Re: How does Oracle determine how much to extend autoextend datafiles?
From: Tanel Poder <tanel_at_tanelpoder.com>
Date: Thu, 14 Mar 2019 18:59:59 -0400
Message-ID: <CAMHX9JKYAam7WtUaiGFA8TgroUN9qOgbD65caGTwwxtdj_T26A_at_mail.gmail.com>
Date: Thu, 14 Mar 2019 18:59:59 -0400
Message-ID: <CAMHX9JKYAam7WtUaiGFA8TgroUN9qOgbD65caGTwwxtdj_T26A_at_mail.gmail.com>
There's a _enable_space_preallocation parameter that controls this and also plenty of MOS notes:
- SMCO (Space Management Coordinator) For Autoextend On Datafiles And How To Disable/Enable (Doc ID 743773.1)
- AUTOEXTEND Grows To Full Size Without Reason (Doc ID 1459097.1)
- Master Note: Overview of Oracle Segment Storage (Doc ID 1491960.1) - *search for SMCO*
-- Tanel Poder https://blog.tanelpoder.com On Thu, Mar 14, 2019 at 2:16 PM Rich J <rjoralist3_at_society.servebeer.com> wrote:Received on Thu Mar 14 2019 - 23:59:59 CET
> On 2019/03/14 12:40, Chris Taylor wrote:
>
> I just thought the extent management in Oracle in LMT - depending on the
> size of the segment, Oracle auto determines the next extent size when using
> automatic space management.
>
> So depending on the size of the segments involved (table, indexes, lobs(?)
> ) those next extents might have been large.
>
> I think I saw an Oracle support document on the extent size thresholds
> (based of segment_size) but I can't lay my hands on it at the moment.
>
>
>
> I'd agree that the algorithm might make the next extent "large", but the
> new extent(s) totaled less than 2GB. Which is why I'm curious what was
> going on that the TS grew by exactly 31GB.
>
> After verifying the top used block in each datafile, I resized them. Sure
> enough, I got back about 28GB of unused/unallocated space. The ERP that
> uses this DB doesn't add/drop segments on-the-fly or anything like that.
> The only event that I can see where the datafiles would autoextend is
> segment growth. And there's a 28+GB difference between how much the
> segment(s) grew and how much the TS grew. Weird.
>
> Thanks for the feedback, Chris!
>
> Rich
>
-- http://www.freelists.org/webpage/oracle-l