Re: Creating DB on partition > 2TB
- From: "Mark Brinsmead" <pythianbrinsmead@xxxxxxxxx>
- To: "William Wagman" <wjwagman@xxxxxxxxxxx>
- Date: Thu, 6 Mar 2008 19:23:26 -0700
Try telling them that if they really want a single 3 TB filesystem, you'll
be happy to provide it but *they* need to pay for the tape drives (and
network, and...) needed to back it up in under 3 hours.
Worst case, thye'll say "okay", and *you* get a bunch of cool new toys. :-)
Usually, though, most people will back down on "requirements" they don't
really need once they find out its gonna cost some serious cash. Not always
though. (I could tell you some stories...)
(On the other hand, you maybe ought to be taking this position on the backup
hardware no matter *how* the storage is sliced. I *hate* it when people
deploy huge new systems and then expect me to pull the required backup
infrastructure out of thin air!)
On Thu, Mar 6, 2008 at 7:10 PM, William Wagman <wjwagman@xxxxxxxxxxx> wrote:
> Mark,
>
> That makes a lot of sense. I just need to convince the requestors.
>
> THanks.
>
>
> Bill Wagman
> Univ. of California at Davis
> IET Campus Data Center
> wjwagman@xxxxxxxxxxx
> (530) 754-6208
>
>
> ------------------------------
> *From:* Mark Brinsmead [mailto:pythianbrinsmead@xxxxxxxxx]
> *Sent:* Thursday, March 06, 2008 6:06 PM
> *To:* greg@xxxxxxxxxxxxxxxxxx; William Wagman
> *Cc:* oracle-l@xxxxxxxxxxxxx
> *Subject:* Re: Creating DB on partition > 2TB
>
> I have to agree with Greg here. There are other things to consider beyond
> (just) IO performance, too.
>
> There is also the question of recover time should come up here as well.
> In building a single LUN (or filesystem, or...) of 3TB, this becomes the
> amount of data you will -- or at least may -- have to recover in the even of
> a serious failure.
>
> Before you say "I have RAID protection at the hardware level", let me ask
> this: Have you even seen somebody reformat the wrong disk, or do *rm -rf*in
> the wrong filesystem? I know
> *I* have! (Seen it, that is. I've never actually *done* that. Oh, no,
> no, no...) :-)
>
> Even in cases where you do have hardware RAID support (and you actually *
> believe* the B.S. that the salesrep pushed about *his* RAID product being
> infallible) there are still plenty of failure modes that can result in the
> need to restore data to an entire LUN or filesystem.
>
> Compartmentalizing your data may not *prevent* such accidents, but it can
> at least *help* to restrict the scope of the damage, and perhaps put you
> in a position where you only need to restore 100GB of data from tape,
> instead of the while 3TB.
>
> (Yeah, yeah, I know: you don't plan to use the *whole* 3 TB. Yet!)
>
>
> --
> Cheers,
> -- Mark Brinsmead
> Senior DBA,
> The Pythian Group
> http://www.pythian.com/blogs
>
>
> On Thu, Mar 6, 2008 at 4:08 PM, Greg Rahn <greg@xxxxxxxxxxxxxxxxxx> wrote:
>
> > Sounds like a DBCA bug from a really large number. Oracle db doesn't
> > have any knowledge of the LUN size, only database file size.
> >
> > I don't know that I would recommend just a single large LUN. I
> > believe that things like the SCSI IO queue are per LUN so with only a
> > single LUN you only get one queue. This probably will limit the
> > number of in-flight IOs.
> >
> > On 3/6/08, William Wagman <wjwagman@xxxxxxxxxxx> wrote:
> > > Greetings,
> > >
> > > I have installed Oracle 10.2.0.3.0 on a server running Centos 4,
> > > 2.6.9-67.0.4.Elsmp. This is a new server and installation. I am using
> > > the DBCA to create the database which initially will only be 2GB. The
> > > partition on which I am attempting to create the database has nearly
> > 3TB
> > > of space and ..
> > <//www.freelists.org/webpage/oracle-l>
> >
> >
> >
>
>
--
Cheers,
-- Mark Brinsmead
Senior DBA,
The Pythian Group
http://www.pythian.com/blogs
Other related posts:
