Re: a quick poll regarding the 11gR2 OFA
- From: Frits Hoogland <frits.hoogland@xxxxxxxxx>
- To: Niall Litchfield <niall.litchfield@xxxxxxxxx>
- Date: Tue, 11 May 2010 23:22:45 +0200
please bear in mind I am not criticizing OFA in any way! I think the basic
principle still stands.
I also think the principle works, and continues to be an important thing for
server/database architecture.
What I am saying here is /u01../u99 is easily misunderstood by non-DBA's, and
even for DBA's it does not tell what the mountpoint is for.
If someone is using it, and it works: fine!
But if you want a logically and descriptive setup, the '/oracle' directory, is
descriptive (mind I say 'directory'), and having mountpoints for the software
and for the slices which hold pieces of the database in mountpoints beneath
that is descriptive in my opinion.
frits
On May 11, 2010, at 11:00 PM, Niall Litchfield wrote:
> On Tue, May 11, 2010 at 8:54 PM, Frits Hoogland <frits.hoogland@xxxxxxxxx>
> wrote:
>
> Anyone know any history of '/u01'? I guess it was used in the past to have a
> second '/usr' on a separate mountpoint, to have a clear place which holds the
> database software. If so, /u02 and further for the datafiles and other
> database files is really bogus.
>
> The original paper is available at
> http://www.method-r.com/downloads/cat_view/38-papers-and-articles (the
> author's current home online) and well worth reading. I'll just quote the
> objectives here, and remember what the typical server config was like in
> 1990, specifically relatively few, directly attached, disks on really rather
> expensive hardware.
>
> Requirement 1. The file system must be
> organized so that it is easy to administer
> growth from: adding data into
> existing databases, adding users, creating
> databases, and adding hardware.
>
> Requirement 2. It must be possible to distribute
> I/O load across sufficiently
> many disk drives to prevent a performance
> bottleneck.
>
> Requirement 3. It may be necessary to
> minimize hardware cost.
>
> Requirement 4. It may be necessary to
> isolate the impact of drive failure across
> as few applications as possible.
>
> Some of these objectives are well addressed by SAN technology, LVMs and so
> on. I think the idea stacks up pretty well. As far as naming goes, well I
> think you can tell a mathematically literate individual who understood UNIX
> filesystems of the time wrote it :(
>
> Name all mount points that will hold sitespecific
> data to match the pattern /pm where p is a
> string constant chosen not to misrepresent the contents
> of any mount point, and m is a unique fixedlength
> key that distinguishes one mount point from
> another
>
> Unfortunately I'm not entirely convinced that current definers of 10 and 11g
> OFA, are quite so rigorous. :(
>
> --
> Niall Litchfield
> Oracle DBA
> http://www.orawin.info
Other related posts:
