In response to Cyril SCETBON <scetbon at echo.fr>:

> I know that it locks everything -> a really bad idea, but thanks

It also fscks up timestamps (commands like now()) and basically doesn't
ensure that your data is consistent between replicas.

> Cause we have a lot of update (14 Millions) to do. It takes 12/14 hours 
> if we want to keep a replication with a tolerated lag. If we load it 
> directly on each host in parallel ot takes 2/3 hours. I've used the 
> alterTableRestore and alterTableForReplication to load data but it stop 
> replication and I don't want it.

I recommend that you investigate this further.  Using execute script() is
a recipe for trouble in all but a few corner cases -- your data probably
isn't one of them.

If it's a bulk load table, why not ditch slony on that particular table
altogether?

If it's not a bulk-load table, then you're risking concurrent updates of
your data (slon vs. your updates) and then you shouldn't be bypassing
slony.

> 
> Bill Moran wrote:
> > In response to Bill Moran <wmoran at collaborativefusion.com>:
> >
> >   
> >> In response to Cyril SCETBON <scetbon at echo.fr>:
> >>
> >>     
> >>> Hi guys,
> >>>
> >>> Is it possible to do it ? I would like to update a field in a table 
> >>> replicated by slony manually on each node, that is to say without using 
> >>> slony, whereas replication continues for other updates.
> >>>       
> >> You can do that with execute script().  Don't know how well that fits
> >> your reqirements.
> >>     
> >
> > Actually, scratch that.  Using execute script() to do DML is a Very
> > Bad Idea.  (I should engage brain before using keyboard ...)
> >
> > I suggest you reconsider your reason for wanting this ...
> >
> >   
> 
> -- 
> Cyril SCETBON


-- 
Bill Moran
Collaborative Fusion Inc.
http://people.collaborativefusion.com/~wmoran/

wmoran at collaborativefusion.com
Phone: 412-422-3463x4023