If your architecture uses MySQL replication and you make some ETL transformations and you observed some incredible values in autoincrement columns (i.e. the value dramatically increases – by hundred of thousands or event millions) – it means you have just experienced a bug.

Very clear description is presented on site http://bugs.mysql.com/.

However, in my case it wasn’t observed in test environment – only production.

How to call this event? Autoincrement is leaping, enormous gaps in autoincrement…

Solution:

Fixed in 5.1+. Documented as follows in the 5.1.64, 5.5.25, and 5.6.6 changelogs:

Statements using AUTO_INCREMENT, LAST_INSERT_ID(), RAND(), or
user variables could be applied in the wrong context on the
slave when using statement-based replication and replication
filtering server options (see
).