[v2] Different databases.yaml based on the provided endpoint

[miran248]

Hey,

Is it possible to have different databases.yaml configs based on the endpoint?
Prod config might have read_replicas which are absent on dev envs.
Currently i have to comment out the field before applying the changes on dev.

Should i create two separate configs using hasura init and then !include the tables, functions and other bits so both configs would use the same definitions?

Thanks,
Miran

[manimovassagh]

Good question. Hasura v2 does not have built-in support for conditional metadata based on the endpoint or environment, but there are a few battle-tested approaches:

1. Environment variables in metadata

You can use env vars in databases.yaml for the connection string itself, but unfortunately read_replicas as a block cannot be conditionally included/excluded via env vars.

2. Separate metadata directories (recommended)

The approach you mentioned is the way most teams handle this. Use hasura init to create two project directories (e.g. hasura-prod and hasura-dev), and share common table/relationship/permission definitions between them via symlinks or the !include tag. Then apply with hasura metadata apply --project hasura-dev or hasura-prod.

3. Script-based approach

Some teams keep a single metadata directory and use a small script that generates the final databases.yaml from a template before running hasura metadata apply. You keep databases.prod.yaml and databases.dev.yaml side by side, and a shell script copies the right one based on an ENVIRONMENT variable before applying metadata.

This keeps your CI/CD simple and avoids the manual commenting/uncommenting.

Option 2 is the most maintainable long-term since it keeps everything declarative.