Hi Everyone, I have a doubt. When we are performing ingestion for realtime, then in that case how do we handle load. From what I understood we need to scale Real time Server? Is it correct, if yes, then when making the API call how does broker know which server to call? Is consistent hashing involved and it figures out itself, in that case when we scale broker then how to deal with that condition?

Pinot ingestion is bounded by the number of upstream partitions, pinot side gives one thread for ingestion per partition. If you want to scale, you can have more Kafka partitions.