Bulletin 164 - 2006 December 11

1. Holidays
2. January tape silo outage and cherax upgrade
3. Accessing HPCCC web pages from external networks
4. New NEC MPI/SX manual
5. Overload queue on cherax
6. Viewing job status on cherax and clusters
7. Removing csxxx usernames on cherax
8. Subversion talk at CSIRO CMAR on 12th December 11am
9. DMF 15th Anniversary

Note: "CSIRO" items can apply to BoM users of cherax and burnet

The days 25th December to 1st January 2007 are not scheduled working days for the HPCCC and the CSIRO HPSC team members, and only limited support will be available, except for operational processing.

[ page top ]

There will be an outage of the tape silos likely in the first week of January, which will affect the CSIRO datastore and the Bureau's SAM-FS. During this outage cherax will have a software upgrade to SLES10 together with SGI's Pro Pack 5. There are no known user impacts with this upgrade.

[ page top ]

A reminder, when outside the Bureau and CSIRO networks you can access HPCCC intranet web pages by selecting the Intranet button and logging in. The web validation system is separate from your system user account, and your regular username and password will not work; you can get login details by selecting the Intranet button while on the Bureau or CSIRO networks, or by contacting HPCCC staff.

[ page top ]

NEC has recently provided a new manual under the manual improvement project for the HPCCC.

The new manual is: SUPER-UX MPI/SX User's Guide and it is accessible from http://www.hpccc.gov.au/hpccc/userdocs/MUG/index.shtml

[ page top ]

There is a new queue on cherax, the overload queue. It is for jobs that can run without being assigned almost exclusively to CPUs and memory.

It has the best turnaround for short single-CPU jobs. It is a really good idea when the ordinary queue is swamped. It is also appropriate for I/O bound jobs. Just include -q overload in your job script.

[ page top ]

qstat - command line query of the batch system, but no scheduling information (for which use showq).

pbstop - terminal based summary, similar to "top".

showq - shows Maui scheduler's view of the order in which it intends to start jobs, Note that we are now running the Maui scheduler on top of the torque batch system, to allow fair-share scheduling and other features. For more information on maui, see its home web site: www.clusterresources.com/pages/products/maui-cluster-scheduler.php. We plan to move to the more capable moab scheduler.

xpbs and xpbsmon - recently installed on burnet to provide a GUI access and monitoring for the batch system.

[ page top ]

When we switched in 2004 to payid usernames for CSIRO users, we preserved the old csxxx usernames as aliases so that scripts with ~csxxx would continue to work. This is using a standard unix capability, but some software doesn't handle the multiple usernames well. Consequently we will be removing those duplicate usernames on 10th January 2007. Note this doesn't apply to csxxx usernames which don't share UIDs with payid usernames. Please check for ~csxxx in scripts.

[ page top ]

NEC experts Joerg Henrichs and Stephen Leak will be giving a presentation on the Subversion (svn) version control system. The place and time are CSIRO Marine and Atmospheric Research, Aspendale from 11-12 on Tuesday 12th December. If you want to attend please contact Aaron McDonough on 03 9669 8133 or to .

[ page top ]

The CSIRO Datastore has been using the DMF (Data Migration Facility) since 14th November 1991. For most of this time we used the Cray version of DMF, but we now use the SGI Linux implementation.

Users can see total holdings at the traffic lights (the status page at http://www.hpccc.gov.au/hpccc/system_status/).

[ page top ]