Scaling Backstage Deployments

There are several different methods for scaling Backstage deployments. The most straight-forward one is to simply deploy multiple identical instances and distributing incoming requests across them. The one requirement for this to work is that all instances need to share the same external resources, such as the database, and optional caching or search services. The Backstage backend plugins will coordinate through the database to share state and coordinate work.

Another method for scaling Backstage deployments is to break apart the backend into multiple different services, each running a different set of plugins. This is a more advanced approach and requires you to be able to route requests to the appropriate backends based on the plugin ID. Both for ingress, but also internal traffic between Backstage backends, which is done by creating a custom implementation of the DiscoveryService interface. See the backend system docs for more details on how to separate your deployment into multiple backend instances.

Lastly, you can also replicate the Backstage deployments across multiple regions. This is not a pattern that there is built-in support for and typically only makes sense to do for individual backend plugins.