to move move the namenode to a different machine (move it
    from machine B to machine A), since the manager didn't automatically
    install Hadoop and assign it as namenode on it's own machine which I
    though it would.

    I've been following this documentation:
    https://ccp.cloudera.com/display/FREE374/Moving+the+NameNode+to+A+Diff
    erent+Host

    The instructions say:
    "Using the command line, make a backup copy of the dfs.name.dir
    directories on the existing NameNode host. Make sure you backup the
    fsimage and edits files. They should be the same across all of the
    directories specified by the dfs.name.dir property. You can view the
    setting for this property in the HDFS Service > Configuration tab. "

    However, where are dfs.name.dir directories? Is this the actual name
    or a template name? I found a dfs directory, and it contains
    directories nn, dn, and snn subsdirectories, each with a 'current'
    directory. Is that what the docs are talking about? Copy over all the
    subdirectories under dfs?

    Currently machine B is everything (NN,SN,DN,Balancer), and I added
    machine A as a datanode. Do I need to have both machines as datanodes
    if I want to store data on them, even if it's already the namenode?

    Thanks!

    Gene





    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.


    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.


    --
    Joey Echeverria
    Principal Solutions Architect
    Cloudera, Inc.

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupscm-users @
categorieshadoop
postedSep 25, '12 at 10:21p
activeSep 27, '12 at 12:01a
posts5
users2
websitecloudera.com
irc#hadoop

2 users in discussion

Joey Echeverria: 3 posts Gene: 2 posts

People

Translate

site design / logo © 2022 Grokbase