Bulletin 150 - 2005 November 18

1. HPCCC SX-6/TX7 file system problems
2. Incident logs
3. SX-6 job memory sizes
4. netCDF developments
5. cherax - new version of gnuplot
6. cherax - upgrade and problems

There were problems with the /bm/home file system on the SX-6/TX7 system from about 15:00 on Tuesday 15th November.

Users were notified by e-mail of the problems.

Measures have been taken to isolate the machine which caused the troubles from mawson and eccles.

Files were restored from a backup. However, if any users have files missing, or are wanting newer versions than were restored from backup, then please contact the HPCCC - we do have a backup of the corrupted file system from Thursday, and may be able to restore files.

[ page top ]

There is a prototype incident log for cherax at http://intra.hpsc.csiro.au/user/incident_reports/ax/

If this is useful to users, the HPCCC will extend this facility to other machines.

[ page top ]

We have recently raised the scheduler's threshold for memory usage on the SX-6 nodes from 90% to 96.5%, to allow some very large memory jobs to run (55 Gbyte).

However, this has caused at least one instance where a large load of jobs on a node caused some jobs to run out of memory.

The HPCCC will be configuring a small amount of swap space on SX-6 nodes to provide a cushion against such failures.

The HPCCC will again re-consider the limits on batch jobs, as outlined in HPCbull 134.2.4.

[ page top ]

HPCCC users may be interested in the following mailing list item.

From: Russ Rew <russ@unidata.ucar.edu>
Subject: Paper on netCDF-4 now available

A paper (really an extended abstract) to be presented at the 2006 AMS meeting in Atlanta is now available online:

http://www.unidata.ucar.edu/software/netcdf/papers/2006-ams.pdf

"NetCDF-4: Software implementing an enhanced data model for the geosciences" by Russ Rew, Ed Hartnett, and John Caron

The paper provides background on the netCDF-4 project, descriptions of additions to the netCDF data model, and recommendations for data providers and developers. It may be useful in understanding the status and future directions for netCDF. We will also be presenting a poster session on this work at the December AGU meeting in San Francisco.

[ page top ]

On cherax, there is a new version of gnuplot in /usr/local/bin/gnuplot (Version 4.0 patchlevel 0). The old version is in /usr/bin/gnuplot (Version 3.7 patchlevel 3).

The new version has many new features - type help in gnuplot and follow the instructions. A major new capability is mouse control - allowing coordinate viewing, zooming, changing of axis scales, etc.

There are some difference in the handling of missing data - see req #6041 (resolved).

[ page top ]

cherax was upgraded from 64 to 128 processors on Saturday 12th November. There are 120 processors in the main partition, leaving a test partition of 8 processors. This will be used for work such as testing new versions of the operating system.

There have been some problems since - see the incident log described above.

[ page top ]