On 7/28/2004 6:21 PM, David Parker wrote:

> I gather from slonik_commands.html that it is not possible to add a
> table to a set that has already been subscribed, and that "merge set" is
> the work-around. Is this intended to be the standard, or is adding a
> table to a subscribed set a possible future feature?

This is the standard and I have no plans to change it.

> 
> I understand that slony itself is not designed to manage failover.
> Assuming that an application detects a failure on an "origin" node in a
> cluster, I assume it is up to the application to find out to which node
> it should fail (based on the sets it cares about, and what nodes were
> subscribed), and then issue the slonik "failover" command. Is this the
> correct understanding?

Neither the application, the database, nor the replication system are 
responsible for detecting the failure of a system. This is the job of 
network management, which should call the slonik failover script you 
provided, reconfigure and restart pgpool and notify the application to 
reconnect.

> 
> Assuming node A fails, and the app fails over to node B, does slony
> provide any tools for re-initializing a new node A (assuming the
> worst-case and it has to be re-built from scratch)? My first thought
> would be to create a new node with pg_dump, then configure it for
> replication and subscribe it, but I don't see exactly how it would be
> brought into sync with updates that happen between the pg_dump/load and
> the new subscription.
> 
> I'd be interested to hear how others are handing this kind of
> failover/failback situation.

Right, if you (or your network management) made the decision to abandon 
node A, you can drop and recreate its database. To rebuild the node you 
will have to "drop node" and rebuild it as if it would be a new replica.


Jan

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