When I started working with microservices, I took the common rule of “two services must not share a data source” a bit too literally. I saw it sta

Sharing Data Between Microservices

submited by

Style Pass

2024-05-06 11:30:09

When I started working with microservices, I took the common rule of “two services must not share a data source” a bit too literally.

I saw it stapled everywhere on the internet: “thou shalt not share a DB between two services”, and it definitely made sense. A service must own its data and retain the freedom to change its schema as it pleases, without changing its external-facing API.

But there’s an important subtlety here that I didn’t understand until much later. To apply this rule properly, we have to distinguish between sharing a data source and sharing data.

An example: the Products service should own the products table, and all the records in it. They expose this data to other teams via an API, a products GraphQL query, and the creation of these records via a createProduct mutation.

The Products service has ownership over the products' source of truth, and no other team should reach into this directly, ever. If they want data out of it, they should ask the Products service for it via the contract (API) they adhere to. Under no circumstance should you allow direct access to the database, or you’ll lose the freedom to make changes to your schema. I learned this the hard way.