That's helpful. Thanks. 

I guess I'm not even so much focused on the traffic as the need for
those db connections to be maintained. The way I read the code around
rtcfg_doActivate, and the main methods for the remoteWorker and
remoteListen threads, every active node gets one of each these threads,
which means a database connection to that node. So even if there is a
way to minimize traffic, for a large number of nodes the number of
database connections active on each node quickly becomes huge. Maybe
there is some logic filtering that number that I'm not seeing. We could
well endup with a deployment with 100+ nodes, and that translates into
more db connections than our machines can handle, I'm pretty sure!

You mentioned making sl_confirm aware of activity on a set-by-set basis,
so I'll try and figure out what that code is all about. One thing that
occurred to me is that the decision to call rtcfg_needActivate could be
based on more criteria than just no_active==true, but your sl_confirm
comment makes me suspect that there are more subtleties involved.

Aside: I've never used IRC, but I've seen it mentioned on the postgres
lists a fair amount. I found www.mirc.com - is there a better IRC
client? Seems like postgres folk use IRC a fair amount....

Thanks again.

- DAP

>-----Original Message-----
>From: Christopher Browne [mailto:cbbrowne at ca.afilias.info] 
>Sent: Friday, January 28, 2005 11:54 AM
>To: David Parker
>Cc: slony1-general at gborg.postgresql.org
>Subject: Re: [Slony1-general] quadratic listener count growth
>
>David Parker wrote:
>
>>but I'm not quite seeing why the listening relationships could not be 
>>made lighter....
>>  
>>
>One thought on this; you really don't need to have SYNCs 
>getting generated on non-origin nodes terribly often.
>
>You might therefore set the -s and -t options to be much 
>higher on nodes that aren't origins.
>
>Where the defaults now are equivalent to -s10000 and -t60000, 
>which overall force there to be a SYNC every 10 seconds, you 
>might instead use -s600000 -t1200000, which would only force a 
>SYNC every 10 minutes.
>
>That would cut down on the amount of traffic at least somewhat.
>
>I'm adding this to the documentation for slon.
>