Closed Mitch-Wheat closed 4 years ago
I agree it is a bit eager, however lt's assume we have 10M rows in a table. and last statistics update goes with full scan, so, rows_samples = 10M this means that we need 9M modifications to update this statistics which is not good as well as we might suffer from that before the condition to run the update will be met.
any other ideas?
as of now, since we do not have a good idea about it I will mark this as closed.
Doing a FULLSCAN of statistics as soon as any modifications are made is too eager. Added the condition mod_rows > sampled_rows * 0.9