On 12-01-03 12:34 PM, Martijn van Oosterhout wrote:
> I responded to I think most of your points in the other email, but
> there is one thing:

>>
>> Unfortunately, the UNSUBSCRIBE request comes in as an event, and it's
>> later in the event stream than the SYNC-of-the-million-deletes, so
>> it's probably not processing when you think it ought to.
>
> In our experience whenever the replication is behind, doing an
> unsubscribe is acted upon immediately. I always thought this was by
> design, though I couldn't work out why. But you're saying it shouldn't
> work at all... That's just weird.

I think Martijn is correct, my understanding of the code in 
slonik_unsubscribe_set (slonik.c) is that the command gets submitted to 
the receiver.



>
> I'll have to pay more attention next time to see exactly what happened.
>
> Have a nice day,