Re: SAME(stripe an mirror everything) - where to put binaries and redos?

On Jul 13, 11:32 am, NetComrade <netcomradeNS..._at_bookexchange.net> wrote:
> On Thu, 12 Jul 2007 15:52:53 -0700, Mark D Powell
>
>
>
>
>
> <Mark.Pow..._at_eds.com> wrote:
> >On Jul 12, 6:01 pm, sybra..._at_hccnet.nl wrote:
> >> On Thu, 12 Jul 2007 13:23:42 -0700, JAGA <james.g..._at_gmail.com> wrote:
> >> >I am planning on migrating our db to a new server on using diskspace
> >> >on our SAN and have been thinking of using SAME methodology but have
> >> >questions. I initialy wanted to just 1 mount point for everything
> >> >(oracle install and all datafiles control files,.....) Now I am
> >> >thinking that it may be better to have 2 filesystems for oracle, 1 for
> >> >the binaries and one for all datafiles but I am still not sure where
> >> >to put the redo's?
>
> >> >Thanks in advance,
> >> >Jaga
>
> >> The filesystem issue is irrelevant if all filesystems live on one
> >> single device.
> >> Obviously redo's need to be separated fronm data, as redo's are
> >> written before the datafiles are written, and insufficien performance
> >> in the redo department will have negative impact on the entire
> >> database.
>
> >> --
> >> Sybrand Bakker
> >> Senior Oracle DBA
>
> >There is no reason to place the redo log files in a different file
> >system unless that file system is on a different set of disks.
>
> The reason is a potential corrupted FS.
> .......
> We run Oracle 9iR2,10gR1/2 on RH4/RH3 and Solaris 10 (Sparc)
> remove NSPAM to email- Hide quoted text -
>
> - Show quoted text -

I am not sure if file system corruption is a serious consideration or not. I am unaware of our having a single file system corruption issue on any of our UNIX platforms in the last 12 years even though we have suffered though several major power hits, UBS failure, and disk failures

On Linux and Windows this might be more of an issue. I cannot say, but with our EMC and other SAN type disk units I know our system administrators do not consider file system corruption in the disk layouts that are provided to DBA.

• Mark D Powell --