[Pgcluster-general] Are data changes in triggers are replicated?

[Dmitry 113]

Hello.

Let me ask a few questions...
Assume I have 2 nodes - node1 and node2; load balancer - lb; replicator -
rep.


*Q1.* General question. If I perform

INSERT INTO t(x) VALUES(1);

on node1 and *there is a trigger* on table t which updates a number of other
tables on node1:

a) are these *nested* updates replicated separately to node2 using separate
requests? or...
b) only the one input query "INSERT INTO t(x) VALUES(1)" is sent to node2
and so - triggers will be executed firstly on node1 and then - on node2?

Seems this question is about deterministic features of pgcluster, but I'm
not sure...

*Q2.* If a trigger on t updates a timestamp column with NOW(), would it
possibly causes different values at node1 and node2 or not?

*Q3.* If a stored procedure P creates another stored procedure C, would a
call to P on node1 create P on node2 too?

*Q4.* Are changes to a SEQUENCE replicated or not? E.g. if I perform INSERT,
and this INSERT uses a sequence (not SERIAL!) to generate the primary key,
would the sequence change be replicated?
*
Q5.* Should I send updation queries always to load balancer, OR I could send
it to node1 or node2 directly, without load balancer?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://pgfoundry.org/pipermail/pgcluster-general/attachments/20080622/01a380c9/attachment.html