On 6/4/07, Dmitry Koterov <dmitry at koterov.ru> wrote:
> Hello.
>
> Seems Slony sometimes tries to play an event created BEFORE the whole
> copying process was started?
> I have the following timing:
>
> 1) in 18:28:28 a row was added to the table tableName
> 2) in 18:28:30 we have DEBUG2 remoteWorkerThread_1: Begin COPY of table
> "public"."tableName"
> 3) about 20 minutes later slony on the slave tried to replay an event (1)
> and failed with "duplicate check" violation (because an element was already
> present in the table thanks to COPY).
>
> So, seems we have to stop any write activity during the start of a COPY
> process to avoid this.

That certainly shouldn't happen. Do you have a detailed test case that
allows this error to be re-produced?

> P.S.
> Not a single, but 2 slave machines were copied sequentially. The first one
> copied and began to replicate successfully, but the second - failed with
> described symphtoms.
> Slony 1.2.9.