Context: I have below configuration
A single Managed Instance Group creating VMs using Instance Template. A single TCP/UDP load balancer that refers to the above Managed Instance Group. (A single backend)
In some examples, I've seen similar configurations where the TCP/UDP load balancer refers to multiple backends (Multiple Instance Group 1, 2, 3, etc). However, Instance Groups already have pretty good regional availability since you can spin them up in different zones.
Question: I wonder if a TCP/UDP load balancer only references one Managed Instance Group, does it load balance by the VM level or it becomes entirely useless to have a load balancer at all because there's only one Instance Group. I am not sure if the health check FOR THE LOAD BALANCER is that smart that can remember which VM was unhealthy and stop distributing traffic out of the other VMs in a single backend.
References:
https://cloud.google.com/load-balancing/docs/internal
https://cloud.google.com/load-balancing/docs/internal#traffic_distribution
