On 4/22/2013 10:05 AM, Moshe Jacobson wrote:
Setup a slave.
Not bad, but the transaction logs would fill up the file system.
Besides, it would not be worth it to set up a whole database cluster
just for this purpose.
Any alternatives?
if your file system supports it (ZFS, for instance), you /could/ take a
volume snapshot/clone, and then mount that and run a separate instance
of PostgreSQL on it, using a different port number. when done with your
reports, you'd drop that instance and snapshot.
we run a separate server (hardware and database) for our reporting
systems as their workload is heavier than the realtime transaction system.
--
john r pierce 37N 122W
somewhere on the middle of the left coast
