PHE Swaps Compute Engine

A service that will pre-calculate swaps for all the products PHE have, and stick the results into a MySQL table to act like a cache. This service needs to run on an instance with as many CPU's as posssible in order to calculate the result's as fast as possible. It should automatically shut down as soon as the results have been calculated.

Expected Environment Variables

This is the full list of environment variables that need to be passed to the docker container for it to run:

The first set of environment variables we need is the connection details for the ETL database.

• ETL_DB_HOST - The hostname or IP address of the ETL database e.g. etl-rds.mydomain.com
• ETL_DB_USER - The user to connect to the database with. e.g. read-only-user
• ETL_DB_PASSWORD - The password to connect to the database with: e.g. myStrongRandomPassword
• ETL_DB_NAME - The name of the ETL database. e.g. food_consolidated
• ETL_DB_PORT - The port to connect with: e.g. 3306 (default mysql port)
• ETL_TABLE_NAME - The name of the ETL database table to build swaps based off. E.g. food_consolidated

The second set of environment variables we need are the connection details for the swaps database where we will put the cached results.

• SWAPS_CACHE_DB_HOST - The hostname or IP address of the database e.g. swaps-rds.mydomain.com
• SWAPS_CACHE_DB_USER - The username to connect to the database with (should have full access). e.g. user123
• SWAPS_CACHE_DB_PASSWORD - The password to connect to the database with: e.g. myStrongRandomPassword
• SWAPS_CACHE_DB_NAME - The name of the database that will hold the swaps cache. E.g. swaps
• SWAPS_CACHE_DB_PORT - The port to connect with: e.g. 3306 (default mysql port)

Image Compression Variables

There are also a set of environment variables necessary for the image compression routine:

• COMPRESSED_IMAGE_QUALITY - the quality level of the compressed images. This should be a number between 1-100. E.g. 60
• S3_IMAGE_BUCKET_KEY - the AWS credential key for accessing the S3 bucket that contains the images
• S3_IMAGE_BUCKET_SECRET - the AWS credential secret for accessing the S3 bucket that contains the images
• S3_IMAGE_BUCKET_NAME - the name of the bucket that contains the images.
• S3_IMAGE_BUCKET_PATH - the path to the folder within the bucket that contains the images. The code will handle if you start with a "/" or not.

Machine Learning Algorithm

One also needs to specify which machine-learning algorithm to use for calculating a category information which is then used for calculating swaps.

• ML_ALGORITHM - needs to be one of "spacy", "sklearn", or "both".

both is here for dev purposes. It will have us generate the categories for both so we can send back the information for picking which algorithm to use in future. In such a scenario, spacy will be the one that is then used when calculating swaps as only one algorithm can be used at a time.

Example Build and Deployment Commands

When you have checked out the codebase, you should be able to build with:

docker build . --tag swaps-compute-engine

If running manually, here is an example deployment command:

docker run -it \
    --restart=no \
    -e ETL_DB_HOST=etl-database.mydomain.com \
    -e ETL_DB_USER=foodUser \
    -e ETL_DB_PASSWORD=myStrongPassword \
    -e ETL_DB_NAME=food_consolidated \
    -e ETL_DB_PORT=3306 \
    -e ETL_TABLE_NAME=food_consolidated \
    -e SWAPS_CACHE_DB_HOST="swaps-database.mydomain.com" \
    -e SWAPS_CACHE_DB_USER=swapsUser \
    -e SWAPS_CACHE_DB_PASSWORD="myOtherAmazingPassword" \
    -e SWAPS_CACHE_DB_NAME="swaps" \
    -e SWAPS_CACHE_DB_PORT=3306 \
    -e COMPRESSED_IMAGE_QUALITY=60 \
    -e S3_IMAGE_BUCKET_KEY="AKIAFAKEBUCKETKEYDD2" \
    -e S3_IMAGE_BUCKET_SECRET="thisIsAFakeBucketSecretdfe2434dgernZW3Wx" \
    -e S3_IMAGE_BUCKET_NAME="my-bucket-of-images" \
    -e S3_IMAGE_BUCKET_PATH="/" \
    -e ML_ALGORITHM="spacy" \
    swaps-compute-engine

Alternatively, you can put the environment variables into an env file, and specify it with --env-file.

Database Requirements

To improve performance, the code in this project is likely t send very large queries to the database.

• The max_allowed_packet_size MySQL configuration variable needs to be increased to above its default of 64MiB.
• The database should have at least 2 GiB of RAM.

Crontab

To achieve the desired effect of starting up, running the calculations, and shutting down, install the crons.conf file on the server hosting the docker container.

Debugging

If you are debugging the categorizer module, any changes you make wont apply immediately, one has to change the script inside the Docker container and then run:

python3 /root/phe-swaps-module/phe_recommender setup.py install