Bah. sql, query, queries, smallint
In the last episode (Dec 30), Rick Faircloth said:
> Aaaahhhh, that's how the auto_increment id problem is solved.
>
> I wonder why they can't do something like Access with its
> ReplicationID or whatever it's called...
Access just used a random 32-bit integer, and you risked generating
dupes. If you go this route it might be better to create a dual-column
primary key, with the first column being the server-id taken from
my.cnf (already visible in the @server_id variable), and the other
column being a standard AUTO_INCREMENT field (but with each server
keeping track of its own max ID).
Oracle manages this by handing out "groups" of IDs to each server. The
first server to add a record gets IDs 1-49, for eample, and the next
server gets 50-99. When a server uses its range up, it asks for
another block (which server is in charge of handing out groups I don't
know. Maybe the one with the highest uptime?)
http://download-west.oracle.com/docs/cd/B10501_01/rac.920/a96598/design.htm#1028013
--
Dan Nelson
dnelson@stripped
