DefaultEndpointRouter results uneven autoscaling in mutiple endpoints scenario

[bjqian]

Describe the Issue

When utilizing two SignalR instances in multiple endpoints scenario:

• SignalR A: Autoscaling is enabled at 80% of quota usage
• SignalR B: Autoscaling is enabled at 80% of quota usage

An issue arises after SignalR A is scaled up: SignalR B is less likely to scale up. This occurs because the DefaultEndpointRouter uses a weighted routing method, which routes connections based on the remaining connection quota of each endpoint.

Expected Behavior

The DefaultEndpointRouter should ensure that instances scale more evenly when autoscaling is enabled.

[bjqian]

Normalize the DefaultEndpointRouter weight using unit:

azure-signalr/src/Microsoft.Azure.SignalR/EndpointRouters/DefaultEndpointRouter.cs

Line 53 in 6fa8775

private ServiceEndpoint GetEndpointAccordingToWeight(ServiceEndpoint[] availableEndpoints)