Unfortunately, I don't think Impala exposes this information in any way. The closest way I know of to get at this would be to look at the Kudu debug webui, which has a page for Scans (on the tablet server's debug page, under "Dashboards) which will show which tablets were actually used in a scan.
On Tue, Oct 9, 2018 at 11:59 AM Boris Tyukin <[email protected]> wrote: > Hello, > > How can I see in Impala that partition pruning happening with Kudu tables? > > When I do an EXPLAIN plan on my query or see profile, I do not see > anything that would indicate that Impala filtered partitions. > > With Impala on HDFS, I can clearly see if only a few partitions were used. > For example > > | 00:SCAN HDFS [stg.events] | > | partitions=2/30 files=2 size=14B > > > With Kudu, I am getting something like this (table is hash partitioned by > source_system_id) > > > 00:SCAN KUDU [cerner.test2] kudu predicates: source_system_id = 333 > mem-estimate=0B mem-reservation=0B > tuple-ids=0 row-size=31B cardinality=unavailable > > > Thanks, > Boris >
