On 7/26/2004 2:00 PM, David Parker wrote:

> Ah, good point. Thanks for the responses, everybody.
> 
> So, there seems to be a consensus that A needs to know about C (or vice
> versa), in my example? Apart from the failover case you just mentioned,
> why is that? Allowing C to be ignorant of A would seem to allow for A to
> fail without having any effect on C's replication. 

Every node needs to know about every node. And every node needs to 
listen for events from every node. That doesn't mean that every node 
needs to talk to every node, it is enough that in your example A and B 
talk to each other, and B and C talk to each other. Yet, B has to 
"forward" all news from A to C and vice versa.

Just because you consider A as "master" doesn't mean that C can be 
entirely ignorant. In Slony no node is ever the master of everything. 
They can own a couple of objects, but they are never master of the cluster.


Jan

> 
> As it is, in an A failure case, does the configuration for C need to
> change? 
> 
> One question just leads to another with me....
> 
> - DAP
> 
> -----Original Message-----
> From: slony1-general-bounces at gborg.postgresql.org
> [mailto:slony1-general-bounces at gborg.postgresql.org] On Behalf Of
> Christopher Browne
> Sent: Monday, July 26, 2004 1:53 PM
> To: Darcy Buskermolen
> Cc: slony1-general at gborg.postgresql.org
> Subject: Re: [Slony1-general] replication chain
> 
> Darcy Buskermolen <darcy at wavefire.com> writes:
>> On July 26, 2004 08:43 am, David Parker wrote:
>>> Say I have 2 geographically separated data centers, East and West, 
>>> where is West is the "warm" backup for East. In East I have 2 
>>> databases, A and B, where B is a "hot" backup for A, so A is the 
>>> "master" in that replication cluster.
>>>
>>> I'm wondering about the best way to set up replication for my 
>>> database C, in the West data center. It is the second level back for 
>>> A, but I would like to avoid adding the load on A of having C making 
>>> requests against it to get updates.
>>>
>>> My thought was that I could set up B as the master for C, so when B 
>>> replicated from A, it would trigger a further replication to C, but C
> 
>>> would only need to connect to B. So the replication chain would be:
>>> A->B->C.
>>
>> This is an exact feature of Slony I.
>>>
>>> Would this work, or does C need a connection to A for some reason?
>> Node B subscribes to node A, and Node C subscribes to Node B, Node A 
>> does not need to know Node C exists.
> 
> You probably also want a plan to be able to reconfigure A and C to be
> able to talk directly supposing B should happen to get "taken out."
> 
> This is also a sort of "exact feature" of Slony-I...
> --
> "cbbrowne","@","ca.afilias.info"
> <http://dev6.int.libertyrms.com/>
> Christopher Browne
> (416) 673-4124 (land)
> _______________________________________________
> Slony1-general mailing list
> Slony1-general at gborg.postgresql.org
> http://gborg.postgresql.org/mailman/listinfo/slony1-general
> 
> _______________________________________________
> 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 #