http://www.slony.info/bugzilla/show_bug.cgi?id=167

Jan Wieck <janwieck at yahoo.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |https://github.com/wieck/sl
                   |                            |ony1-engine/commit/cfdf5c51
                   |                            |1e390775fdeb43dc37b07dffc61
                   |                            |a3b20

--- Comment #1 from Jan Wieck <janwieck at yahoo.com> 2010-12-08 07:20:44 PST ---
I have combined the EXPLAIN_INTERVAL feature with the new log selection query
into one commit for review here:

https://github.com/wieck/slony1-engine/commit/cfdf5c511e390775fdeb43dc37b07dffc61a3b20

The new query seems to be completely independent of the sl_log_* size with
respect to the delay for the first row. Since the plan ALWAYS requires a Sort
step at the top and that sort does not include wide rows, we can assume that
the entire result set at that moment is in local memory of the backend unless
there is a mismatch between workmem and max sync group settings.

The delay for the first row is now consistently only a small fraction of the
entire processing time for a sync group. This completely obsoletes any need for
sync pipelining. 

As a side note, I was curious if Londiste/SkyTools is suffering from the same
problem and was surprised that their logic already resembles what I came up
with. They have a stored PL/pgSQL procedure that assembles the query while we
do it in C, but the query structure is very similar.

-- 
Configure bugmail: http://www.slony.info/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
You are the assignee for the bug.