On July 17, 2007 02:33 pm, Al Rosenthal wrote:
> Hello,
>
> Thank you for your reply.  I think that you may be correct.  I looked at
> the link that you suggested and that sounds very much like the problem. 
> Some of the bytea data is greater than 64 mb.  I am running 8.2.3 on
> windows.  I have tried changing all of the registry parameters for TCP in
> windows but to no avail. I would try changing the winsock parameters but it
> doesn't sound like that has made a difference either.  When I do an insert
> on this table it can often be a large amount of binary data (250+ pages of
> scanned documents).  So far, the PQexecParams statement works fine and
> never complains.  Maybe some of the internal buffers in winsock are not
> getting released.
>
> Think maybe I'm stuck with a bunch of pg_dumps and storage on a removable
> hard disk.  Pity.  A restore on a db of that size would take almost a day.
> Not much for availability.

I'd try bumping this thread over to the postgresql mailing list and see if the 
windows hackers there have any other ideas.

>
>
> Again thanks.
>
> Al Rosenthal
>
> > Based on the descriptions you have given, I'm fairly convinced that this
> > is
> > not a slony problem, but rather a PostgreSQL more specificaly a Windows
> > problem,   During the subscribe phase of the replication slony basicly
> > mimics
> > pg_dump | psql.   The fact that you've moved the slon processes to a
> > linux machine yet still get this error,  pretty much confirms this for
> > me.  Some googeling around leads me to see the following:
> > http://archives.postgresql.org/pgsql-ports/2005-11/msg00000.php  Which
> > sounds
> > like it could be exactly what you are experiencing.
> >
> >
> >
> > Darcy Buskermolen
> > The PostgreSQL company, Command Prompt Inc.
> > http://www.commandprompt.com/

-- 
Darcy Buskermolen
Command Prompt, Inc.
+1.503.667.4564 X 102
http://www.commandprompt.com/
PostgreSQL solutions since 1997