Skip to content

Latest commit

 

History

History
40 lines (27 loc) · 1.47 KB

README_TASKVINE_UNL.md

File metadata and controls

40 lines (27 loc) · 1.47 KB
Raw

Execute materialize_branches from coffea-casa dev and using the t3 at UNL

Conda environment with taskvine workers:

  1. In a terminal, ssh to swan.unl.edu
  2. From there, ssh to t3.unl.edu
  3. Create a conda environment to get the taskvine workers executable:
    1. conda create -c conda-forge -n coffea-casa-min python=3.10.12 ndcctools
  4. Optional, once: add basic coffea related packages to environment, not needed if using docker images at execution
    1. conda activate coffea-casa-min
    2. conda install -c conda-forge xrootd
    3. pip install fsspec-xrootd dask distributed coffea

Execute notebook from coffea-casa-dev

  1. In materialize_branches.ipynb notebook, change executor = "dask" to executor = "taskvine".
  2. In a terminal, ssh to swan.unl.edu
  3. From there, ssh to t3.unl.edu
  4. Get a condor token: mkdir -p ~/.condor/tokens.d && condor_token_fetch > ~/.condor/tokens.d/my_token
  5. Activate the environment: conda activate coffea-casa-min
  6. Launch the workers:
casa_user_id=btovar-40nd-2eedu    # get user id from a terminal at coffea-casa: h=$(hostname); echo ${h#jupyter-}

casa_host=$casa_user_id.dask-worker.cmsaf-dev.flatiron.hollandhpc.org

docker=hub.opensciencegrid.org/coffea-casa/cc-dask-alma8:2024.04.05

number_of_workers=1

cores=12

memory=$((cores*3000))

disk=$((cores*6000))

vine_submit_workers -Tcondor -t300 --ssl -E"--transfer-port 1088" --cores $cores --memory $memory --disk $disk --docker-universe $docker $casa_host 8786 $number_of_workers