Robert Haas wrote:
> But Tom's point about XIDs and LSNs seems like it kind of puts a
> bullet through the heart of the whole idea. Now, before you can move
> the database (or table, or whatever) between clusters, you've got to
> rewrite all the data files to freeze XIDs and, I don't know, zero out
> LSNs, or something. And if you're going to rewrite all the data, then
> you've pretty much lost all the benefit of doing this in the first
> place. In fact, it might end up being *slower* than a dump and
> restore; even an uncompressed dump will be smaller than the on-disk
> footprint of the original database, and many dumps compress quite
> well.
These are the same limitations pg_upgrade has, except it freezes the
system tables of the new cluster (very small) and then moves the clog
files from the old cluster to the new cluster to match the user files.
No way to really merge two different cluster clog files.
-- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB
http://enterprisedb.com
+ It's impossible for everything to be true. +