On 12-02-27 08:53 AM, Ulas Albayrak wrote:
> Hi,
>
> I noted a few specific rows in the sl_log_x tables and later, after
> they had been removed from the sl_log_x table, checked for them in the
> subscriber but they never got inserted. The answer to the question is
> answer #2, Slony thinks All changes earlier than 1.5 hours ago have
> been replicated. I know that changes were made before that time to the
> tables. Those specific tables continuously receive new data.
>
> But if Slony thinks the data has been replicated the question arises
> of why? Do the worker thread on the provider get some sort of
> acknowledgement that the data has  been transferred to the provider?
> And if so, what is this ack based on?
>

The slon threads figure out which rows in sl_log belong to a particular 
SYNC event based on the transaction id's stored as part of the sync 
event.  Slon then replicates these rows and stores a confirmation (that 
gets sent back to the master) in sl_confirm.

Do you have any rows in sl_confirm with con_origin=$master where the 
con_seqno is bigger than the largets ev_seqno in sl_event with 
ev_origin=$master ?  (this situation should not be possible).



> /Ulas
>