Hi, on Hive 3.1.4 we have a COUNT(*) query which returns more than one rows, instead of exactly one. I created this test table: CREATE TABLE test_case ( cod_pers STRING, cod_address STRING, PRIMARY KEY (cod_pers) DISABLE NOVALIDATE ) PARTITIONED BY (num_snapshot BIGINT) ; and populated it with more than 500,000 rows in 3 partitions. To find how many people changed address, I wrote this query: WITH data1 AS ( -- eliminate duplicates select t.cod_pers, t.cod_address, count(*) as num_address from test_case as t group by t.cod_pers, t.cod_address ), data2 AS ( -- find changes per person select s.cod_pers, count(*) as num_changes from data1 s group by s.cod_pers having count(*)>1 ) select count(*) as num_all_changes from data2 as gg ; Instead of obtaining a single row with the total number, the query returns 4 rows : +------------------+ | num_all_changes | +------------------+ | 63 | | 58 | | 64 | | 59 | +------------------+ 4 rows selected (1.252 seconds) 4 is also the number of reducers used by the query. If I add a "LIMIT 1" clause, the query works as expected, and a final reducer is added in the query execution. The same happens adding "GROUP BY 1" clause.   I tried to COMPACT the table and have the statistics recalculated, but nothing changed. I know I can rewrite the query, but I'm not looking for a work-aroud: I wonder if it's a known bug of CBO and if someone else experimented the same behavior (and also if a patch is available).   Thanks in advance ... View more