I had asked this in a github issue but I'm reposting here to try and get an answer from a wider audience.
Has any thought gone into how kafka-connect-jdbc will be impacted by SQL transactions committing IDs and timestamps out-of-order? Let me give an example with two connections. 1: begin transaction 1: insert (get id 1) 2: begin transaction 2: insert (get id 2) 2: commit (recording id 2) kafka-connect-jdbc runs and thinks it has handled everything through id 2 1: commit (recording id 1) This would result in kafka-connect-jdbc missing id 1. The same thing could happen with timestamps. I've read through some of the kafka-connect-jdbc code and I think it may be susceptible to this problem, but I haven't run it or verified that it would be an issue. Has this come up before? Are there plans to deal with this situation? Obviously something like bottled-water for postgresql would handle this nicely as it would get the changes once they're committed. Thanks for any insight, Mark. Original github issue: https://github.com/confluentinc/kafka-connect-jdbc/issues/27