The first question; When we enable

aggregateMetrics

to pre-aggregation as it is consumed, pinot aggregates data based on fields which defined in

dimensionFieldSpecs

and

dateTimeFieldSpecs

. Can pinot aggregates data only based on fields which defined in

dimensionFieldSpecs

while applying pre-aggregation using

aggregateMetrics?

The second question; We can set time to generate segments for real-time table using

realtime.segment.flush.threshold.time

config. Let's assume current hour is 10:25. When i set

realtime.segment.flush.threshold.time

to

1 hour

, pinot creates segment with startTime 10:25, and it will close this segment when time is 11:25. As a result, start/end time of that segment is 1025 1125. But when the new hour starts, I want pinot to close segment.. Start/end time of that segment should be 1000 1100. How can i achieve that?

1. aggregating only on dimensionFieldSpecs can result in wrong results right. If you want aggregation on specific dimensions - use startree index 2. Use realtime to offline converter minion task. This will periodically merge multiple segments and create time partitioned segments

When i use the star-three index, pinot only looks the start-tree document, right? It still manages time boundary using time column. So, if i have columns like that;

dimensions: sellerId, brand, category, productId, productName

metrics: totalCount

timeFields: eventDate, orderDate

i want to get aggregated results based on orderDate, brand, category, productId for specific seller, category or brand. Some possible queries;

select productId, sum(totalCOunt) 
from t 
where sellerId = x and orderDate > Y and orderDate < Z and category = '
group by productId

select category, sum(totalCOunt) 
from t 
where sellerId = x and orderDate > Y and orderDate < Z 
group by category

select brand, sum(totalCOunt) 
from t 
where sellerId = x and orderDate > Y and orderDate < Z 
group by brand

I have to create star-tree index on sellerId, brand, category, productId and orderDate, right? In that case, what happens when i want to get productName?

That’s right. You can create index on a specific column