DEEP-Hybrid-DataCloud training dashboard

⚠️ Warning: This is a fork from the INDIGO PaaS Orchestrator - Simple Graphical UI that is being customized to accomodate ML/DL workloads over the DEEP services. This is still work in progress. A preliminary version is deployed here.

Functionality

IAM authentication
Display user's deployments
Display deployment details and template
Delete deployment
Create new deployment

The deep-training-dashboard is a Python application built with the Flask microframework; Flask-Dance is used for Openid-Connect/OAuth2 integration.

The docker image uses Gunicorn as WSGI HTTP server to serve the Flask Application.

How to deploy the dashboard

Register a client in DEEP-IAM with the following properties:
- redirect uri: https://<DASHBOARD_HOST>:<PORT>/login/iam/authorized.
- scopes: openid, email, profile, offline_access.
- introspection endpoint enabled.
Clone the tosca-templates repository to get a set of tosca templates that the dashboard will load, e.g.:

git clone https://github.com/indigo-dc/tosca-templates

Create a config.json file in /app (see the example) an replace the values with your IAM_CLIENT_ID, IAM_CLIENT_SECRET and TOSCA_TEMPLATES_DIR. If you want that the reload requests (to update Tocas and modules list) from Github to be authenticated (so to ensure that they only come from your Github webhooks) you have to set GITHUB_SECRET to be the same as Github's webhook secret (see "Keeping the Dashboard updated" below).

{
    "IAM_CLIENT_ID": "my_client_id",
    "IAM_CLIENT_SECRET": "my_client_secret",
    "IAM_BASE_URL": "https://iam.deep-hybrid-datacloud.eu",

    "ORCHESTRATOR_URL": "https://paas.cloud.cnaf.infn.it/orchestrator",
    "SLAM_URL": "https://paas.cloud.cnaf.infn.it:8443",
    "CMDB_URL": "http://paas.cloud.cnaf.infn.it/cmdb",
    "IM_URL": "https://paas.cloud.cnaf.infn.it/im",
    "MONITORING_URL": "https://deep-paas.cloud.ba.infn.it/monitoring-wrapper",

    "TOSCA_TEMPLATES_DIR": "../tosca-templates/deep-oc",
    "COMMON_TOSCAS": {
        "default": "deep-oc-marathon-webdav.yml",
        "minimal": "deep-oc-marathon-minimal.yml"
    },
    "MODULES_YML": "https://raw.githubusercontent.com/deephdc/deep-oc/master/MODULES.yml",
    "GITHUB_SECRET": "",

    "SUPPORT_EMAIL": "deep-support@listas.csic.es",

    "EXTERNAL_LINKS": [
        {
            "url": "https://marketplace.deep-hybrid-datacloud.eu",
            "menu_item_name": "DEEP Marketplace"
        },
        {
            "url": "https://docs.deep-hybrid-datacloud.eu",
            "menu_item_name": "Documentation"
        },
        {
            "url": "https://deep-hybrid-datacloud.eu",
            "menu_item_name": "DEEP-Hybrid-DataCloud project page"
        }
    ],

    "LOG_LEVEL": "info",
    "ENABLE_ADVANCED_MENU": "yes"
}

Enable HTTPS

You need to run the deep-training-dashboard on HTTPS (otherwise you will get an error); you can choose between
- enabling the HTTPS support
- using an HTTPS proxy
Details are provided in the next paragraphs.

Enabling HTTPS

You would need to provide

a pair certificate/key that the container will read from the container paths /certs/cert.pem and /certs/key.pem;
the environment variable ENABLE_HTTPS set to True

Run the docker container:

docker run -d -p 443:5001 --name='deep-training-dashboard' \
           -e ENABLE_HTTPS=True \
           -v $PWD/cert.pem:/certs/cert.pem \
           -v $PWD/key.pem:/certs/key.pem \
           -v $PWD/config.json:/app/app/config.json \
           -v $PWD/tosca-templates:/opt/tosca-templates \
           indigodatacloud/deep-training-dashboard:latest

Access the dashboard at https://<DASHBOARD_HOST>/

Using an HTTPS Proxy

Example of configuration for nginx:

server {
      listen         80;
      server_name    YOUR_SERVER_NAME;
      return         301 https://$server_name$request_uri;
}

server {
    listen        443 ssl;
    server_name   YOUR_SERVER_NAME;
    access_log    /var/log/nginx/proxy-paas.access.log  combined;

    ssl on;
    ssl_protocols TLSv1 TLSv1.1 TLSv1.2;
    ssl_certificate           /etc/nginx/cert.pem;
    ssl_certificate_key       /etc/nginx/key.pem;
    ssl_trusted_certificate   /etc/nginx/trusted_ca_cert.pem;

    location / {
        # Pass the request to Gunicorn
        proxy_pass http://127.0.0.1:5001/;

        proxy_set_header        X-Real-IP $remote_addr;
        proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header        X-Forwarded-Proto https;
        proxy_set_header        Host $http_host;
        proxy_redirect          http:// https://;
        proxy_buffering         off;
    }
}

Run the docker container:

docker run -d -p 5001:5001 --name='deep-training-dashboard' \
           -v $PWD/config.json:/app/app/config.json \
           -v $PWD/tosca-templates:/opt/tosca-templates \
           indigodatacloud/deep-training-dashboard:latest

⚠️ Remember to update the redirect uri in the IAM client to https://<PROXY_HOST>/login/iam/authorized

Access the dashboard at https://<PROXY_HOST>/

Keeping the Dashboard updated

If you want the Dashboard to keep updated with the changes in the TOSCA repos or the modules list you will have to configure a Github webhook in those repos (for example [1] and [2]) so that any pushes in those repos trigger an update in the Dashboard.

The webhooks have to be configured as following:

Payload URL: <dashboard_url>/reload
Content type: application/json
Secret: Has to be the same as GITHUB_SECRET in the config.
Enable SSL is you are running over HTTPS and have valid certificates.
Just the push events.
Mark as Active.

Repo examples:

Performance tuning

You can change the number of gunicorn worker processes using the environment variable WORKERS. E.g. if you want to use 2 workers, launch the container with the option -e WORKERS=2 Check the documentation for ideas on tuning this parameter.

How to build and run the docker image

git clone https://github.com/indigo-dc/deep-training-dashboard.git
cd deep-training-dashboard
docker build -f docker/Dockerfile -t deep-training-dashboard .

To run the created image you have to export the config.json file (with your credentials) inside the docker container:

docker run -d -p 5001:5001 -v $PWD/config.json:/app/app/config.json deep-training-dashboard

The dashboard will be accessible at http://0.0.0.0:5001 . You can also choose to run image hosted on DockerHub:

docker run -d -p 5001:5001 -v $PWD/config.json:/app/app/config.json indigodatacloud/deep-training-dashboard

How to setup a development environment

git clone https://github.com/indigo-dc/deep-training-dashboard.git
cd deep-training-dashboard
python3 -m venv venv
source venv/bin/activate
pip3 install -r requirements.txt

Start the dashboard app with Flask:

FLASK_app=orchdashboard flask run --host=0.0.0.0 --cert cert.pem --key privkey.pem --port 443

or with Gunicorn:

gunicorn --certfile=cert.pem --keyfile=key.pem --bind 0.0.0.0:443 orchdashboard:app --daemon

Troubleshooting

SSL Cert Verification

If you see problems with the SLAM interaction, you would need to specify the certificate to be used to verify the SSL connection. You can pass the path to a CA_BUNDLE file or directory with certificates of trusted CAs setting the parameter SLAM_CERT in the config.json file:

{
  ...
  "SLAM_URL": "https://indigo-slam.cloud.ba.infn.it:8443",
  "SLAM_CERT": "/path/to/certfile"
}

If you are running the docker container, you need to ensure that the cert file is available inside the container in the path set in the SLAM_CERT parameter, i.e. you would use a bind mount (-v $PWD/certfile:/path/to/cerfile)

References:

https://2.python-requests.org/en/master/user/advanced/#ssl-cert-verification

indigo-dc / deep-training-dashboard