Re: [grpc-io] gRPC client fallback from load-balancer to round robin DNS endpoint on LB outage

Will take a look. Thank you Larry and Terry! Much appreciated! On Monday, November 6, 2023 at 4:10:09 PM UTC-8 Terry Wilson wrote: > Krishna, > > I'll mention that we also have a short guide on custom LBs > that can help you > with the

Re: [grpc-io] gRPC client fallback from load-balancer to round robin DNS endpoint on LB outage

Krishna, I'll mention that we also have a short guide on custom LBs that can help you with the fundamentals. On Monday, November 6, 2023 at 3:53:01 PM UTC-8 Larry Safran wrote: > Yes, you will need a custom NameResolver. > You could have

Re: [grpc-io] gRPC client fallback from load-balancer to round robin DNS endpoint on LB outage

Yes, you will need a custom NameResolver. You could have your NameResolver do one of the following 1) Extend DnsNameResolver (as is done by the internal GrpcNameResolver) and override the doResolve method 2) Directly extend NameResolver and have a field with a DnsNameResolver created through the

Re: [grpc-io] gRPC client fallback from load-balancer to round robin DNS endpoint on LB outage

Hey Larry, Thank you for your response! Are there any examples out in the wild on how to do this? Also the part that is confusing me is the name resolution. Do we have to add a custom name resolution policy as well so that we get both the LB and round robin IP addresses in

Re: [grpc-io] gRPC client fallback from load-balancer to round robin DNS endpoint on LB outage

This is a perfect scenario for a custom LB policy. Since LB policies form a tree, you could easily have your policy delegate to either a PickFirst or RoundRobin policy depending on whether or not the LB is up. Just have your policy define a ForwardingLoadbalanceHelper whose updateBalancingState

[grpc-io] gRPC client fallback from load-balancer to round robin DNS endpoint on LB outage

Hello All, I am working on a gRPC client that connects to a gRPC service fronted by an LB but would like the client to automatically fallback to using round-robin DNS to directly connect to the backend servers when the LB goes down. How do I go about this? Would I need to implement my own