This flow example has two main components,
https://funcx.readthedocs.io/en/latest/quickstart.html
ssh gl-login1.arc-ts.umich.edu
module load python/3.10.4
virtualenv
source bin/activate
python3 -m pip install globus-compute-endpoint
python3 -m pip install globus-automate-client
Need to start this on same host always. An endpoint per slurm configuration is required.
globus-compute-endpoint configure gl-login1-standard
globus-compute-endpoint list
# copy in example parsl config to use slurm
# this can be skipped and test will run on the login node to veryify
cp config.py ~/.globus_compute/gl-login1-standard/config.py
# edit config.py and update account, slurm settings
# example defaults to using a whole 36 core 180gb node
globus-compute-endpoint start gl-login1-standard
globus-compute-endpoint list
#should list running
# save UUID
python3 reg_func.py
# save UUID
globus-automate flow deploy --title "ARC Example Flow" \
--definition flow-def.json --input-schema input-schema.json
Edit input.json
for your own path and server. Default values are for the
web UI only and must be provided when calling on the CLI. The example
input.json
has all the required variables.
globus-automate flow run <uuid> --flow-input
input.json --label test-cli1