MANUAL_INSTALLATION.md

Manual Setup

Prerequisites

• Cloud Foundry CLI installed: https://console.ng.bluemix.net/docs/starters/install_cli.html
• Local python version must be 3.5.x. Older python versions will not work.
• Virtualenv installed
• Linux/OS X are recommended for your local development environment.

Setup the source code

git clone https://github.com/snowch/movie-recommender-demo
cd movie-recommender-demo/web_app
virtualenv venv --python=python3.5 
source venv/bin/activate
pip3.5 install -r requirements.txt

Configure Cloudant

• Create a Cloudant service in Bluemix.
• Create the file web_app/etc/cloudant_vcap.json (see web_app/etc/cloudant_vcap.json_template for an example)

Configure Message Hub (optional)

• Create a MessageHub service in Bluemix.
• Create the file web_app/etc/messagehub_vcap.json (see web_app/etc/messagehub_vcap.json_template for an example)
• Edit manifest.yml to uncomment the messagehub configuration
• When the web application gets deployed, a topic 'movie_ratings' will be created for you

Configure Redis (optional)

• Create a Redis service in Bluemix.
• Create the file web_app/etc/redis_vcap.json (see web_app/etc/redis_vcap.json_template for an example)
• Edit manifest.yml to uncomment the redis configuration

Configure BigInsights (optional)

• Create a BigInsights Basic service in Bluemix. Ensure you select Hive and Spark as optional components.
• Create the file hdp_app/etc/vcap.json with your Message Hub details (see hdp_app/etc/vcap.json_template for an example)
• Create the file hdp_app/etc/bi_connection.properties with your BigInsight details (see hdp_app/etc/bi_connection.properties_template for an example)
• Follow the steps in the README to build and deploy the spark code

Setup the Cloudant databases

This only needs to be performed once.

cd web_app
./run.sh db_all

If you want to use BigInsights:

cf set-env movie-recommend-demo BI_HIVE_USERNAME changeme
cf set-env movie-recommend-demo BI_HIVE_PASSWORD changeme
cf set-env movie-recommend-demo BI_HIVE_HOSTNAME changeme
cf restage movie-recommend-demo

Run web_app locally

cd web_app
./run.sh

You should see a URL output:

 * Running on http://0.0.0.0:5000/ (Press CTRL+C to quit)

Open this to try out the webapp locally.

Push web_app to bluemix

• edit manifest.yml
  • provide a unique host
  • change the name of the services to reflect
    • your cloudant service name
    • your redis service name
    • your message hub service name

Then run the following commands:

# change directory to the folder with manifest.yml

cd ..

# When you login, you will be prompted to select the space and org.
# ensure you choose the space/org where you created your Cloudant 
# service and Redis service 

cf login ...

# verify that cf is pointing at the right space/org

cf target

# you can now push the web app to Bluemix

cf push

To create recommendations

• create an account in the webapp
• rate some movies
• create a new DSX project, then
  • upload the [Step 06 - Install Spark Cloudant](../notebooks/Step 06 - Install Spark Cloudant.ipynb) notebook to DSX
  • run the Step 06 - Install Cloudant Python notebook
  • upload the [Step 07 - Cloudant Datastore Recommender](../notebooks/Step 07 - Cloudant Datastore Recommender.ipynb) notebook to DSX
  • follow the instructions in the notebook to setup the cloudant_credentials.json file
  • click: Cell -> run all
• when finished, navigate back to the web application and click on 'Get Recommendations'
• if recommendations were generated without error:
  • save a new version of the notebook: File -> Save Version
  • setup a DSX schedule job to run the version every hour