Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
Impala 2.9.0
-
None
-
ghx-label-2
Description
Currently, our DataPartition functionality allows only for a single partition specification (e.g. HASH and the corresponding columns).
Kudu allows tables to have multilevel partitioning schemes (eg. first HASH then RANGE). In the short term, a pending change (IMPALA-3742) will represent Kudu partitioning schemes with a new TPartitionType of KUDU along with the id of the table to take the partitioning scheme from.
In the long run, we should allow for representing these multilevel partitioning schemes directly. This will allow us to make smarter choices about when to repartition data, e.g. during a grouping aggregation if the input is from Kudu and already partitioned correctly.
Attachments
Issue Links
- relates to
-
IMPALA-3118 Bucketed table support
- Open