Hey all,

I was reviewing the Kafka connect JDBC driver, and I had a question. Is it
possible to use the JDBC driver with a look-back configured? The reason
that I ask is that there are some known issues with using a modified
timestamp:

Slide 14 here explains one with Oracle:

https://qconsf.com/sf2007/dl/QConSF2007/slides/public/JeanLucVaillant_LinkedIn.pdf?path=/QConSF2007/slides/public/JeanLucVaillant_LinkedIn.pdf

There is also some SCN-related discussion here:

https://github.com/linkedin/databus/wiki/Databus-for-MySQL

Though that is more specific to MySQL.

I am concerned that using insert IDs might not even be good enough
(assuming my tables were immutable, which they're not), since I believe
some DB storage systems might have the same issue. I think InnoDB's pkey
auto increment ID commit order is even tunable based on config.

I would rather get some duplicates than lose data if at all possible. Can I
configure the JDBC driver to subtract some number from the offset to
prevent (or drastically reduce) lost data?

Cheers,
Chris

Reply via email to