[20044] in Athena Bugs
Long job support needed
daemon@ATHENA.MIT.EDU (Thomas E Cavin)
Wed Dec 5 17:31:59 2001
From: Thomas E Cavin <cavin@MIT.EDU>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <15374.41054.245888.762183@lap1-wccf.mit.edu>
Date: Wed, 5 Dec 2001 17:31:58 -0500
To: Athena Bugs list <bugs@mit.edu>
CC: Tom Cavin <tec@ai.mit.edu>
Hi Folks,
This isn't strictly a bug report, but more of a plea for support of long
jobs in some manner.
I have two labs that are running private Athena workstations and Athena
KNFS servers on moderately sized RAID systems. The type of work they do
involves getting large datasets in multiple ~1 MB files and having Matlab
scripts do the processing using a custom analysis package called SPM.
The problem is that some of these scripts can take upwards of 20 hours to
run.
The preferred way of working is to use a client system and access the data
via the KNFS mounts. Unfortunately, once the tickets expire the jobs
break, and some of these researchers have lost days of work restarting jobs
and hoping for the best.
One possible solution is to copy lockers to the local client systems. The
problem with this solution is that it becomes a maintenance nightmare.
Is there any way to get something like 24 hour tickets? What is the basic
limit on ticket lifetime?
I think they could live with a solution that let them reset their tickets
before they left for the day and renew them when they come back in the next
day in order to keep a continuous job running beyond the normal maximum
ticket length. (I don't know enough about the technical aspects of
Kerberos to know how practical this is.)
