On Wed, Jun 20, 2007 at 08:59:35AM -0500, Jeff Amiel wrote:
> themselves were nearly identical in size.....but the indexes in some 
> cases were significantly different (in one case, the master node disk 
> size for index for a specific table was 6 gigabytes...on the subscriber 
> node, it was only 1 gig).

There remain some cases where indexes don't get compacted
appropriately.  I believe that some of this is addressed in 8.2 and
yet some more in 8.3.

> I am assuming that  empty index entries are not reused and that only a 
> re-index on the indexes (on the master node) will recreate (and 
> shrink)....but this requires a lock on the entire index, correct?

Right.  In later versions than 8.1, you can do indexing on the side,
which is incredibly cool and would solve your problem.  But it won't
help you in this case.  It isn't an exclusive lock -- you can read --
but it does block writes.

A

-- 
Andrew Sullivan  | ajs at crankycanuck.ca
However important originality may be in some fields, restraint and 
adherence to procedure emerge as the more significant virtues in a 
great many others.   --Alain de Botton