Collecting ganglia monitoring records

Preparation steps:

1. Login as root and create a regular user account on host mon-lcg.projects.cscs.ch
2. Login as user and put a valid certificate in your home dir.
   If creating on a machine anew start with a grid-cert-request command and follow all the instructions
   If bringing a certificate (usercert.pem and userkey.pem files) from another machine:
   % mkdir .globus
   % chmod g-w .globus
   % cp userkey.pem usercert.pem .globus/
   % chmod 600 .globus/userkey.pem
   
3. Collect certificate information
   % grid-cert-info -subject
   % grid-cert-info -issuer
4. Send the output from above commands to the person who deals with registering users on a remote machine. For Phoenix cluster info the points of contact are Kerstin Voss kerstinv@upb.de and Georg Birkenheuer birke@upb.de. After the confirmation that the user is added o the grid-map file can collect and transmit the data.

Executiion steps:

1. Initialize certificate proxy
   % grid-proxy-init
2. Run script to collect the data.
   % ./GetDump
3. Transfer the resulting YYYYMMDD-HHmm.tar.gz file to the remote machine
   % globus-url-copy file:///home/nadya/20070213-1108.tar.gz gsiftp://arminius-grid.uni-paderborn.de/tmp/20070213-1108.tar.gz
4. send email notification to the points of contact listed above. Specify the file name and any other info if needed.

Script GetDumps provides for collection of the ganglia records. The script assumes the following (all can be changed):

• it is run on the host mon-lcg.projects.cscs.ch where ganglia records are stored
• ganglia's DB location is /var/lib/ganglia/rrds/. Two current subdirectories that hold records for the nodes are PHOENIX-services/ and PHOENIX-workers/
• repeated running of the script will create another gzipped tar file with dumps inside.

-- NadyaWilliams - 13 Feb 2007