Miscellaneous

Getting Help

Tmux

Lighthouse and GreatLakes use multiple login nodes for load balancing/redundancy. To persist a session across login nodes, change where tmux creates its sockets:

# Goes in your ~/.bashrc
export TMUX_TMPDIR=~/.local/state/tmux
mkdir -p -m 700 $TMUX_TMPDIR

This won’t apply until you’ve killed tmux on all login nodes with tmux kill-server and loaded your shell (Log out and back in again)

Setting up Turbo / Scratch

# Create a folder on turbo
mkdir "/nfs/turbo/coe-venkvis/$(whoami)"

# Symlinks to make things nice
ln -s "/nfs/turbo/coe-venkvis/$(whoami)" ~/turbo
ln -s "/scratch/venkvis_root/venkvis/$(whoami)" ~/scratch

Globus

If you’re moving data between clusters, use Globus:

UM ARC Endpoints (don’t go using some rando endpoint)

Brock’s Guide to Data Archiving

Brock Palen created a great guide to data archiving, highly recommend for anyone with lots of data that needs to be saved somewhere:

  • Tools for finding files
  • Various compression tools (gzip, bzip2, xz, lz4) and their trade offs
  • Tar: What it is and why you must use it on DataDen
  • Archivetar - An all-in-one tool for archiving data and uploading it to DataDen
    • Parallel compression/tar tools built-in
    • Can module load it on GreatLakes or Artemis module load archivetar

OnDemand

Provided by ARC OnDemand let’s you interact with Artemis (Or Great Lakes) via a web browser to:

  • Launch a Jupyter Notebook (Even one with a GPU)
  • Launch a Virtual Remote Desktop (You can even connect to it with a native app)

To start visit: lighthouse.arc-ts.umich.edu

You’re still billed for the resources you use, it’s just a different way of accessing them