Changelog¶

Recent changes to the DIPC HPC infrastructure, policies, and documentation.

2026¶

April 20, 2026¶

New email address for HPC account management: accounts-hpc@dipc.org

To streamline our support process and ensure a faster response to your requests, we have created a new email address dedicated exclusively to account administration: accounts-hpc@dipc.org

From now on, all inquiries regarding account creation, renewals, and management (including password resets) must be sent to this new address.

You will also receive all account-related notifications and updates from this address.

Please continue to use support-hpc@dipc.org strictly for technical incidents and queries regarding the use of our resources.

We kindly ask you to use the appropriate channel for your future messages to avoid any delays.

March 24, 2026¶

New documentation site launched

We have completely rewritten our user documentation. The new site covers all three clusters (Hyperion, Atlas EDR, Atlas FDR) with practical guides, real examples, and up-to-date information. Highlights include a Quick Start guide, a batch script generator, expanded guides for containers, GPU computing, and a searchable software catalog. Feedback is welcome at support-hpc@dipc.org.

February 15, 2026¶

SSH key authentication is now mandatory

Password-based access to all DIPC HPC clusters has been permanently disabled. All users must authenticate using SSH keys. If you haven't registered a key yet, follow the instructions in the Account Setup guide. If you have lost access, contact support-hpc@dipc.org.

January 1, 2026¶

Anaconda3 removed, Miniforge3 is the new default

Due to Anaconda's licensing restrictions, Anaconda3 has been fully removed from all clusters. Miniforge3, a lightweight conda installer backed by the open conda-forge channel, is now the only supported Python distribution.

If you previously used module load Python to access Anaconda3, switch to module load Miniforge3. Existing conda environments will continue to work, but we recommend recreating them with conda-forge packages for long-term compatibility. See the Python guide for details.