I’m currently working on a university project and trying to explore the possibility of scaling a service in a way that the newly created replicas are accessible only by specific clients.
Why I want to do this approach, as opposed to the usual horizontal scaling, is to test whether it can help reduce unused resources from the replicas by making a new one with less resource than normal and to avoid disrupting other clients.
My current plan involves identifying clients through their HTTP headers and automatically, but I’m open to alternative suggestions. Some people have recommended using a service mesh (currently we are using linkerd), Flagger’s A/B testing, or nginx ingress configuration-snippets to achieve this. However, I’m unsure how to properly implement these solutions.
My final goal is to able group multiple “unusual” (as in exceeding normal request) to the newly created service.
Image bellow is for an illustration on what I want to achieve. Any technical guidance or additional suggestions would be greatly appreciated!
Also, pardon for my bad english