From: | George Neuner <gneuner2(at)comcast(dot)net> |
---|---|
To: | pgsql-general(at)postgresql(dot)org |
Subject: | partitioned tables |
Date: | 2015-05-20 11:23:11 |
Message-ID: | 7rjola5b0pg6pskoa94qlmgsdnkd3r3ln1@4ax.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
Hi all,
I've been using Postgresql (9.3) for a while, but I'm new to
partitioned tables.
1. I have noticed that when the partition trigger function returns
NULL (to stop operation on the parent table), the operation always
reports no (zero) rows affected - even when rows have been affected.
That's a problem for error checking in my client applications. Is
there a way to get the number of rows affected regardless of the
trigger?
2. I need to do upserts as opposed to just inserts. Does it make
sense to try to redirect updates to the current active partition in a
trigger (as with insert) or is it better to choose the update
target(s) using a where clause?
3. Do the child tables all need separate indexing. I've seen
conflicting information on the web - particularly in regards to
defining a primary key on the parent table.
4. I need to keep 15 months of data in circular fashion. Is there a
clever way of selecting partition by date when there are 15 of them?
Or is it waterfall time (add/delete tables every month)?
Thanks,
George
From | Date | Subject | |
---|---|---|---|
Next Message | Melvin Davidson | 2015-05-20 12:26:00 | Re: Slaves show different results for query |
Previous Message | PT | 2015-05-20 11:09:11 | Re: pg_xlog Concern |