Re: ORACLE AND UNIX QUESTION

In article <781720147snz_at_sambusys.demon.co.uk>

           psb_at_sambusys.demon.co.uk "Paul Beardsell" writes:

> In article <2aa.553.846%mpcbbs_at_ibase.org.br>
> Carlos.Netto_at_ibase.org.br "Carlos Netto" writes:
>
> > I use raw-devices on a production machine. I don't know anything about
> > performance improvement. I choose to use raw-devices because it's possible
> > to loose a data-file when using fsck. But be sure, it's harder to administrate>
> That's like saying that I leave the lawnmower outside so that we don't trip
> over it in the lounge.
>
> I've heard some arguments for using raw partitions but that isn't a good one.
> Just put your tablespaces in their own filesystem(s) if you want to
> segregate then from the other files and other users.
>
> And put the lawnmower in the shed.
>

This is a little unfair? This actually happened to me (fsck losing a database file) a couple of times, albeit using some hardware which wasn't as stable as it ought to have been (that's why the machine crashed and needed fsck). Our policy on raw versus filesystem data files is easy; if we're running on a 3/486/Pentium with only a couple of disks (with SCO as the OS) we use filesystems; for our bigger systems (SVR4), especially if they support good async IO, we use raw partitions (usually on mirrored & striped disks to get all the resilience and speed goodies). BTW, to answer the point earlier on in the thread about overwriting your partitions with filesystems, the way to stop this is to set permissions rw-r----- oracle,dba on the device nodes (character and block) for the partition. This way, only the DBA and the server processes can do stuff to the partition. And the DBA can't edit /etc/vfstab, right?

-- 
Tom Cooke               tom_at_tomcooke.demon.co.uk             +44 (0)1782 748027
North Staffordshire Hospital Computer Centre, Stoke-on-Trent, Staffordshire, UK