Re: Logical replication from 11.x to 12.x and "unique key violations"

Thomas Kellerer <shammat@gmx.net> writes:
> Tom Lane schrieb am 20.07.2020 um 20:04:
>> Yeah, duplicate keys does seem odd here.  Can you provide a self
>> contained example?

> I'll try, but this is a production system.
> Extracting the necessary anonymous data will be tricky.

If this is a PG bug, it should be possible to reproduce it with
completely random/generated data.  The key ingredient that you
have and the rest of us don't is the process and timing by which
the primary key values are introduced.

> Is there any chance the version difference might cause this?
> And a slightly outdated 11.x at that?

Hmmm ... I do not recall any recent bug fixes that seem to match
this symptom, but replication isn't a part of the code that
I'm the world's best expert on.  In any case, we do offer as
standard advice that you should reproduce a problem on the latest
minor release before filing a bug report.

            regards, tom lane