Deployment Patterns

This article describes deployment patterns for retool-onpremise. One common deployment pattern is to have multiple instances for prod, staging and dev. Environments allow you to configure different endpoints for the same resource - for example, you can create one resource with three different endpoints for prod, staging and dev. App developers can then use the dev or staging resources when they are developing, and only switch to the prod resource once they have tested the new features.

One issue with this approach is the lack of any sort of barrier when developers are switching to prod resources. If a user has access to a resource, they have access to all environments of that resource. Ideally, app developers would have a guarantee that a new feature will not be deployed into production until it has been tested successfully.

Deploying multiple instances of Retool, in conjunction with protected apps, can solve this problem. Developers can add new features in the test instance, and test new feature branches on test and staging before merging the feature branch into the master branch. After the feature branch has been tested, and merged, the code can be promoted to the prod instance.

AWS Architecture

This architecture was deployed on AWS, using EC2 for the Retool deployments and Postgres RDS for resources.