To clean up a failed concatenate job, do the following:
========================================================

0) Put a .ignore file in the control area of the dataset!

-----------------------------------------------------------
1) Delete the data files:
cd to the "data assembly" area (for instance zrua5z):
> cd /cdf/data00/RU/_zrua5z/  
 
Are there any files here?  Write down the names, and rm these.
(my*.root files don't count but rm them anyway)

------------------------------------------------
2) Delete the file "orphan" entries in the DFC:

>dfc_delete_orphans.csh  dfc_amit zrua5z

The names should be the same as noted down in step 1)

------------------------------------------------------
3) List the completed filesets and delete them. If
   some filesets were made in earlier concat jobs which
   were successful, leave them.
> list_ru zrua5z
----> RU1536.0
        .
        .
        .

> dfc_delete_fs dfc_amit RU1536.0
                            .
                            .
                            .

Now unstatic these fileset(s) that were deleted:

> stage_update_csh C RU1536.0  
                       .
                       .
                       .
------------------------------------------------------
4) Remove these deleted fileset(s) from tape archive list:

> rm /cdf/data00/DATA_INFO/ontape/putontape_RU11536.0
                                            .
                                            .
                                            .
-----------------------------------------------------
5) concat.started and concat.done files.  These contain
   the list of files given to concat jobs and completed
   by concat jobs, respectively.  So if a concat job
   crashes or must be repeated, they must be restored to 
   their state before the crashed/repeated concat job began.
If no previous concat jobs were successful,
   go to control area, redo  concat.started file:

> cd /cdf/data00/DATA_INFO/RU/_zrua5z
> cd log
> mv concat.started concat.started.old
> touch concat.started
> chmod  g+w concat.started
DON'T TOUCH the  concat.done file.

Note that if this is a second pass of input files being
  concatenated, edit concat.started to keep the first pass
  files, assuming they were successful, and remove the
  second pass files. Don't forget the chmod g+w.
 
If a concat job was successful but needs to be repeated,
  then also restore concat.done and remember that it also needs 
  chmod g+w.
-------------------------------------------------------
You can restart the concatenate job now:  Remove the .ignore
file (and concat.ignore file if it exists).
=========================================================