Re: large InnoDB table partitioning without explicit PK

[Rick James]

> * from Table1 where col2 like '2014-11-16 1%';
-->
WHERE col1 >= '2014-11-16 10:00:00'
AND col1 < '2014-11-16 10:00:00' + INTERVAL 10 HOUR

> [Miko] You mean analyse/check/optimise/repair table, is it correct?

No. Things like walking through the table to delete parts of it, or to study the values. As I say, subtle.

> [Miko] There will be a "fresh" installation with empty DB from the very beginning.
> But I still need to test performance degradation. This is my biggest headache I have to say.

Because of the eventual size of the table (and other issues), your testing should get a lot closer to the eventual size. We are seeing some issues now, but there may be bigger issues as it grows. The insertion rate going from 5M -> 50M scares me the most. (Hence, I have dwelled on it.) Having all the activity in a single partition that is 1% of the total makes it all feasible.

The concern over the binlog may be solved via RBR.