Bulletin 121 - 2004 Jul 23

1. Moving of HPCCC staff and offices
2. Recent interruptions
3. SYSTEM CHANGES - reminders
4. SYSTEM CHANGE NOTICE 2004-A004 - ERS upgrade - load balance option

The HPCCC offices and staff are scheduled to move from 150 Lonsdale Street to 700 Collins Street starting on 29th July.

There may be some disruption to HPCCC and CSIRO HPSC WWW services.

There was an TX7 failover around 0940 on Thursday 22nd July.

All running SX-6 jobs were terminated as a precautionary measure during the recovery from the TX7 failover. Please re-submit your jobs.

In the early afternoon after the failover, we implemented a new feature ('SYNC') in GFS, which provides a more reliable model for GFS writes, and would allow us to let jobs continue after a TX-7 failover.

However, the system showed some slowness - ls commands, and job throughput, and after advice and information gathering by NEC Tokyo, we reverted to the previous system at around 17:30.

SYSTEM CHANGE NOTICE 2004-C002 Separation of CSIRO multi-CPU jobs has been delayed until Monday 26th July.

SYSTEM CHANGE NOTICE 2004-B002 SXCross structure change will occur on Monday 26th July

SYSTEM CHANGE NOTICE 2004-C001 Local Corrections and tempd will occur on Tuesday 27th July

DATE OF CHANGE: 2004-07-26 starting 11:00 AEST

SYSTEMS AFFECTED: SX-6/TX7 system

SOFTWARE AFFECTED: ERS II

IMPACT: ERS functions will be unavailable for up to 30 minutes during the upgrade.

Jobs will continue to run, but new jobs will not start during the upgrade.

DETAILS:

1. The upgraded ERS provides corrections, and also provides a new option for balancing the load across nodes by allocated resources, rather than just by utilisation.
2. Sometime after the upgrade, the new option will be invoked, along with changes to the ERS CPU usage borders, to reduce over-commitment of multi-processor jobs on nodes.

CONTACT: Ann Eblen

NOTICE APPROVAL: HPCCC Manager