Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
Description
Currently, Druid Storage adapter issues a Segment metadata Query every time the query is of type Select or Scan. Not only that but then every input split (map) will do the same as well since it is using the same Serde, this is very expensive and put a lot of pressure on the Druid Cluster. The way to fix this is to add the schema out of the calcite plan instead of serializing the query itself as part of the Hive query context.
Attachments
Attachments
Issue Links
- contains
-
HIVE-14518 Support 'having' translation for Druid GroupBy queries
- Closed
-
HIVE-18993 Use Druid Expressions
- Closed
- relates to
-
HIVE-18957 Upgrade Calcite version to 1.16.0
- Closed