Hello all,
I'd like to post here me experience of using new Linux kernels both at
home and on enterprise server.
At home I'm using new branch since 2.3.x - it works really fine for
home work, multimedia, etc. Nice work, thanks.
But on our enterprise server It's something awful.
Server hardware is: PIII 500Mhz, Intel 440GX, 512 ECC RAM, 2xAdaptec 7xxx +
Mylex AccelRAID 250 disk subsystem.
Server software is 2 instances of Oracle 8i (8.1.6 and 8.1.7).
No kernel modules, no inet services, only Oracle.
On 2.2.x this run more or less stable (it usually crash or hang once
per month, but it's acceptable).
I've installed 2.4.x there. Just immedualtely I've noticed performance
improve, responce time improve.
BUT: All kernels prior to 2.1.4-ac8 hangs during first few hours of
work on heavy disk (Mylex) activity.
2.1.4-ac8 was the first kernel which was able to work nore then 24
hours. But on 26'th our of work it crashed with:
Kernel panic: Aiee:, killing interrupt handler!
In interrupt handler - not syncing
So, again I've downgraded kernel to 2.2.18 again :(
Can I know your thoughts about target market of 2.4.x kernel? I assume
that the goal is to make it feature-rich multimedia desktop system?
Personally I now look how to run Oracle under FreeBSD which is much
more stable on the same hardware at high load (corporate Squid).
--
Best regards,
Andriy mailto:[email protected]
> I've installed 2.4.x there. Just immedualtely I've noticed performance
> improve, responce time improve.
That is good news
> BUT: All kernels prior to 2.1.4-ac8 hangs during first few hours of
> work on heavy disk (Mylex) activity.
Ok Im glad to know we have made some progress there
> 2.1.4-ac8 was the first kernel which was able to work nore then 24
> hours. But on 26'th our of work it crashed with:
>
> Kernel panic: Aiee:, killing interrupt handler!
> In interrupt handler - not syncing
What is the oops data before the kernel panic. I need that to debug the
driver. Also did you build the DAC960 support with gcc 2.96-x x<74 ?
> Can I know your thoughts about target market of 2.4.x kernel? I assume
> that the goal is to make it feature-rich multimedia desktop system?
And a rock solid server and anything else people want it to be.
Alan> What is the oops data before the kernel panic. I need that to debug the
Alan> driver. Also did you build the DAC960 support with gcc 2.96-x x<74 ?
My system compiler is:
gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)
Shoud I upgrade it to gcc 2.95.x or 2.96.x?
And since my last post 2.4.1-ac8 crashed again with:
kswapd: Cannot dereference NULL pointer.
On the next crash I'll try write down all traces.
BTW, is there some way to log it somewhere to file?
--
Best regards,
Andriy mailto:[email protected]
On Sun, Feb 11, 2001 at 08:01:48PM +0200, Andriy Korud wrote:
> Alan> What is the oops data before the kernel panic. I need that to debug the
> Alan> driver. Also did you build the DAC960 support with gcc 2.96-x x<74 ?
> My system compiler is:
> gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)
> Shoud I upgrade it to gcc 2.95.x or 2.96.x?
>
> And since my last post 2.4.1-ac8 crashed again with:
> kswapd: Cannot dereference NULL pointer.
>
> On the next crash I'll try write down all traces.
> BTW, is there some way to log it somewhere to file?
Considering this is an enterprise-server, you should have a serial-console
to it; this will allow you to log all error-messages. Alternatively
(not as nice, though), is to connect a printer to the parallel-port and
use the parallelconsole option.
Then there's always the LKCD-patches, by SGI, if I'm not all wrong.
/David Weinehall
_ _
// David Weinehall <[email protected]> /> Northern lights wander \\
// Project MCA Linux hacker // Dance across the winter sky //
\> http://www.acc.umu.se/~tao/ </ Full colour fire </
> Alan> driver. Also did you build the DAC960 support with gcc 2.96-x x<74 ?
> My system compiler is:
> gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)
> Shoud I upgrade it to gcc 2.95.x or 2.96.x?
No that one is fine. I have a known problem with DAC960 and cvs gcc or
gcc 2.96.x x<74 where struct sizes change and it doesnt work.
> On the next crash I'll try write down all traces.
> BTW, is there some way to log it somewhere to file?
On a crash we don't log it to a file. The problem is that you cannot be
sure of the state of the system so your logging might itself crash or
do more harm.
The important bit is the EIP value and the call trace.
On Sun, Feb 11, 2001 at 07:13:29PM +0100, David Weinehall wrote:
> Considering this is an enterprise-server, you should have a serial-console
> to it; this will allow you to log all error-messages. Alternatively
> (not as nice, though), is to connect a printer to the parallel-port and
> use the parallelconsole option.
I wrote I little palm app some time ago that can capture serial
console output. If anyone is interested I'll build a tar ball with
sources an binary.
regards Christian
--
THAT'S ALL FOLKS!
On Mon, Feb 12, 2001 at 05:32:39PM +0100, I wrote:
> I wrote a little palm app some time ago that can capture serial
> console output. If anyone is interested I'll build a tar ball with
> sources and binary.
It is now availiable at http://www.mathematik.uni-ulm.de/~ehrhardt/serial/
Sorry for the slightly off topic post but I think enough people
showed interest to justify this.
regards Christian Ehrhardt
--
THAT'S ALL FOLKS!