equinox mit

Contains code that process mitgcm output for the EQUINOx project

• sandbox/ : first trials with xmitgcm, basic stuff

• launch/ : scripts required to launch jupyterlabs and dask clusters

• doc/ : conda and git info

• ...

---

install

All scripts require python librairies that may be installed with conda according to the following instructions here

---

run on datarmor:

After having installed all libraries, and cloned this repository, go into mit_equinox/datarmor.

method 1 (preferred):

./launch-jlab.sh

Follow instructions that pop up from there.

The spin up of dask relies on dask-jobqueue:

from dask_jobqueue import PBSCluster
cluster = PBSCluster()
w = cluster.start_workers(28*2)

from dask.distributed import Client
client = Client(cluster)

Kill jobs once done with computations in a notebook with:

cluster.close()

or in a shell with python kill.py.

Clean up after computations: ./clean.sh

method 2 :

For 8 nodes for example:

./launch-dask.sh 8
./launch-jlab.sh wait

Follow instructions that pop up from there

Once you are done computing, kill the relevant jobs.

Clean up after computations: ./clean.sh