Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Migrated to Confluence 5.3

...

  1. Copy or move files to new directory (on a production system, you should stop NMIS from polling by editing the crontab and appending a "#" to the lines for NMIS polling). Pick one of these options, C is recommended.

    1. Copy files to a new directory

      Code Block
      cp -rp /usr/local/nmis8/database /opt/nmis8
    2. Move files to a new directory

      Code Block
      mv /usr/local/nmis8/database /opt/nmis8
    3. If you have a large database and copying or moving files takes a significant amount of time then we will prefer to use rsync.  In this case do not stop NMIS from polling at this time, it will be done in a later step. We will do a "hot" copy now and follow it with a "cold" copy.

      Code Block
      #do a hot copy where nmis is still updating
      rsync -av --inplace /usr/local/nmis8 /opt/
  2. Check permissions on the copied files. The above commands should preserve owners and permissions, but double check anyway

    Code Block
    ls -l /opt # nmis8 should be owned by nmis, group nmis
    ls -l /opt/nmis8/ # all should be owned by nmis, group nmis
    
    #if the have incorrect permissions run these commands: 
    chown -Rv nmis:nmis /opt/nmis8
    chmod -Rv g+w /opt/nmis8 
  3. If you have a large database, now is the time to stop NMIS from polling.  As mentioned previously, the safest way to do this is by commenting out the lines that run NMIS polling, be sure to do all of them, just to be safe. Another option (and not necessarily recommended) is to stop the cron daemon with: service crond stop #Do Not forget to restart it later.  Now we will run the rsync copy one more time to get any updates: 

     

    1. Stop polling.  Wait for nmis to stop updating. 

      Code Block
      crontab -e #comment out all nmis.pl lines
      # continue running the following command until you no longer see nmis running 
      ps aux | grep nmis.pl 
      # make sure you don't see a line like this:
      # /usr/bin/perl ./nmis.pl
    2. Now that NMIS is no longer polling we can run a "cold" copy to get the last updates that NMIS made while we were "hot" copying. 

      Code Block
      sudo rsync -av --inplace /usr/local/nmis8 /opt/
  4. Edit config files so that NMIS will write data to our new location, edit /usr/local/nmis8/conf/Config.nmis 

    Code Block
    # this line
    line 13: '<nmis_data>' => '<nmis_base>',
    # should change to
    line 13: '<nmis_data>' => '/opt/nmis8',
  5. To test, move the existing data directores to new folders and symbolically link the new database location to the old one

    Code Block
     mv /usr/local/nmis8/database /usr/local/nmis8/database_old
     ln -s /opt/nmis8/database /usr/local/nmis8/database

    In a new installation, the symbolic link will not be necessary, but when changing the database location in a running NMIS8 installation, putting the SYM link in will help during the update.

  6. Run a manual update, you should see no errors. 

    Code Block
    /usr/local/nmis8/bin/nmis.pl type=update debug=1
    # you should not see any errors like this:
    # createRRD, ERROR (localhost) unable to create /opt/nmis8/database/health/server/localhost-reach.rrd: creating '/opt/nmis8/database/health/server/localhost-reach.rrd': No such file or directory  
  7. Open or reload the NMIS web page and everything should still look the same.
  8. After confirming there are no errors you can now safely delete the old database directory. 

    Code Block
    themeEmacs
    rm -rf /usr/local/nmis8/database_old
  9. Now re-enable your cron jobs.  If you commented lines out, remove the comments.  If you stopped crond, start it again: service crond start

...

We now have NMIS saving all databases to a different folder (which could be on a different partition).

If you have copied data, you will need to delete the files you no longer need to free up disk space.

This article describes a high level process for relocating data, the final process used by an organisation should be tested before running on a production system.