Re: Async IO on Oracle 9.2.0.5

"Krispyqube" <kris.plakkat_at_gmail.com> wrote in message news:1120160196.352273.90720_at_g49g2000cwa.googlegroups.com...
> Hi
>
> We are trying to figure if oracle is indeed using async io
>
> there are two articles/notes i referred to in the process and of course
> the metalink. We did relink the Oracle kernel
>
> 1. http://download-east.oracle.com/oowsf2004/1449_wp.pdf See page 14
>
> 2. http://www.grabinar.com/simon/aio.html
>
>
> i am still not sure if the server is doing async io and the other issue
> is when we run quries on the DB i see a lot of Io waits
> Checks we have done. The OS is enabled with async io looks like its the
> threaded aysnc io since kernel io is only in 2.6 +
> we are using
>
> Linux extapp1 2.4.21-27.ELsmp #1 SMP Wed Dec 1 21:59:02 EST 2004 i686
> i686 i386 GNU/Linux
>
> Oracle 9.2.0.5 with the aio related patches
>
>
> nm $ORACLE_HOME/bin/oracle | grep io_getevent
> U io_getevents@@LIBAIO_0.1
>
> ran a sql query which dd full table scans
> and trace the pid with
> strace -p pid
>
> Among other systems calls you will see either:
> a. synchronous calls
> read(3, "783|37567799|213483944|2004-10-1"..., 4096) = 4096
> b. asynchronous calls
> io_submit(0xb6a93000, 0x1, 0xbfff236c) = 1
> io_submit(0xb6a93000, 0x1, 0xbfff236c) = 1
> io_getevents(0xb6a93000, 0x1, 0x400, 0xbffeb794,
>
>
> i am only seeing case a. read
>
>
> On lthe linux box
> rpm -q --whatprovides libaio
> libaio-0.3.96-5
>
> and its enabled
> Oops i forgot Oracle is also enabled with the asynch io parameters
> filesystemio_options=asynch
> disk_asynch_io=true
>
>
> is this a fair test to confirm aio is happening or is there any other
> definitive method to confirm if we are indeed using asynch io
>
> thanks for any help
>
> regds
> Kris
>

Did you reboot the server after making async io config changes ? I don't know if it's necessary on Linux but it sure is on AIX.

Matthias Received on Thu Jun 30 2005 - 14:55:43 CDT