This file is /cdf/run1/documents/development/cdfsga_disk.note Last update was September 29th, 2000 We have seven kinds of disk on cdfsga. 1) User disk. Directories such as ~wicklund are found on the user disks. Intended for storing permanent files: for example user's fortran programs, tc-shell scripts, TeX files and so on. This disk space is backed up. Note: The disk backups protect against the physical failure of a whole disk. They provide either limited or no protection against inadvertent deletion of files. The backups will be up to three days behind the current contents of the disk. No doubt, you will wish to periodically back up your files onto tape. The default allocation per user is 50 Mbytes (100,000 512 byte blocks) which is five times the default on the FNALD VAXes. The current usage can be seen using the command > quota -v Note: the units reported by this command are 1024 byte blocks. A typical command sequence for summarizing your disk usage in your home area is: > cd > du -s `ls` Netscape may cache files in your user area, in the directory $HOME/.netscape/archive This default can fill up your user area with files that aren't very useful. Here is how to change the defaults in Netscape so that the cache is not stored on the user disk: Click on "Edit" in the top line, which produces a submenu. Then click on "Preferences..." in the submenu. This produces a menu box. Then click on the triangle in front of "Advanced". This should reveal more menu items. Then click on "Cache" Then type into the "Cache Folder:" box text similar to /spool/username/.netscape/archive/ Then click on the "OK" button to commit the change. 2) Disks managed by the physics groups. These are the main repositories of disk space on cdfsga. A typical name is /data01/jet These disks are managed by the physics groups. These disks are not backed up by the system adminstrators at all. The physics groups have responsibility for maintaining copies of the data which resides on these disks. Users requiring write access to the physics group disks should consult the conveners of their group, and then contact one of the people listed below to request write access. top group: M. Kruse mkruse@fnal.gov exotics group: S. Lammel lammel@fnal.gov bottom group: S. Tkaczyk tka@fnal.gov ElectroWeak group: M. Lancaster lmark@fnal.gov QCD group: A. Bhatti bhatti@fnal.gov For better performance, physics groups are encouraged to store CDF YBOS data in UNIX format on these disks. This is done via the Analysis Control command ANA>> OUTPUT FORMAT UNIX Note: Please try to use -o /spool/$LOGNAME/job.log -e /spool/$LOGNAME/job.err qualifiers to the qsub command when submitting a batch job, instead of placing these log files on a physics group disk. The disk quotas on /spool help to prevent the physical disk from filling up. Here are examples of how to find files of a particular physics group which have not been read in over 180 days and which are larger than 500 kBytes: > find /data*/top -atime +180 -size +1000 -print > find /data*/exotic -atime +180 -size +1000 -print > find /data*/bottom /data*/sin2beta -atime +180 -size +1000 -print > find /data*/ewk -atime +180 -size +1000 -print > find /data*/qcd /data*/jet -atime +180 -size +1000 -print 3) Backed up disks managed by the Run II upgrade group. /cdf/software Source code of CDF releases /usr/people/cvsuser Source code of ZOOM releases /cdf/upgrade Source code of test versions /usr/prod2 Run II products These three disk areas are backed up. Note: The disk backups protect against the physical failure of a whole disk. They provide either limited or no protection against inadvertent deletion of files. The backups will be up to three days behind the current contents of the disk. No doubt, you will wish to periodically back up your files onto tape. Note: Please try to use > cd /spool/$LOGNAME before running executables, rather than > cd /cdf/upgrade This is so core files will be deposited on the /spool disk and not on the /cdf/upgrade disk. Users requiring write access to the backed up disks managed by the Run II upgrade group should contact Robert Harris rharris@fnal.gov and request write access to group cdfupg 4) Data disks managed by the Run II upgrade group. /data33/validation /data62/software /data64/upgrade /data69/upgrade /data81/trig /data82/trig /data83/upgrade /data84/upgrade /data92/software /spool2 These disks are meant to hold executables, libraries and monte carlo samples. These disks are not backed up by the system adminstrators at all, so please don't store source code on them. Here is a way to find files on these disks which have not been read in over 180 days and which are larger than 250 kBytes: > find /data*/validation /data*/upgrade /data*/trig /data*/software \ ? -atime +180 -size +500 -exec ls -l {} \; Note: These Run II upgrade disks do not appear when using the command: > quota -v so don't bother trying quota -v for the Run II upgrade disks. Note: Please try to use -o /spool/$LOGNAME/job.log -e /spool/$LOGNAME/job.err qualifiers to the qsub command when submitting a batch job, instead of placing these log files on a upgrade group disk. The disk quotas on /spool help to prevent the physical disk from filling up. Users requiring write access to the data disks managed by the Run II upgrade group should contact Robert Harris rharris@fnal.gov and request write access to group cdfupg_d 5) Scratch disk. Directories such as $USR_SCRATCH/wicklund are found on these disks. ($USR_SCRATCH is also known as /spool). Intended for storing files that are temporary: for example user's executables, user's object libraries, PostScript versions of Tex files, the Netscape cache, and so on. These disks are not backed up. Please keep a copy of important files on 8mm cassette. Please periodically delete unused files from this disk. The default allocation per user is 250 Mbytes (500,000 512 byte blocks) which is two and a half times the default on FNALD. Old files are periodically deleted, (even when the quota is not exhausted). The current usage can be seen with the command: > quota -v Note: the units reported by this command are 1024 byte blocks. If you run out of space on the scratch disk, please contact your physics group, and arrange for access to the disk space of that physics group. Please do not store large files (over 50 Mbytes) on /spool, instead please store such files on the disk space of your physics group. Please do not store pad files on /spool, instead please store such files on the disk space of your physics group. If you have files that are not temporary, please contact your physics group, and arrange for access to the disk space of that physics group. 6) CDF code disks. Directories such as /cdf/code, /cdf/offln, and /cdf/software are found on these disks. These disks are managed by the CDF code librarians. These disks are backed up. 7) Staging disk. When files from the STK silo are requested, they are automatically copied to these disks. Older files are deleted to make room for newer files. The same generic file names are used to refer to the files whether they are actually in the STK or actually on disk. Also note: the same generic file names are used on both FNALD and on cdfsga. An example of use is: ANA>> INPUT FILE/FATMEN "//FNAL/CDF/B615/PAD01/ELE/CCF904A" If you would like to have a set of data files (or a set of Monte Carlo files) copied into the STK silo, please contact: Bill Robertson robertson@hep.phy.duke.edu Note: The disk backups protect against the physical failure of a whole disk. They provide either limited or no protection against inadvertent deletion of files. The backups will be up to three days behind the current state of the disk. The disk backups overwrite older backup tapes, on a two week cycle. The list of the disks being backed up on cdfsga can be found in files: /usr/products/fmb/v6_5b/fmb_files/cdfsga.A /usr/products/fmb/v6_5b/fmb_files/cdfsga.B /usr/products/fmb/v6_5b/fmb_files/cdfsga.C /usr/products/fmb/v6_5b/fmb_files/cdfsga.D /usr/products/fmb/v6_5b/fmb_files/cdfsga.E /usr/products/fmb/v6_5b/fmb_files/cdfsga.F /usr/products/fmb/v6_5b/fmb_files/cdfsga.G /usr/products/fmb/v6_5b/fmb_files/cdfsga.H /usr/products/fmb/v6_5b/fmb_files/cdfsga2.A /usr/products/fmb/v6_5b/fmb_files/cdfsga2.B /usr/products/fmb/v6_5b/fmb_files/cdfsga2.C /usr/products/fmb/v6_5b/fmb_files/cdfsga2.D /usr/products/fmb/v6_5b/fmb_files/cdfsga2.E /usr/products/fmb/v6_5b/fmb_files/cdfsga2.F /usr/products/fmb/v6_5b/fmb_files/cdfsga2.G /usr/products/fmb/v6_5b/fmb_files/cdfsga2.H If the disk doesn't appear in one of these files, it is not backed up.