1) jsub command
There is a new version of jsub now installed that allows transparent access to
data files in the silo. If you specify the file name(s) to be processed (key
words INPUT_FILES and OUTPUT_FILES) beginning with /mss (i.e. a stub file) then
the data will be staged out of the silo and copied to the farm. There is no
longer any need to run jget before submitting the job, or to run jput after
job completion. Data files with names beginning with anything other than /mss
will be treated as before.
2) jkill command
The new jkill command permits user termination of queued and executing jobs on
the farm. You can kill (or signal) individual jobs or groups of jobs. Please
refer to the jkill man page for me details.
3) jput/jget new functionality.
A new option (-w) that causes the command to block until the tape copy has
been done. This is useful in scripts for example, where you may want to stage
the tape and wait until the operation is complete before trying to access the
data.
Multi-file copies. Jput and jget can now copy multiple files and recursively
copy directory trees. However, please note that there is a limit of 50 files
that will be moved in one operation. If you find that you need more than
this, consider making a tar archive of the files before running jput. This
will usually be far more efficient.
Creation of stub file directories on the fly. jput now will create a
directory under /mss that you specify as the destination for your stub files.
However, it will only do this if the requesting user has write permission in
the parent directory. For most users this means they will only be able to
create mss directories under /mss/home/<user>.
The man pages will be updated to reflect the new functionality.
Any problems to : farm_support@jlab.org