RICC for CCJ users

1. Overview
   All CCJ users can use a new computing facility RICC (RIKEN Integrated Cluster of Clusters), which is maintained by RIKEN IT division, since October 1, 2009. You can find several general documentations about RICC in ( English / Japanese ). Actually, 20 computing nodes are available with PHENIX analysis environment as exclusive use for CCJ users.


2. Current configuration (reserved for CCJ users)

   CPU Intel(R) Xeon(R) CPU X5570, Quad-core 2.93GHz (Nehalem) x 2CPU per node Memory 12GB per node Number of nodes for exclusive use 20 nodes (1 for interactive + 19 for batch jobs) Number of CPU core for batch jobs 152 core Operating system Scientific Linux 5.3 (x86_64) Batch job system Condor 7.2.4



3. Get Account
   Please fill and modify the blue words in the following paper ( English / Japanese ), and send it to Yasushi Watanabe (watanaby -at- riken.jp) as an e-mail attachment.


4. How to play
   • Login: Since your home directory is same as for CCJ, you can login to the interactive node at RICC by the ssh from ccjgw or ccjsun after you get account.
     
     (e.g.)

     [user:ccjgw]$ ssh mpc2001

   
   
   • PHENIX environment: "-a" option is required to take over the default environmental variables.
     
     

     [user:mpc2001]$ source /opt/phenix/bin/phenix_setup.csh -a

     
     or
     
     

     [user:mpc2001]$ source /opt/phenix/bin/phenix_setup.csh -a pro.??

   
   
   • Batch job: Condor scheduler, which is much familiar with PHENIX analyzer, is available at RICC. If you do not know about it, please refer to a general manual of Condor in this URL. Firstly, you have to prepare a script for job submission such as,
     
     (sample.cmd)

     Executable = your-script.csh Universe = vanilla getenv = true Output = your-job-$(Process).out Error = your-job-$(Process).err Log = your-job-$(Process).log Arguments = $(Process) Queue 10

     
     and then, just type as follows to submit it. The line "getenv = true" is necessary to take over the environemntal variables in current interactive node.
     
     

     [user:mpc2001]$ condor_submit sample.cmd

     
     You can check the status of your jobs and clusters by the several command as follows.
     
     

     [user:mpc2001]$ condor_q

     [user:mpc2001]$ condor_status

   
   
   • Working area: Each computing node has enough disk space in "/job_tmp" for the input/output of your process, and you can get the location by a variable of "$CCJ_JOBTMP". So, please do not access to the NFS server directly in your jobs and copy from/to the "/job_tmp" by rcpx command for the files of large size. It is to reduce the load of NFS servers (ccjnfs11-15 and ccjnfs20) as described in this URL.