Variable query plan in pbxt.join_nested test, MariaDB 5.1.41
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
PBXT |
In Progress
|
High
|
Vladimir Kolesnikov |
Bug Description
As described by Kristian:
Using lp:~maria-captains/maria/5.1-merge:
(see end of mail for diff).
It looks like the test is not deterministic, and chooses a slightly different
execution plan between different test runs and platforms for the table t3 and
t4.
Do you have any suggestions about how to deal with this?
-------
CURRENT_TEST: pbxt.join_nested
--- /var/lib/
+++ /var/lib/
@@ -1008,13 +1008,13 @@
1 SIMPLE t1 ALL NULL NULL NULL NULL 3 100.00 Using where; Using join buffer
1 SIMPLE t2 ALL NULL NULL NULL NULL 3 100.00 Using where
1 SIMPLE t3 ALL NULL NULL NULL NULL 2 100.00 Using where
-1 SIMPLE t4 ref idx_b idx_b 5 test.t2.b 1 100.00 Using where
+1 SIMPLE t4 ref idx_b idx_b 5 test.t2.b 1 100.00
1 SIMPLE t5 ALL idx_b NULL NULL NULL 3 100.00 Using where
1 SIMPLE t6 ALL NULL NULL NULL NULL 3 100.00 Using where
1 SIMPLE t7 ALL NULL NULL NULL NULL 2 100.00 Using where
1 SIMPLE t8 ref idx_b idx_b 5 test.t5.b 1 100.00 Using where
1 SIMPLE t9 ALL NULL NULL NULL NULL 3 100.00 Using where; Using join buffer
-Note 1003 select `test`.`t0`.`a` AS `a`,`test`.`t0`.`b` AS `b`,`test`.`t1`.`a` AS `a`,`test`.`t1`.`b` AS `b`,`test`.`t2`.`a` AS `a`,`test`.`t2`.`b` AS `b`,`test`.`t3`.`a` AS `a`,`test`.`t3`.`b` AS `b`,`test`.`t4`.`a` AS `a`,`test`.`t4`.`b` AS `b`,`test`.`t5`.`a` AS `a`,`test`.`t5`.`b` AS `b`,`test`.`t6`.`a` AS `a`,`test`.`t6`.`b` AS `b`,`test`.`t7`.`a` AS `a`,`test`.`t7`.`b` AS `b`,`test`.`t8`.`a` AS `a`,`test`.`t8`.`b` AS `b`,`test`.`t9`.`a` AS `a`,`test`.`t9`.`b` AS `b` from `test`.`t0` join `test`.`t1` left join (`test`.`t2` left join (`test`.`t3` join `test`.`t4`) on(((`test`
+Note 1003 select `test`.`t0`.`a` AS `a`,`test`.`t0`.`b` AS `b`,`test`.`t1`.`a` AS `a`,`test`.`t1`.`b` AS `b`,`test`.`t2`.`a` AS `a`,`test`.`t2`.`b` AS `b`,`test`.`t3`.`a` AS `a`,`test`.`t3`.`b` AS `b`,`test`.`t4`.`a` AS `a`,`test`.`t4`.`b` AS `b`,`test`.`t5`.`a` AS `a`,`test`.`t5`.`b` AS `b`,`test`.`t6`.`a` AS `a`,`test`.`t6`.`b` AS `b`,`test`.`t7`.`a` AS `a`,`test`.`t7`.`b` AS `b`,`test`.`t8`.`a` AS `a`,`test`.`t8`.`b` AS `b`,`test`.`t9`.`a` AS `a`,`test`.`t9`.`b` AS `b` from `test`.`t0` join `test`.`t1` left join (`test`.`t2` left join (`test`.`t3` join `test`.`t4`) on(((`test`
Changed in pbxt: | |
status: | New → In Progress |
This could be tricky. Basically, the way PBXT works this is possible, because the statistics can depend on the speed of threads running in the background.
So this is likely to be a matter of timing.
We will give this test a try, and see if we also get unstable results.