CPI Viewer Backend

A data visualization backend that uses saspy python module inside of flask to connect to SAS OnDemand server and retrieve data from CPI government files.

1. Set up free SAS OnDemand account and configure things

• https://welcome.oda.sas.com
• In SAS OnDemand, create a new directory cpi_viewer/data. Upload the files in cpi_data.zip to this directly. Also, create a new direcotry cpi_viewer/sas7bdat_data and upload the files from sas7bdat.zip into there.
• In SAS OnDemand, copy the specific path to the series file used in macro1.txt. For me, it is /home/u63731510/cpi_viewer/data/cu.series. Set this directly in the proc import of macro1.txt.
• Similarly in macro2.txt, replace the two paths with your own specific ones that include your SAS OnDemand account id.

2. Create the Docker image for the flask app

• in cpi-viewer-container file, add a .env file with your SAS OnDemand username and password. It should look like this: OMRUSER=johndoe@gmail.com OMRPW=password
• Within the cpi-viewer-container directory, execute the docker build command: docker build -t <docker_account_name>/<image_name>:1.1 .

3. Run and test the flask app locally

• To run the image locally, execute the docker run command: docker run -p 8000:8000 <docker_account_name>/<image_name>:1.1
• Test the first route with this sample call: http://localhost:8000/getAvailableSeries?1=SAF11211
• Test the second route with this sample call: http://localhost:8000/makeGraphReadyData?1=1938%2C2020%2CQuarterly%2Ccu.data.11.USFoodBeverage%2CCUUR0000SAF11211

4. Deploy the flask app to the Azure cloud (works on free account)

• Once you have the docker image built, upload it to your docker public registery with this command: docker push <docker_account_name>/<image_name>:1.1
• In Azure, create a new "Container App".
• For ingress options, set it to "enable ingress", ingress type = http, target port = 8000
• Set environmental variable NUM_WORKERS = 4. This value can be set higher, but this is a safe number. As a rule of thumb, the number of processes that can be spawned efficiently is 2*number of cores + 2.
• Set SAS_REQ_TIMEOUT = 5.
• Set the 'Scale' setting to be between 1 and X instances so the container app dosen't shut down, if it is 0 it will power off in some time and will need a warmup start again.
• Bump up CPU to 1.0 and memory to 2.0 as the default is 0.5 and 1.0 for the container app