Procedure to recreate the scratch

There are two steps to recreate the scratch filesystem.

1. Recreate the filesystem structure: Make sure that the following directories exist in the scratch filesystem:

   # Create Nordugrid skeleton. After this, you need to restart ARC services.
   cd /gpfs
   mkdir -p home/nordugrid-atlas/sessiondir
   mkdir -p home/nordugrid-atlas/cache
   chown nordugrid-atlas.atlas home/nordugrid-atlas -R
   chmod 1777 home/nordugrid-atlas/sessiondir
   mkdir /gpfs/home/egee/                                   # Required for /home/egee/, WNs and CREAM have a symbolic link in /home/egee to /lustre/scratch/home/egee
   mkdir /gpfs/home/egi_PP/                                # Required for PP
   mkdir /gpfs/tmpdir_pbs                                    # Where the jobs run
   
   #/CREAM_CE/cream01/                # <-- Only if the cream01 stores the shared filesystem to store the cream sandbox
   
   

2. Destroy user accounts and all data related to the old grid users existing in the system.

   /opt/cscs/sbin/clean_grid_accounts.bash

3. Run YAIM
4. Set the sticky bit and correct permissions in tmpdir_pbs

   chmod 1777 tmpdir_pbs/ 

5. Make it visible for our local scripts

   echo OK > /gpfs/chk_mount

-- MiguelGila - 2011-10-21