Profiles to Subscription Receipts

Speed up entitlement lookups by having a simple API to see what products someone bought (receipts)

epe-receipts.conde.io/graphql

Wiki home: https://cnissues.atlassian.net/wiki/spaces/CMSG/pages/327942320/Entitlement+Receipts+Service

Requirements

• Be aware of rate limiting - See the RateLimit-* headers to check your quota. In the GraphQL Playground, you can set auth by adding something like this to HTTP HEADERS:

  {
  "Authorization": "Basic YW5kaTp0ZXN0"
  }
  

• Use HTTP Basic Authorization - For now, you can specify any password, but a specific username/password will be required in the future. For example, Segmenter-Entitlements uses https://segent:hunter2@epe-receipts.conde.io/graphql

Best practices

• Set a User-Agent - Again, this helps us understand who is calling, but a User-Agent can have more information, like version info. Between HTTP Basic Auth and User-Agent, we can slice traffic by application, and application version. In the future, requests without a User-Agent may be severely rate limited
• Name your queries - This adds context to your queries so if there's a problem, it's easier to debug. Also, it's a GraphQL best practice
• Avoid "refresh" - This is an extremely costly query, and puts pressure on a lot of systems. In the future, applications that refresh too often will get rate limited

Glossary

Definitions are stored as living documentation in the GraphQL schema to avoid duplicating work.

GraphQL

Basic entitlement query:

mutation ExampleFetch {
    fetchProfile(input: { profile: { amgUuid: "24851284-1fae-4818-b92e-56da11555c6f" } }) {
        profile {
            viewEntitlements {
                product
                starts
                expires
                __typename
            }
        }
    }
}

Only return CDS results

mutation ExampleOnlyCDS {
    fetchProfile(
        input: {
            profile: { amgUuid: "24851284-1fae-4818-b92e-56da11555c6f" }
            refreshPolicy: "refresh"
        }
    ) {
        didRefresh
        didRefreshReason
        profile {
            viewEntitlements {
                ... on CdsViewEntitlements {
                    product
                    starts
                    expires
                }
            }
        }
    }
}

View what we know about a profile

query ExampleGetProfileByEmail {
    profile(email: "active@tt.com") {
        amgUuid
        receipts {
            providerId
            v
            timestamp
        }
        viewEntitlements {
            product
            starts
            expires
        }
    }
}

Force a refresh of a profile

mutation ExampleRefresh {
    fetchProfile(
        input: {
            profile: { amgUuid: "24851284-1fae-4818-b92e-56da11555c6f" }
            refreshPolicy: "refresh"
        }
    ) {
        profile {
            amgUuid
            receipts {
                providerId
                timestamp
            }
            viewEntitlements {
                product
                starts
                expires
            }
        }
    }
}

DynamoDB schema

entitlements-receipts

The original data from providers is stored in entitlements-receipts with amgUuid as the partition key and the providerId as the sort key. This is just a cache. All this data can get retrieved again.

CDS Global
---
amgUuid: '58383a82-eadc-4172-8685-ca800e7b9500'
providerId: 'cds.<accountNumber>.<magCode>'
data: <json body>
timestamp: 1531254336997

CDS UK
---
amgUuid: '58383a82-eadc-4172-8685-ca800e7b9500'
providerId: 'cdsuk.<URN>.<originator>'
data: <json body>
timestamp: 1531254336997

Apple Receipt
---
amgUuid: '58383a82-eadc-4172-8685-ca800e7b9500'
providerId: 'apple.<sha1 of base64 receipt data>'
data: <json verified receipt data>
timestamp: 1531254336997

Andros aka CNEE Legacy receipt
---
amgUuid: '58383a82-eadc-4172-8685-ca800e7b9500'
providerId: 'andros.<coupon code>'
data: <json body>
timestamp: 1531254336997

Using providerId as a sort key?

Instead of throwing everything into one document, we could use amgUuid as the partition key and provider id as a sort key with a query.

• GetItem is slightly faster and cheaper than Query
• Although we may have thousands of possible products and providers, users won't hit the 400kb limit 🤞

Extra DynamoDB fields

There are some special fields you may see:

• v: The current schema version (set by SCHEMA_VERSION in config.js)
• _b: This item was touched by the ElasticSearch backfill process

Special cases

• CDS status "D" donors. They bought a subscription, but as a gift. We pretend these don't exist

Development instruction

Prequisites

• nodemon available globally, npm i -g nodemon
• You must have AWS CLI access and an active AWS_SESSION_TOKEN. See: https://cnissues.atlassian.net/wiki/spaces/DEVOPS/pages/6717524/AWS+CLI+Access
• A variety of environment variables set. I need to do a better job of documenting this. For now, run this at the top level: VAULT_TOKEN=$(cat ~/.vault-token) VAULT_SECRET_PATH=secret/colab/entitlements/staging/app node index.js

Starting the app

npm install
lerna bootstrap
npm run dev