Date: Wed, 22 Sep 2004 15:25:10 -0400 (EDT)
From: Donald Watrous <watrous@cs.rutgers.edu>
To: watrous@cs.rutgers.edu
Subject: Subject: Reminder /filer/tmp is not backed up

This message is being sent to all users using space on LCSR's
experimental filer mounted on the research cluster as /filer/tmp.

Files on this filer are *NOT BACKED UP*.  There are a couple of
implications of this:

  1) Once you delete a file, it's gone.  Calls to the operator or
     systems staff will not get your file back.

  2) Catastrophic failure of the filer will result in all data being
     lost.  While we believe the filer is as stable as possible (that
     is, it is raided so any single disk failure will not cause data
     loss -- to lose data, there would have to be two simultaneous
     disk failures), occasional glitches have resulted in extended
     (on the order of hours) downtime to recover.

Of these two threats to your data, user error is the much greater
probability.

One more point: should the filer become full, we reserve the right to
delete files in order to make it usable for the user community.
Reasonable steps will be taken before files are deleted based on the
urgency of the situation: an attempt will be made to contact owners of
files before they are deleted.  Staff members files will be considered
before user files.  And the possibility of compressing files instead
of deleting them will also be considered.

For your own peace of mind, you should put nothing on the filer if you
would be seriously inconvenienced if it disappeared.  It is good to
use as a repository for a large amount of data downloaded from the net
or generated by yourself.  These files could be relatively easily
downloaded or generated again.  On the other hand, the results of long
term analysis, programs which have a lot of work invested in them,
etc., should not exist only on the filer.

Several options available for online storage within LCSR are discussed
at
    http://www.cs.rutgers.edu/~watrous/disk-space.html

Backed up space is obviously recommended for important data.