BlueCrystal 5 Early Access FAQs

This FAQs page is specific to BlueCrystal 5 (BC5) and is intended to provide timely facility-specific information for users of the system during the early access and post-launch periods.

The content on this page will be updated over time as the system configuration is tuned and issues are encountered and resolved. Where possible this page links to the main documentation rather than repeating it, and the content will eventually be merged into the main documentation and this page deleted.

Early Access

Please note that access to BC5 is in early access while the facility is configured, tuned, and brought to full capacity.

If you encounter issues with the service, please check the section on known issues and workarounds to see if there is a solution or workaround. If your issue is not covered in known issues and workarounds or if you would like to provide general feedback on the service and user experience, please contact the BriCS helpdesk.

Who can access BC5?

Access to BC5 is described on the access page. Permanent academic staff at the University of Bristol can apply for resources via the University of Bristol allocation portal.

How do I get access to BC5?

Once a project is allocated on BC5, the Project Lead receives an email invite to join the project in the BriCS portal. Follow the instructions in the setup tutorial to access the portal and accept the invitation.

To add project members, the Project Lead sends invites by following the managing projects guide. Invited members follow the setup tutorial to accept.

How do I find the allocated resources for my project on BC5?

Allocated and consumed resources can be viewed in the BriCS portal. See the accounting guide for how resource usage is calculated.

How do I log in to BC5?

Follow the login guide, selecting the "BlueCrystal 5" tabs where shown. You will need to set your UNIX short name in the BriCS portal and use clifton to obtain an SSH certificate.

University of Bristol VPN

When away from campus, or connected to a network other than eduroam, you must be connected to the University of Bristol F5 VPN to log in to BC5.

How can I use my software/module?

BC5 follows the same self-service model for software as the other BriCS facilities: see Can you install my application for me? and the Installing and Running Software tutorial for the general approach. This tutorial and the Python, Spack and container guides were written for Isambard-AI and Isambard 3, so some of their content does not apply to BC5. In particular, Arm-specific (aarch64) instructions and the GPU examples (GH200 GPUs) do not apply, as BC5 uses x86_64 CPUs and A100 GPUs.

EESSI

BC5 provides EESSI (European Environment for Scientific Software Installations), an externally-maintained standard set of packages, including build tools, compilers and commonly-used software. Search the online list of available software to see what is available.

To access EESSI, run:

$ module use /cvmfs/software.eessi.io/init/modules
$ module avail EESSI
EESSI/2023.06 (D)    EESSI/2025.06    EESSI/2026.06
$ module load EESSI/2025.06

module avail EESSI lists the available EESSI versions, with (D) marking the default; this is EESSI/2023.06 at the time of writing. Once an EESSI version is loaded, module avail lists the software it provides. These commands must also be run in any Slurm job script that uses EESSI.

How can I run jobs on BC5?

BC5 uses the Slurm workload manager. See the Slurm guide for the main commands. Some examples in the guide are specific to Isambard-AI and Isambard 3 and will need minor modification for BC5.

Use the --partition option of sbatch and srun to select a partition; if it is not specified, jobs are submitted to the default partition (marked with an asterisk below). Partitions group the different types of compute node, see the system specifications for node details.

Partition name	Nodes	Maximum walltime	Notes
general (*)	371 × AMD EPYC CPU: 32 core, 128 GiB RAM	24h	For general purpose CPU workloads
highmem	4 × AMD EPYC CPU: 56 core, 448 GiB RAM	24h	For CPU workloads with high memory requirements
gpu	4 × AMD EPYC CPU + NVIDIA GPU: 16 core, 128 GiB RAM, 1 × NVIDIA A100 GPU	24h	For GPU workloads

Where can I store my data on BC5?

BC5 offers the same shared storage spaces as the other BriCS facilities (but not node-local storage, see below), with paths and quotas given on the storage spaces page. Storage policies are still being tuned during early access, so the page may not yet reflect BC5 completely: for example, its data retention table does not yet include BC5. Storage environment variables and node-local storage are affected by known issues.

How do I get support?

Support is provided through the BriCS helpdesk; see Getting Support for how to submit a ticket. This includes reporting documentation errors or omissions, and general feedback.

Known issues and workarounds

During early access we would be grateful to receive reports of any issues you encounter. Please contact the BriCS helpdesk to open a support ticket.

While working to resolve reported issues we will record known issues and workarounds in the table below.

Issue	Description	Workaround
Storage space environment variables are not available on BC5	The environment variables containing BriCS storage space paths listed on the storage spaces page (e.g. SCRATCH, PROJECTDIR) are not set on BC5. The HOME variable is set correctly.	Use the paths to the storage spaces directly where needed. If your workflow depends on the environment variables, please set these variables manually.
Node-local scratch storage is not available on BC5	The node-local per-user /local/user/<UID> storage space is not configured on BC5 compute nodes currently.	Use one of the other available user storage spaces (e.g. user home) for storing temporary workload data.