Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
None
-
None
Description
Large group by on 40 billion records has different results when run. Implying something is broken with disk agg/query execution
Setup:
mcsSetConfig HashJoin TotalUmMemory 10% |
mcsSetConfig DBBC NumBlocksPct 10 |
mcsSetConfig RowAggregation AllowDiskBasedAggregation Y
|
systemctl stop mariadb-columnstore
|
systemctl start mariadb-columnstore
|
 |
create database if not exists test; use test; |
create table test_256 (a varchar(30)) ENGINE=Columnstore DEFAULT CHARACTER SET=utf8; |
cpimport test test_256 /data/split_twohundredfiftysixth_1
|
Problem Query:
time mariadb test -qse "select count(*) from (select count(a) from test_256 GROUP BY a) s;
|
Data in first comment
Proof
[root@ip-172-31-19-93 data]# time mariadb test -qse "select count(*) from (select count(a) from test_256 GROUP BY a) s;" |
count(*)
|
142064923
|
real 0m40.239s
|
[root@ip-172-31-19-93 data]# time mariadb test -qse "select count(*) from (select count(a) from test_256 GROUP BY a) s;" |
count(*)
|
142022851
|
real 0m38.298s
|
[root@ip-172-31-19-93 data]# time mariadb test -qse "select count(*) from (select count(a) from test_256 GROUP BY a) s;" |
count(*)
|
142016464
|
real 0m37.936s
|