RE: Filewatcher frustration
- From: "Mark W. Farnham" <mwf@xxxxxxxx>
- To: <mcdonald.connor@xxxxxxxxx>, "'oracle-l'" <oracle-l@xxxxxxxxxxxxx>
- Date: Mon, 23 Nov 2009 10:20:37 -0500
I'm not sure whether people remember that the reason it became possible to
put the various dump and trace files to different places in the first place
(circa 1988 with V6) was that it facilitated putting the output to different
mount points and inode branches. In the era of much slower (and usually with
no write cache) disk farms and file systems, it was not hard at all to make
writing trace information, alert logs, and the various dump areas THE
bottleneck to Oracle database server throughput.
What I have not measured is whether the diag_dest scenario can become a
bottleneck, but I suspect this is not an issue because of the combination of
write cache, better handling of huge inode branches, and disk throughput.
Since they have adrci, though, I tend to agree with Niall that they didn't
have to break things to put this in place. No reason adrci couldn't read a
configuration instead of making customers carefully put symbolic links in
place to make old script sets work. (And I don't know what you do in the
operating systems of the convicted monopolists to make the equivalent fix
up.) The configuration file (or reading init or spfile) would also have
served as a dictionary of what lives where.
Still, I appreciate Oracle trying to make improvements. A little more
creativity in reducing the affect of such changes on the installed base
would be appreciated even more.
And yes, this is definitely an exadata_fusion_grid_soa_bpel issue. But why
did you leave out RAC and SAME?
_____
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Connor McDonald
Sent: Monday, November 23, 2009 9:18 AM
To: oracle-l
Subject: Re: Filewatcher frustration
The secret is to put an enhancement request in to get the diag_dest
parameter changed to:
exadata_fusion_grid_soa_bpel_dest = ....
which will get the marketing bods salivating sufficiently to get some work
done on it :-)
On Mon, Nov 23, 2009 at 2:40 PM, Niall Litchfield
<niall.litchfield@xxxxxxxxx> wrote:
I don't see any reason at all why a new facility shouldn't have its logging
and tracing infrastructure mandated as part of the design. (actually I think
retrofitting the old components and enhancing adrci for them would be a good
plan as well, but probably not one that sores highly on the marketing bumf).
<snip>
Other related posts:
