On 12/7/2004 5:11 PM, Hannu Krosing wrote:

>> On 12/6/2004 8:23 PM, James Black wrote:
>>> Hello, all,
> 
>>> merge set (id = 1, add id = 999, origin = 1);
>>
>> not only is it kosher, but the only way it works. The two sets you want
>> to merge must have the same origin and an identical list of subscribers.
>> Due to some sillyness in the way "subscribe set" currently works
>> internally, you unfortunately cannot issue any "wait for event" to have
>> the slonik script waiting until all the subscribers have finished
>> copying the initial data for the new tables. So you will have to put the
>> "merge set" part into a separate script run later.
> 
> Has anyone else noticed that if he forgets to run "merge set" later, there
> will be noticably reduced performance until he does ?
> 
> In other words, replicating the same tables using multiple sets is
> slower/harder for slony than replicating them using one set.
> 
> If it is a known problem, can anybody explain, why it must be so ?

This is not a known problem. All I could think of is that this might 
screw up the planner because the sl_log_1 selection query becomes more 
complicated.

> 
> Coult the reason be that cleanup thread fails to truncate sl_log_1 in
> this case ?

it never gets truncated in the first place.


Jan

> 
> -----------
> Hannu
> 
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at gborg.postgresql.org
> http://gborg.postgresql.org/mailman/listinfo/slony1-general


-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #