Skip to content

CLAUDE.md

Latest commit

 

History

History
82 lines (59 loc) · 3.09 KB

CLAUDE.md

File metadata and controls

82 lines (59 loc) · 3.09 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Overview

This repository contains a unified SLURM batch script for launching Positron (an IDE) on Alpine (CU Boulder) and amc-bodhi (CU Anschutz) HPC clusters. The script allocates compute resources, provides SSH connection instructions, and includes a one-time setup subcommand.

Architecture

The repository is intentionally minimal:

  • positron-remote.sh: Unified SLURM batch script supporting both Alpine and amc-bodhi clusters
    • Cluster-specific config (partition, memory, proxy host) via a case statement
    • setup subcommand automates SSH key exchange and scratch symlink creation
    • Cluster is selected via argument: ./positron-remote.sh alpine or ./positron-remote.sh bodhi
  • Uses a ProxyJump SSH configuration pattern to connect through the login node to the allocated compute node

Usage

Submit the job (the script self-submits to SLURM):

./positron-remote.sh alpine    # Alpine (default)
./positron-remote.sh bodhi     # amc-bodhi

One-time setup (run from local machine):

./positron-remote.sh setup alpine
./positron-remote.sh setup bodhi

Monitor job status:

squeue -u $USER

Cancel the job when done:

scancel <JOB_ID>

View connection info:

cat logs/positron-<JOB_ID>.out

Key Configuration Parameters

Cluster-specific SLURM settings are passed via sbatch CLI overrides:

  • --time: Maximum job duration (currently 8 hours)
  • --mem: Memory allocation (24gb Alpine, 20gb bodhi)
  • --cpus-per-task: Number of CPU cores (currently 4)
  • --partition: Cluster partition (amilan for Alpine, normal for bodhi)
  • --qos: Quality of service tier

These parameters should be adjusted based on computational requirements. Alpine documentation: https://curc.readthedocs.io/en/latest/compute/alpine.html

How It Works

When you run ./positron-remote.sh alpine:

  1. The script checks if it's running under SLURM (via the POSITRON_SLURM_EXEC environment variable)
  2. If not, it submits itself to SLURM using sbatch with cluster-specific overrides and exits
  3. The cluster name is passed to the SLURM execution via POSITRON_CLUSTER environment variable
  4. When SLURM runs the script on a compute node, it displays the SSH connection info in the log file

SSH Configuration Pattern

The script generates a temporary SSH config entry using:

  1. ProxyJump: Routes connection through the cluster's login node
  2. Dynamic hostname: Uses the allocated compute node's hostname
  3. Job-specific alias: positron-{alpine|bodhi}-${SLURM_JOB_ID} for easy identification
  4. ForwardAgent: Enables SSH agent forwarding for git operations on the compute node

This pattern enables Positron's Remote-SSH extension to connect directly to the compute node while respecting the cluster's security model.

Prerequisites

  • Positron or VS Code with Remote-SSH extension (recommended in .vscode/extensions.json)
  • SSH key configured for cluster access (use ./positron-remote.sh setup to automate)
  • Access to Alpine or amc-bodhi HPC cluster