Re: Insert into partitioned tables and specifying the partition name
Date: Sat, 10 Apr 2010 09:52:33 -0700
Message-ID: <k2qa9c093441004100952wc2e2cfe4ycd9d3de2754eb4dc_at_mail.gmail.com>
Autoextend is the other operation that can cause issues in this case. I know there are bugs logged against this scenario, but as you mentioned, the increment for the autoextend should be large enough such that the allocation is infrequent (doing anything too frequently when it can be avoided is a bad thing).
Personally I'm not a big fan of autoextend - I understand where/why it can be desirable but I'd rather size out stuff by calculation and monitor it and adjust as necessary. That's not to say there is anything wrong with it.
I think that bigfile is a fine idea, its just that there are certain
"corner/edge" cases where it can cause a little headache, and if using
RAC it can exacerbate this a bit. The two specifically that come to
mind are
1) autoextend increment being too small
2) segment next extent being too small
Both of these put the file header block (block 2) in demand and when
there is only one file and one header block it can cause (undesired)
waits, however, making the next allocations large enough should
alleviate these issues.
The other issue that I am vaguely aware of is that RMAN is unable to use parallelism for bigfile backups until 11.2 (IIRC).
On Sat, Apr 10, 2010 at 4:47 AM, amonte <ax.mount_at_gmail.com> wrote:
> After checking the datafile autoextend looks like it can cause the problems
> I am seeing since the autoextend is set to 100M only. Just changed it to
> 512M. Will see how it affects the performance
>
> Finally is bigfile really a good idea in 10g?
-- Regards, Greg Rahn http://structureddata.org -- http://www.freelists.org/webpage/oracle-lReceived on Sat Apr 10 2010 - 11:52:33 CDT