[Completed] Urgent Scheduled Maintenance for ASPIRE 2A on 22 January 2025, 9AM to 6.30PM

Dear ASPIRE 2A Users,

We are pleased to announce that the ASPIRE 2A urgent scheduled system maintenance has been completed. You may proceed to access the systems as per usual.

Should you have any questions or need assistance, please contact our Helpdesk via the Service Desk Portal or email us at [email protected] if you have any questions.

Thank you for your understanding and patience.

Warm regards,

The NSCC Team

Urgent Scheduled Maintenance for ASPIRE 2A on 22 January 2025, 9AM to 6.30PM

Dear ASPIRE 2A Users,

We wish to inform you of an upcoming urgent scheduled system maintenance for ASPIRE 2A to enhance its long-term reliability, uptime, and stability.

Maintenance Details:

  • Start: 22 January 2025 (Wednesday), 9:00 AM SGT
  • End: 22 January 2025 (Wednesday), 6:30 PM SGT
  • Duration: 9.5 hours

Purpose:

To replace the faulty Cooling Distribution Unit (CDU) Actuator.

Impact During the Maintenance Period:

  • Unavailable Compute Resources:
    • 33% of CPU compute nodes
    • All GP-GPU compute nodes
    • 75% of large memory nodes
  • Available Compute Resources:
    • 66% of CPU compute nodes
    • All AI-GPU compute nodes
    • 25% of large memory nodes
  • Users will still be able to log in and access the ASPIRE 2A system during this time.

We apologise for any inconvenience this may cause and thank you for your understanding. Should you have any questions or need assistance, please contact our Helpdesk via the Service Desk Portal or email us at [email protected] if you have any questions.

Thank you for your understanding and patience.

Warm regards,

The NSCC Team

Removal of OpenMPI Modules on ASPIRE 2A

Dear Users,

The following OpenMPI modules will be removed from ASPIRE 2A on 3 February 2025:

openmpi/4.1.4-aocc4.0

openmpi/4.1.5-aocc4

openmpi/4.1.5-gcc11

openmpi/4.1.5-icc22

openmpi/4.1.5-icc24

These modules were built statically, and following the update of PBS to a newer version, they no longer function correctly across multiple nodes. While MPI processes can still be spawned within a single compute node, these modules fail to communicate across nodes. As a result, applications built with these versions of OpenMPI will not run successfully on multiple nodes.

To ensure continuity in your workflows, we recommend transitioning to the following OpenMPI versions:

openmpi/4.1.2-hpe(default)

openmpi/4.1.6-gcc11

openmpi/4.1.7-gcc11

openmpi/4.1.7-icc24.2.1 

openmpi/5.0.5-gcc11

openmpi/5.0.5-icc24.2.1 

If you encounter any issues while using these versions, please do not hesitate to contact our helpdesk via the Service Desk Portal or email us at [email protected].

Thank you for your understanding and cooperation.

Warm regards,

NSCC Team