On Fri, Oct 20, 2006 at 02:36:48PM -0600, paul cannon wrote:
> On Wed, Oct 18, 2006 at 10:12:54PM +0000, Christopher Browne wrote:
> > We've discovered Another Problem that warrants holding on a little bit
> > for 1.2.1.
> > 
> > <http://gborg.postgresql.org/project/slony1/bugs/bugupdate.php?1591>
> > 
> > It seems that this line explains the problem: "ERROR
> > remoteHelperThread_1_1: large log_cmddata for actionseq fetch 100 from
> > LOG; not found"
> > 
> > The problem is that the "large object search" code only consults
> > sl_log_1 to look for "large tuples".
> > 
> > Where it occurs is fairly well identified; it will take a couple days to
> > get the fix working and tested :-(.
> 
> So say we've already found ourselves in this situation. Is there
> anything that can be done to get slony un-wedged? We don't even have any
> large objects.

I guess they were large enough. Turning up sync_max_rowsize to its
maximum still wasn't enough to skip the special large object
optimization handling.

I came up with a lame solution, the patch for which I'll attach to the
bug referenced above. I'm sure you slony people can come up with
something a lot more efficient and whatnot, but this at least lets
my replication work again.

-- 
paul