Need "realtimeRowsConsumed" metric for Low Level Consumer too

[dongxiaoman]

Currently we are not seeing an aggregated per-host realtimeRowsConsumed metric when using LLC. That metric is useful because currently per-partition-per-table metric is too much detail meant for debugging.

It is possible to do metric aggregation to have this "per-host realtimeRowsConsumed" metric with some metric limit like the 250(450?) series in DataDog StatsD, it is not possible to get that many series into it.

[xiangfu0]

I feel we can have table level aggregated metrics for each host. Just per host level metrics might be too coarse?