After a configuration update, it is possible for the number of replicas to experience a large change (e.g. from 1k replicas => 2k), overwhelming the other parts of the infrastructure.
We should have a tunable scale up rate of change parameter such as "scale_up_replicas_per_second" that can control how fast we increase replica counts.
Note: it can default to infinity which is the policy today.
After a configuration update, it is possible for the number of replicas to experience a large change (e.g. from 1k replicas => 2k), overwhelming the other parts of the infrastructure.
We should have a tunable scale up rate of change parameter such as "scale_up_replicas_per_second" that can control how fast we increase replica counts.
Note: it can default to infinity which is the policy today.