Thanks for the reply. I really want to have something generic to garbage
collect old data. This is pretty easy to do with re-writing check
constraints but much more verbose if you have drop and recreate rules,
since the column lists are different for each table. Also I have several
related tables to deal with, which adds to the fun. I think I will go
back to a non partitioned system at this point. I think the complexity
involved in the solution is not worth the gain over plain old delete.
-----Original Message-----
From: Scott Marlowe [mailto:scott.marlowe@gmail.com]
Sent: Tuesday, October 23, 2007 2:51 PM
To: Brad King
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] Partitioning and deadlocks
On 10/23/07, Brad King <brad.king@channeladvisor.com> wrote:
> Let me re-phrase this to see if I can get any response. Has anyone
used
> partitioning with inheritance successfully ? Can you point to any
> specific concurrency strategies for garbage collecting old partitions
on
> a live system ? Thank you.
I've never truncated on a live one. It is pretty easy to just update
the update/insert rules/triggers to ignore the old table, then drop
it. I'm guessing you could do that, then recreate the table and get
good performance.
