| name | lawrencium-hpc |
| description | Answer questions about high-performance computing at Lawrence Berkeley National Laboratory (LBNL) — the Lawrencium (LRC) CPU cluster and Einsteinium GPU cluster. Covers accounts & MFA, logging in, Slurm job scheduling and example scripts, the Software Module Farm (compilers, Python/R/Julia, MPI, CUDA and other libraries, PyTorch/TensorFlow/Ray, VASP, Ollama/vLLM), Open OnDemand and Jupyter, data transfer with the lrc-xfer node and Globus, and the supported research (condo) clusters. Use whenever a question concerns Lawrencium, Einsteinium, LBNL/Berkeley Lab research computing, or running jobs on these systems. Fetch the Markdown source from the scienceit-docs GitHub repository instead of scraping HTML from scienceit-docs.lbl.gov. |
| license | BSD-3-Clause |
Lawrencium HPC documentation
This skill lets you answer Lawrencium/Einsteinium HPC questions from the
Markdown source of the Science IT documentation. The source of truth is the
lbnl-science-it/scienceit-docs
GitHub repository — the pages live under docs/ (mainly docs/hpc/, plus the Globus
data-transfer pages under docs/data/) on the main branch and are fetched directly,
so this skill works without a local checkout of the repo.
How to use this skill
- Fetch the raw Markdown, not the website. Every page on
https://scienceit-docs.lbl.gov is generated from a .md file under docs/hpc/
in the GitHub repo. Fetch the raw Markdown (see Fetching a page below) with
your web-fetch tool. Do not scrape the rendered HTML — the Markdown is the
source of truth, is cheaper to read, and has no navigation/markup noise.
- Route with the doc map below. Match the user's question to one or more pages
in the map, then fetch those files. Start narrow (the most specific page) and
widen only if needed.
- Ground every answer in what you read. If a topic is not covered by these
files, say so plainly rather than guessing. Never invent partition names, module
names, file paths, commands, email addresses, or URLs.
- Note on MkDocs syntax. The Markdown uses some MkDocs Material extensions —
??? question / !!! note admonitions, === "Tab" content tabs, {{ ext }}
macros, and {: ...} attribute lists. Treat these as formatting wrappers; read
through them to the underlying content.
Fetching a page
The doc map below lists each page by its repository path (e.g. docs/hpc/faqs.md).
To fetch the raw Markdown, prepend the raw base URL for the main branch:
https://raw.githubusercontent.com/lbnl-science-it/scienceit-docs/main/
Citing sources
When an answer draws on a page, cite it with its public site URL (not the raw
GitHub URL). Convert the repository path by removing the leading docs/, dropping
the .md extension, adding a trailing /, and prefixing
https://scienceit-docs.lbl.gov/:
Cite the public site URL, not the raw file path or the GitHub URL.
Scope
In scope: Lawrencium (LRC), Einsteinium, and the supported research clusters on the
LBNL Supercluster — accounts, access, Slurm, software, data transfer (the lrc-xfer DTN
and Globus), and Open OnDemand. Globus is covered because it is the primary way LBNL
researchers move data to and from Lawrencium; the connector pages for cloud storage
(Google Drive, AWS S3, Google Cloud Storage) are included as part of Globus. Out of
scope for this skill: cloud compute services (AWS/GCP accounts and VMs) and general
(non-LBNL) computing. Decline or redirect out-of-scope questions.
Doc map
Paths are relative to the repository root. Titles mirror the site navigation.
Overview & getting started
- Lawrencium Overview —
docs/hpc/index.md — What Lawrencium is: the LBNL Condo Cluster Computing (LC3) platform, part of the LBNL Supercluster.
- Live Cluster Status —
docs/hpc/status.md — Current cluster / queue status.
- Getting Started —
docs/hpc/getting-started.md — The primary ways to obtain access to Lawrencium.
- Frequently Asked Questions —
docs/hpc/faqs.md — Common questions: getting an account, submitting a first job, transferring data, max walltime, and more.
- Acknowledgement —
docs/hpc/acknowledgement.md — How to acknowledge Lawrencium in publications.
Account management
- Project Accounts —
docs/hpc/accounts/project-accounts.md — Eligibility and the PI-sponsored project/condo model.
- User Accounts —
docs/hpc/accounts/user-accounts.md — Requesting an account via the MyLRC portal and OTP token setup.
- Logging in —
docs/hpc/accounts/loggingin.md — SSH login to lrc-login.lbl.gov with PIN+OTP; login-node etiquette.
- Multi-Factor Authentication —
docs/hpc/accounts/mfa.md — Setting up and managing OTP/MFA tokens.
Computing systems
- CPU Cluster (Lawrencium) —
docs/hpc/systems/lawrencium.md — The general-purpose Lawrencium cluster: hardware and partitions.
- GPU Cluster (Einsteinium) —
docs/hpc/systems/einsteinium.md — The institutional GPU cluster for ML/DL workloads (es0/es1/es2 partitions).
- Supported research clusters — condo/PI clusters that share the Supercluster infrastructure:
- ALSACC —
docs/hpc/systems/supported/alsacc.md
- CATAMOUNT —
docs/hpc/systems/supported/catamount.md
- CATSCAN —
docs/hpc/systems/supported/catscan.md
- DIRAC1 —
docs/hpc/systems/supported/dirac1.md
- ETNA —
docs/hpc/systems/supported/etna.md
- MHG —
docs/hpc/systems/supported/mhg.md
Data transfer
- Using the lrc-xfer DTN —
docs/hpc/data-transfer-node.md — The dedicated data transfer node for moving data to/from the cluster.
- Globus —
docs/data/globus.md — What Globus is, LBL's Globus UI at globus.lbl.gov, the managed endpoints (Lawrencium, Google Drive, AWS S3, Google Cloud Storage), and setting up a Globus Connect Personal endpoint.
- Globus for Lawrencium —
docs/data/globus-instructions.md — Logging in to globus.lbl.gov, finding the lbnl#lrc endpoint, and transferring files to/from Lawrencium.
- Globus for Google Drive —
docs/data/globus-google-drive.md — Using the LBNL Gdrive Access endpoint and creating a guest collection for a Google Drive path.
- Globus AWS S3 Connector —
docs/data/globus-aws-s3-connector.md — Configuring the LBNL AWS S3 collection with an IAM access key to transfer to/from an S3 bucket.
- Globus Google Cloud Storage Connector —
docs/data/globus-google-cloud-storage-connector.md — Configuring the LBNL Google Cloud Storage collection with your LBL credentials to transfer to/from a GCS bucket.
Running jobs
- Slurm Overview —
docs/hpc/running/slurm-overview.md — Slurm resource manager: TRES, associations, partitions, QoS, submitting jobs.
- Example Scripts —
docs/hpc/running/script-examples.md — Job scripts for common cases: multi-core, GPU, low-priority condo, and long-running jobs.
- Monitor Jobs —
docs/hpc/running/monitor-jobs.md — Checking running jobs by Slurm job ID and inspecting usage.
- GNU Parallel —
docs/hpc/running/gnu-parallel.md — Grouping many (often serial) tasks into one Slurm submission across a node's cores.
Software — general
- Software Module Farm —
docs/hpc/software/software-module-farm.md — The SMF software collection and its usage on Lawrencium.
- Module Management —
docs/hpc/software/module-management.md — Using the Lmod environment-module system (module load, module avail, ...).
Software — compilers
- GCC —
docs/hpc/software/compilers/gcc.md — Available gcc versions and how to load them.
- Intel —
docs/hpc/software/compilers/intel.md — The LLVM-based Intel oneAPI compilers (icx, icpx, ifx).
- NVHPC —
docs/hpc/software/compilers/nvhpc.md — The NVIDIA HPC SDK.
Software — languages
- Python —
docs/hpc/software/languages/python.md — System Python and environment/module options.
- Julia —
docs/hpc/software/languages/julia.md — Julia as a module.
- R —
docs/hpc/software/languages/R.md — The R module.
Software — machine learning
- PyTorch —
docs/hpc/software/ml/pytorch.md — The ml/pytorch module.
- Ray —
docs/hpc/software/ml/ray.md — Distributed/parallel Python execution across nodes.
- TensorFlow —
docs/hpc/software/ml/tensorflow.md — The ml/tensorflow module.
- AlphaFold3 —
docs/hpc/software/ml/alphafold3.md — AlphaFold 3 model and database on Lawrencium.
Software — applications
- VASP —
docs/hpc/software/applications/vasp.md — Licensed VASP 6.4.1 and the access-request process.
Software — MPI
- Open MPI —
docs/hpc/software/mpi/openmpi.md — Loading openmpi under a compiler.
- Intel MPI —
docs/hpc/software/mpi/intelmpi.md — Loading Intel MPI under the Intel oneAPI compilers.
Software — libraries
- FFTW —
docs/hpc/software/libraries/fftw.md — FFTW under an MPI library.
- HDF5 —
docs/hpc/software/libraries/hdf5.md — HDF5 under a compiler + MPI.
- NetCDF —
docs/hpc/software/libraries/netcdf.md — NetCDF under a compiler + MPI.
- MKL —
docs/hpc/software/libraries/mkl.md — Intel MKL under gcc or Intel compilers.
- CUDA —
docs/hpc/software/libraries/cuda.md — CUDA Toolkit under the gcc tree.
Software — profiling
- GNU Gprof —
docs/hpc/software/performance/gnu-gprof.md — Profiling C/C++/Fortran with gprof.
- Intel VTune —
docs/hpc/software/performance/intel-vtune.md — Finding hotspots with Intel VTune Profiler.
Software — LLMs
- Ollama —
docs/hpc/software/llms/ollama.md — Running open-weight LLMs via the ollama module.
- vLLM —
docs/hpc/software/llms/vllm.md — Serving open-weight LLMs via the vllm module.
Open OnDemand
- Open OnDemand Overview —
docs/hpc/openondemand/overview.md — Browser-based interactive apps at lrc-ondemand.lbl.gov.
- Jupyter Server —
docs/hpc/openondemand/jupyter-server.md — Running Jupyter notebooks on Lawrencium compute.
- Ollama with Jupyter and VS Code —
docs/hpc/openondemand/ollama-jupyter-vscode.md — The Ollama + JupyterAI + VS Code Continue app for local LLMs.
- Adding Packages and Kernels —
docs/hpc/openondemand/packages-kernels.md — Managing Python packages and Jupyter kernels (miniforge3).