Currently the helm chart prototype only supports configuration for global affinity settings. Best practice is to define anti-affinity rules between multiple replicas of an instance so they can be forced to schedule on different nodes, but that is currently not allowed.
If possible, add support for per-service affinity settings. As a nice-to-have, allow configuration at the global level as well. Also consider whether tolerations and node selectors could be treated in the same way (per-service settings).
Currently the helm chart prototype only supports configuration for global affinity settings. Best practice is to define anti-affinity rules between multiple replicas of an instance so they can be forced to schedule on different nodes, but that is currently not allowed.
If possible, add support for per-service affinity settings. As a nice-to-have, allow configuration at the global level as well. Also consider whether tolerations and node selectors could be treated in the same way (per-service settings).