[Bucardo-general] Swap replication latency

[Alex Balashov]

On 05/04/2011 07:54 PM, Rosser Schwarz wrote:

> On Wed, May 4, 2011 at 4:51 PM, Alex Balashov<abalashov at evaristesys.com>  wrote:
>
>> How does that work for swap, though, where data may be getting modified
>> via the same application triggers on the other master?
>
> In a swap sync, whichever side is the logical slave for a given
> replication event has its triggers disabled for that event.  For pg>=
> 8.3, that's done by setting session_replication_role to 'replica',
> which is transactional, IIRC. Concurrent activity shouldn't be
> affected.

Thank you for that clarification.

That does put me back to square 1, though, because if there is any 
suspected cause of slowness in any of the INSERT or UPDATE statements, 
it is definitely due to the somewhat ponderous triggers attached to them.

If the triggers are disabled on the receiving side, what could possibly 
be taking so long?  INSERT statements are inherently fast, and UPDATE 
statements rely on indexes in their WHERE clause that we have verified 
are there as they should be.

-- 
Alex Balashov - Principal
Evariste Systems LLC
260 Peachtree Street NW
Suite 2200
Atlanta, GA 30303
Tel: +1-678-954-0670
Fax: +1-404-961-1892
Web: http://www.evaristesys.com/