Hi,

I know that we have not so far implemented a way to use COPY on
multiple nodes, and then add the table to replication, without
causing the associated tables to get blown away and re-copied.

I have a use-case I've been working on where what I was originally
planning to do was load >10M records in little hunks of INSERT
statements.  My colleague, however, just asked me, "Isn't there some
way we could cheat and bulk load it?"  I initially said no, but then
I thought of this, and I'm wondering whether anyone can think of a
reason it _can't_ work before I go away and start testing.

The idea is that I have an empty table that is already added to
replication.  This table is not being accessed by any application
(it's part of the support for the new application.)  I have no
indexes on it.  

Couldn't I just generate a set of files to hand to COPY on each node,
populate the table that way, then set the sequence numbers up
correctly, add the indexes, and be ready to go?  Or is there
something I am overlooking in my excitement to speed up my
application migration so much?

A

-- 
Andrew Sullivan  | ajs at crankycanuck.ca
If they don't do anything, we don't need their acronym.
		--Josh Hamilton, on the US FEMA