Frank,
It's very unfortunate that this bug (bz #165140) is still
not resolved
as both Oracle and SUSE eng. teams are looking into problem.
Your analysis is to the mark. Hopefully we will have some
fix (other
than disabling aio) for this annoying aio bug.
Bye,
Arun
<frank.westheider westheider.de> wrote:
> Hi Arun,
>
> Sorry, but we filed a bug...with both parties, Novell
AND Oracle.
> We escalated this case at Novell, because it's a kernel
bug...no
change for
> the last 4-6 weeks.
> But...as you see...no solution after about 3 months...
>
> I took a deep look into the kernel-code, espacially the
part of the
bug in
> aio.c
> As far as i see, it looks like a list-corruption of the
list of
outstanding
> io-requests.
> So i don't think that it is driver-specific...it looks
like a general
bug.
>
> That it is working for some customers: it depends on
the duration and
number
>
> of parallel requests if it happens or not...and these
parameter
determine
> only the timeline, when it is occuring.
>
> Bye
> Frank
>
> ----- Original Message -----
> From: "Arun Singh" <Arun.Singh novell.com>
> To: "'SuSE MailingList'" <suse-oracle suse.com>
> Sent: Monday, December 18, 2006 7:13 PM
> Subject: Re: AW: [suse-oracle] kernel bug
>
>
> Hi Frank,
>
> This patch has fixed/worked for some customers (not
all). I will
highly
> recommend to file bug as this will help to raise
severity of this
bug. This
> is a serious kernel AIO bug and only workaround at this
time is to
disable
> AIO (as you mentioned).
>
> Regards,
> Arun
>
> <frank.westheider westheider.de> wrote:
>> Hi!
>>
>> The patch is not helping.
>> We are currently running 2.6.5-7.282.PTF.165140-smp
>>
>> We already tried this...opened a TAR with Oracle,
openend an issue
with
>> Novell...got 2 fixes from Novell, but both are not
helping around
the bug.
>> The database crashes after approx. 1 week of heavy
load and you have
to
>> restart the machine to free the ipc-resources.
>>
>> After several attempts to speed up Novell fixing
this issue...we
gave up
>> and
>> turned off Async-IO. Now we're hoping for the next
service-pack to
fix
>> this
>> issue.
>>
>> Sad...but Novell was not really helping us....they
are currently
only
>> stealing time and hoping, that we close the case
and work without
>> async-io.
>>
>> Bye
>> Frank
>>
>> -----Ursprüngliche Nachricht-----
>> Von: Arun Singh [mailto:Arun.Singh novell.com]
>> Gesendet: Montag, 18. Dezember 2006 17:03
>> An: SuSE MailingList
>> Betreff: Re: [suse-oracle] kernel bug
>>
>> Hi Toni,
>>
>> <Anton.Dischner med.uni-muenchen.de>
wrote:
>>> an advice for Oracle 9.x was to
>>>
>>> >Correct procedure will be upgrade to
latest kernel (.282) and
ask
>>
>>> Novell
>>> >Tech support to provide a fix (PTF) for
this problem.
>>>
>>> Is this the way for Oracle 10.1.0.5 too?
>>>
>>> We are running kernel 2.6.5-7.257
>>
>> Yes, Please apply latest kernel which is .282 and
file bug with NTS,
If
>> it doesn't help.
>>
>> Regards,
>> Arun
--
To unsubscribe, email: suse-oracle-unsubscribe suse.com
For additional commands, email: suse-oracle-help suse.com
Please see http://www.suse.com/oracl
e/ before posting
|