Jeff Frost wrote:
> We're going to be bringing up a new master and had planned to add it 
> to replication, let it sync, then do a graceful switchover.  But, that 
> would leaev it with a node id of 5.  Since the old master (node id = 
> 1) is going away, I'm wondering if it's possible to modify the node id 
> of the new master?
>
No, there is no notion of modifying the node ID; the ID is very deeply 
intertwined into replication state across all the nodes, with the result 
that trying to alter the ID would be likely to lead to potential for 
*enormous* confusion within the replication system.

There is no reason to have a "node 1" around; many of the tests assume 
its absence, and in most of our production clusters, we use node numbers 
that are *somewhat* meaningful.

For instance, several clusters consist of nodes:
  8141
  8142
  8143
  8194

Where:

- "81" indicates that it's a PostgreSQL 8.1 instance  (we may soon have 
nodes starting with "83")
- 4 versus 9 indicates which data centre the node is at
- below all of that, we have a more-or-less serial value, 1,2,3,4

It's not quite wonderful, but it's mnemonic enough to be useful...