/training-2023-CSDMS

Primary LanguageJupyter NotebookCreative Commons Zero v1.0 UniversalCC0-1.0

Materials for 2023-CSDMS Clinic 1.2:

"Emerging open-source Python-based infrastructure for scalable, data-proximate analysis and visualization of model output (a.k.a. the Pangeo Framework)"

Start

  1. login to https://nebari-workshop.esipfed.org:

    username: <your email address>
    password: 123  (you will be prompted to change it)
    
  2. Start a "regular instance" server (the default). This launches JupyterLab.

  3. In the JupyterLab interface, open a terminal ("File"->"New"->"Terminal")

  4. In the terminal, type:

    cd /shared/users
    python start.py
  5. In the file browser, browse to the 'shared'->'users' and then your username folder.

  6. Browse to "training-2023-CSDMS"

  7. Click the "clinic_overview.ipynb" notebook to launch it.

Deployment notes

  1. Deployed Nebari 2023.5.1 on AWS using username/password for authentication

  2. In Keycloak, added a users group, with same permissions as analyst (allowing dask gateway cluster creation). To do this go to the users group, select Role Mapping, at the bottom of that page is a Client Roles drop down that you can use to add and remove roles for that group. Add dask-developer.

  3. We created a CSV list of users with columns: first_name, last_name, email

  4. Converted the CSV to a JSON file importable into keycloak using generate_keycloak.ipynb in this repo

  5. At the time of the May 13 workshop, destroying on AWS with nebari destroy didn't everything:

[terraform]: Destroy complete! Resources: 2 destroyed.
[terraform]: Stage=stages/07-kubernetes-services failed to fully destroy
Stage=stages/05-kubernetes-keycloak failed to fully destroy
Stage=stages/04-kubernetes-ingress failed to fully destroy
Stage=stages/03-kubernetes-initialize failed to fully destroy
Stage=stages/02-infrastructure failed to fully destroy
ERROR: not all nebari stages were destroyed properly. For cloud deployments of Nebari typically only stages 01 and 02 need to succeed to properly destroy everything

This AWS force destroy script from @iameskild should destroy everything, but be careful, as it destroys the current cluster, which may not be the cluster you indended if the nebari destroy already destroyed the targeted cluster!