Estimation info of planner is seriously wrong under ch-benchmark #42961
Labels
affects-6.1
This bug affects the 6.1.x(LTS) versions.
affects-6.5
This bug affects the 6.5.x(LTS) versions.
affects-7.0
sig/planner
SIG: Planner
type/enhancement
The issue or PR belongs to an enhancement.
Bug Report
Please answer these questions before submitting your issue. Thanks!
Similar issue #17185
1. Minimal reproduce step (Required)
Run daily ch-benchmark regression tests.
2. What did you expect to see? (Required)
Estimation info is similar to the real.
3. What did you see instead (Required)
ch-benchmark Q5
tidb will choose the unexpected join method. The estimate row count of
HashJoin_56
is far more bigger than the real one. It may cause the chosen join algorithm(Broadcast Join) is not the best one because the size of hash table when using broadcast join is bigger than the shuffle way, which means it may bring worse performance.4. What is your TiDB version? (Required)
master
The text was updated successfully, but these errors were encountered: