Re: Corruption of Data Blocks on Oracle 8.1.7.1 and 8.1.7.0 and 9.0.1 under Redhat 7.1

Well,

thanks for the info on new certified linux distros, didn't check ML for quite a while :)

As of what piece of software may be causing corruption on asynch i/o - this is hard to guess without knowing your hardware config of disk subsystem (i.e. is it an IDE/ATA or SCSI, type, vendor and model of controller, etc.), but first thing I'd blame when it comes to corrupted data being written on disk is controller drivers. Generic drivers coming with kernel may be inappropriate for your particular controller, or, if you are using vendor drivers, they may have a bug which manifests in data corruption.

--
Vladimir Zakharychev (bob@dpsp-yes.com)                http://www.dpsp-yes.com
Dynamic PSP(tm) - the first true RAD toolkit for Oracle-based internet applications.
All opinions are mine and do not necessarily go in line with those of my employer.


"Eric P. Meyer" <eric_at_salmontrap.com> wrote in message
news:dd4cd687.0109081321.31abc8f0_at_posting.google.com...


> Vladimir,

>

> Thanks for your answer.

>

> According to Oracle's metalink, Redhat 7.0 is officialy certified &

> supported for Oracle 8i. And RedHat 7.1 is now certified with the

> 9.0.1 version.

> I did indeed set DBWR_IO_SLAVES to 4 and it is still much slower than

> with asynch io's.  Also, do you have any idea which piece of software

> would cause corruption thru the asynch io's?

>

> Eric

>

>

> "Vladimir M. Zakharychev" <bob_at_dpsp-yes.com> wrote in message

news:<9nd04l$l87$1_at_babylon.agtel.net>...


> > Eric,

> >

> > first of all, RedHat is not an officially supported Linux distribution for

> > all Oracle releases (SuSe 7.1 is the only one, plus kernel 2.4.4 and

> > glibc 2.2 are the only supported versions of kernel and C rtl).

> > As of disk_asynch_io = false being much slower than true, there is

> > a plain hint in Oracle documentation for this parameter:

> > ------

> > If you set DISK_ASYNCH_IO to FALSE, then you should also set DBWR_IO_SLAVES to a value other

than


> > its default of zero in order to simulate asynchronous I/O.

> > ------

> > Setting DBWR_IO_SLAVES to, say, 5 will create 5 additional DBWR processes

> > and Oracle will simulate asynchronous i/o using all 6 DBWRs. You may then

> > tune the number of io slaves depending on throughput you achieve.

> >

> > --

> > Vladimir Zakharychev (bob@dpsp-yes.com)                http://www.dpsp-yes.com

> > Dynamic PSP(tm) - the first true RAD toolkit for Oracle-based internet applications.

> > All opinions are mine and do not necessarily go in line with those of my employer.

> >

> >

> > "Eric P. Meyer" <eric_at_salmontrap.com> wrote in message

> > news:dd4cd687.0109070929.78afb288_at_posting.google.com...

> > > Does anyone know of any issue with Asynchronous IO's on Redhat 7.1

> > > with Oracle?

> > >

> > > I had disk_asynch_io=true in my init file, and I tried all 3 versions

> > > of Oracle and all got corrupted. Then I tried 9.0.1 with

> > > disk_asynch_io=false and it seems fine. My kernel was initialy 2.4.2

> > > and is now 2.4.4.

> > >

> > > Any idea where my problem lies? IO are much slower with

> > > disk_asynch_io=false.

> > >

> > > Eric Meyer

> > > eric_at_salmontrap.com