After trying to find the reason for missing transactions reported in my previous messages, I started examining Slony 
source code and came upon a comment about an unconfirmed suspicion of missing transactions in cleanupEvent(). After that 
I increased the cleanup_interval setting to 6 hours so that cleanupEvent() would keep a lot more data around. This did 
the trick and for a week or so we didn't see a single row go missing.

Today I reduced the cleanup_interval back down to 10 minutes and ran the usual scripts. Rows went missing again. This 
leads me to believe that cleanupEvent() does indeed remove rows that are still needed.

Just like previous times there were no crashes and there's nothing unusual in either Slony or PG logs.

Regards,
Aleksander


Aleksander Kmetec wrote:
> Hi,
> 
> Aleksander Kmetec wrote:
>> in the past that are missing. Unfortunately I can't determine if they 
>> were all made during the same transaction or not.
> 
> I have now performed some additional checks and it turns out that all 
> the changes have indeed most likely been performed inside a single 
> transaction by a script which performs daily imports from an external 
> database.
> 
> This script can sometimes take hours to complete (it's heavily throttled 
> in order to have minimum impact on performance) and it also deletes many 
> rows only to replace them with exact same ones.
> 
> Could either of this have caused a whole transaction to disappear from 
> replication?
> 
> Regards,
> Aleksander
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at lists.slony.info
> http://lists.slony.info/mailman/listinfo/slony1-general
>