Hi Team, Considering I am using

replicaGroupStrategyConfig

to use

Partitioned Replica-Group Segment Assignment

and I give a column and number of instances per partition, I have two questions here: 1. What is the method used for doing this partition? 2. If I am partitioning based on a column and I want to partition a particular value of the column separately how can this be achieved?

1. You can use murmur partition, but note that the implementation of the function has to match on Pinot side and upstream where you are partitioning your input

I don't see an option to set the partition function in the

replicaGroupStrategyConfig

From this doc I see the

replicaGroupStrategyConfig

is placed under

segmentConfig

so I am assuming

segmentPartitionConfig

under

tableIndexConfig

is mandatory for using

replicaGroupStrategyConfig

?

No, partitioning and replicaGroup are orthogonal to each other.

I am aiming to achieve something like this

We have lots of tenants and I want to classify these tenants into 2 replica groups with 10 partitions and 1 server under each partition. And going forward if the load increases I can either increase the server count or split the org that is causing the increase in latency into a separate partition.

so I was hoping to use something like this in table config

{
  "instanceAssignmentConfigMap": {
    "OFFLINE": {
      "tagPoolConfig": {
        "tag": "Tenants_OFFLINE"
      },
      "replicaGroupPartitionConfig": {
        "replicaGroupBased": true,
        "numReplicaGroups": 2,
        "numPartitions": 10,
        "numInstancesPerPartition": 1
      }
    }
  },
  "segmentsConfig": {
    "replicaGroupStrategyConfig": {
      "partitionColumn": "tenantId",
      "numInstancesPerPartition": 1
    },
    ...
  },
  ...
}

But I was confused on how I can set the partitioning function here and how I can move a particular tenant into its own partition. So based on the doc you shared, if I set the

segmentPartitionConfig

with a custom plugin implementation will that suffice?