Lab Server Guide

Operational guide for researchers using the lab's Ubuntu compute server over SSH or remote desktop.

Important

This repository documents a real research system. Do not commit passwords, private keys, tokens, unpublished data, participant data, or administrator credentials.

Server at a glance

Item	Value
SSH endpoint	142.244.63.221:22
Operating system	Ubuntu 22.04 LTS
GPU	NVIDIA RTX A6000, 48 GB
Private working storage	/home/USERNAME
Large data storage	/media/Data
Remote desktop	xrdp over an SSH tunnel only
Workload scheduler	None; coordinate GPU-intensive work with other users

SSH host-key fingerprint:

ED25519 SHA256:kiCGXITxBmFlUtyAEsxdLH9jSSSBa1McELyjoq/Thd0

If the fingerprint changes unexpectedly, stop and contact the lab administrator before accepting it.

Use this repository

The repository is public so lab members can access the guide without repository invitations. Server access still requires an individually assigned account:

git clone https://github.com/ValleyC/lab-server-guide.git
cd lab-server-guide

The repository is organized by workflow: detailed material is under docs/, reusable client helpers are under scripts/, and maintenance requests use the GitHub issue template. If the repository is cloned on the server, run its read-only session check with:

bash scripts/check-session.sh

Quick start

SSH

ssh USERNAME@142.244.63.221

Confirm the session:

whoami
hostname

For SSH keys, client configuration, and VS Code Remote SSH, see SSH access.

Remote desktop

RDP port 3389 is not exposed directly. Start an SSH tunnel, then connect the RDP client to 127.0.0.1:13389.

Windows users can download both files in scripts/windows and run connect-lab-rdp.cmd. macOS and Linux users should follow Remote desktop.

Storage

Use /home/USERNAME for source code, environments, credentials, and private work. Use /media/Data for large datasets and other capacity-heavy files. The Data volume is shared infrastructure; privacy depends on Linux permissions, not folder names.

See Storage and permissions before moving data.

Guide map

Task	Guide
SSH, keys, host aliases, VS Code	SSH access
Graphical desktop over an encrypted tunnel	Remote desktop
/home, /media/Data, privacy, permissions	Storage and permissions
Conda, Python, local software installation	Software environments
GPU inspection and long-running jobs	GPU and jobs
scp, rsync, SFTP, VS Code transfer	File transfer
Command reference	Common commands
Known failure modes	Troubleshooting
Shared-system expectations	Usage policy
Resolved issues and pending maintenance	Maintenance status

Non-negotiable practices

• Use only your assigned account. Never share user or administrator credentials.
• Do not use the administrator account for research work.
• Do not reboot the server, install kernel/GPU drivers, edit system mounts, or change network configuration.
• Check GPU ownership before starting a memory-intensive process. There is no scheduler.
• Keep reproducible environment specifications with each project.
• Do not assume that data under /media/Data is private.
• Report security concerns privately rather than in a GitHub issue.

Contributing

Corrections and additions are welcome. See CONTRIBUTING.md. Keep examples generic and redact infrastructure secrets and research data.