Edward Gao

(the code changes themselves look pretty reasonable, I just want to doublecheck on testing + figure out a plan for getting test coverage)

@subodh1810 I've only skimmed this PR (and basically haven't looked into that issue at all), but I think they should be aligned? https://github.com/airbytehq/airbyte/issues/14628 is probably easier if we have unified...

actually: for https://github.com/airbytehq/airbyte/issues/14628, doesn't this PR resolve most of it? * mysql / other DB sources still need to be fixed, but they can be based on this PR *...

merged in master + fixed conflicts + enabled catalog assert for the CDC tests (these were already passing, at least on my laptop!)

/test connector=connectors/source-postgres > :clock2: connectors/source-postgres https://github.com/airbytehq/airbyte/actions/runs/2749458333 > :white_check_mark: connectors/source-postgres https://github.com/airbytehq/airbyte/actions/runs/2749458333 No Python unittests run > ### Build Passed Test summary info: ``` All Passed ```

summarizing some [slack discussion](https://airbytehq-team.slack.com/archives/C03C4AVJWG4/p1663192739940639): There's maybe a technically feasible implementation in platform (where it detects a lack of checkpointing and forces the destination to commit). This has some advantages (possibly-reduced...

@yurii-bidiuk could you show an example of what an exception stacktrace looks like? (i.e. if the connector has a `throw new RuntimeException("whatever")`, does that also get wrapped correctly?) And is...

Does the original stacktrace never get logged? That seems problematic. Losing the original stacktrace will make debugging connector crashes a lot more difficult. https://logging.apache.org/log4j/2.x/manual/json-template-layout.html says there are ways to add...

I'd pretty strongly prefer to have the exception stacktrace logged immediately, rather than captured inside the sync summary log. Otherwise it could hide important information (e.g. timestamp).

does this mean that anywhere else we use the `LOGGER.log("message", e)` call, it needs to be replaced by the ExceptionUtils.getStackTrace style? (that doesn't sound ideal) also @yurii-bidiuk @grishick feel free...