[torqueusers] php based web interface for qstat viewing

arnaud.renard at univ-reims.fr arnaud.renard at univ-reims.fr
Thu Apr 19 02:21:44 MDT 2007


Hi all,

i'm new to this mailing list, so sorry if i have stupid questions.

I'm looking for a php based web interface for qstat viewing, pearhaps like it was describes there :
http://www.supercluster.org/pipermail/torqueusers/2006-November/004711.html
Can you give me some url ?

-thanks-


-- 
Arnaud Renard, 
tél : +33 326 91 85 91 - fax : +33 326 91 33 97
http://cosy.univ-reims.fr/~arenard/
Université de Reims Champagne-Ardenne - CReSTIC SysCom
UFR Sciences Exactes et Naturelles
Département de Mathématiques et Informatique
Moulin de la Housse - BP 1039
51687 Reims Cedex 2. 


  ----- Original Message ----- 
  From: Fabio Martinelli 
  To: torqueusers at supercluster.org 
  Sent: Wednesday, April 18, 2007 11:03 PM
  Subject: [torqueusers] cloning the installation jobs for a generic LCG site


  Hi all,

  I manage an LCG Grid site in Italy with about 15 HPC linux nodes,
  normally every node see the application software from a 
  NFS storage element and the application software is installed
  from outside my site by someone that submit an 'installation job': 
  when the job arrive it runs on a node and it modifies the NFS area.
  I never know when these jobs could arrive. 

  this is a terrible deployment strategy for me because:
  1) you don't know when/where/why something change in your cluster
  2) you can't check if something is changed after the installation
     because you don't have a package manager, so you don't know
     which is a correct status for the software ( permissions? crc? owner? size?)
  3) the NFS area is a single point of failure
  4) the NFS area is a performance bottleneck for all the nodes: there is not
     the concept of 'cache', so during the time the nodes read many times
     the same files
  5) my nodes have hard disks with a lot of storage not used: sure almost
     every node can store all the application software ( that's about 90Gb )
     without problems.

  all the others Grid sites are designed in this way!

  for at least mitigate this scenario I want:
  1) Install the application software everywhere it's possible: where it's 
     not possible continue to use NFS for now ( or AFS with a big cache? or what else? )
  2) run tools like tripwire or samhain to understand when/where/why something change 
     on each node ( or what else? )
  3) the most important point for Torque/Maui: run N copy of the installation job when that arrives,
     one for each node that can store the application software.

  I well know the list of users that run the install job, they are about 15:
  so how may solve this in Torque? do you have suggestions for a better deployment design?
  I want more security and performances from my cluster.

  many thanks in advance,
  Fabio Martinelli




  -- 
  ------------------------------------------------------ 
   Dr. Fabio Martinelli 
   Grid Computing Lab 

  institute INFN Tor Vergata 
    address via della ricerca scientifica, 00133 Roma 
      phone +39 06 7259 4113 
            +39 06 7259 4036 
      email fabio.martinelli at roma2.infn.it 
            fabio.martinelli at cern.ch 
        web http://grid.roma2.infn.it/ 





     




------------------------------------------------------------------------------


  _______________________________________________
  torqueusers mailing list
  torqueusers at supercluster.org
  http://www.supercluster.org/mailman/listinfo/torqueusers
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20070419/b22796e4/attachment-0001.html


More information about the torqueusers mailing list