Server Access and Paths

This page summarizes practical server access for ADAMACS users.

Public-doc note:

internal endpoint values are redacted here; request current IPs/hostnames via lab ops private channels.

Current server IPs

MAIN_SERVER: <MAIN_SERVER_IP>
- primary DataJoint DB host
- blob storage host
- share host for ingest uploads
- CPU worker host
GPU_SERVER: <GPU_SERVER_IP>
- GPU worker host for model-heavy pipelines
BACKUP_SERVER: <BACKUP_SERVER_IP>
- backup host (rsync pulls + borg)

Primary access modes

SSH

ssh <username>@<MAIN_SERVER_IP>

For GPU host:

ssh <username>@<GPU_SERVER_IP>

Detailed GPU usage policy and workflows:

Infrastructure -> GPU Server Usage

VS Code remote tunnel / remote SSH

Either:

Remote SSH directly to server hosts, or
VS Code tunnel workflows if configured by lab ops.

Example tunnel CLI flow on server:

curl -Lk 'https://code.visualstudio.com/sha/download?build=stable&os=cli-alpine-x64' --output vscode_cli.tar.gz
tar -xf vscode_cli.tar.gz
./code tunnel

Reference:

https://code.visualstudio.com/docs/remote/tunnels

SSH key setup (recommended)

Generate key pair locally:

ssh-keygen -t rsa -b 4096 -C "your_email@example.com"

Copy public key to server:

ssh-copy-id <username>@<MAIN_SERVER_IP>

Then test:

ssh <username>@<MAIN_SERVER_IP>

Standard upload path convention

Raw/consolidated data should live under:

/datajoint-data/data/<username>

Do not store raw datasets in home directories.

Jupyter on remote host

Start on server:

jupyter lab --no-browser --port=8080

Forward locally:

ssh -L 8080:localhost:8080 <username>@<MAIN_SERVER_IP>

Open local browser:

http://localhost:8080

Access troubleshooting quick checks

Ping host IP.
Validate VPN status if remote.
Check username/password or key registration.
Check home directory quotas and permissions.
Confirm your account has DB and share permissions.