Web UI with FirecREST

Goal

Create a Flask web application that interfaces HPC services at CSCS using the FirecREST API.

Edit src/client.py and replace the body of every function with the _with_f7t suffix using PyFirecREST (recommended) or the FirecREST API directly.

Prerequisites

• API key training credentials
• Docker or Podman installed

Configuration

1. Copy and edit the configuration file:

   cd use-case-UI/src
   cp config.py.orig config.py
   

2. Fill in your credentials in config.py:

   class DevConfig(Config):
       OIDC_CLIENT_ID = "<CLIENT_ID>"
       OIDC_CLIENT_SECRET = "<CLIENT_SECRET>"
       USER_GROUP = "<GROUP>"
       OIDC_AUTH_BASE_URL = "https://auth.cscs.ch"
       OIDC_AUTH_REALM = "firecrest-clients"
       FIRECREST_URL = "https://api.cscs.ch/hpc/firecrest/v2"
       SYSTEM_NAME = "daint"
       SYSTEM_RESERVATION = "<RESERVATION>"
   

Tip

For debugging, leave the default values for SBATCH_TEMPLATE, PROBLEM_INI_FILE, PROBLEM_MSH_FILE, and POST_TEMPLATE.

For a real test case, use:

PROBLEM_INI_FILE = 'inc-cylinder.ini'
PROBLEM_MSH_FILE = 'inc-cylinder.msh'
SBATCH_TEMPLATE = "cylinder.sh.tmpl"
POST_TEMPLATE = "post_proc.sh.tmpl"

This requires Sarus.

Build and Run

cd use-case-UI
make build
make run

Note

The Makefile uses the docker command. To use podman instead, replace docker with podman in the Makefile.

The application runs at http://localhost:9091. Logs are written to log/client.log:

tail -f log/client.log

Using the Application

1. Open http://localhost:9091 in your browser.
2. Set the simulation parameters (number of steps, nodes, job name, etc.).
3. Click Submit Jobs to run the simulation. New jobs appear in the list and result files in the working directory.
4. Click Start Postprocessing to launch a post-processing job and view the animated GIF result.