2007-08-28 18:40:30

by Pete Monroe

[permalink] [raw]
Subject: Crash report 2.6.22.5

Hi,

Sorry there's not more to go on here.

A 32-bit firewall running the kernel LVS virtual server to fan out to
a dozen webservers ran fine for a year using 2.6.17.13, but won't
last more than four hours or so with 2.6.22.5. Another server,
different hardware and vendor but same purpose, also crashed with
2.6.22.5 after a few hours. It had previously run 2.6.20.11. Nothing
on the screen, nothing in the logs.

I'm attaching zipped dmesg (both kernel versions), .config and lspci
-v output for one of the machines, a Dell Intel dual-Xeon box. The
other machine is a dual Athlon box. Both use SCSI drives (the
attached Dell uses MPT Fusion, the other one Adaptec.) Intel ethernet
on both.

I did enable the Slub allocator in 2.6.22.5, figuring that if it is
going to be the default in 2.6.23 that it's probably solid in .22.5.

PLMK if any more info would be useful.

Thanks,
Pete


Attachments:
(No filename) (891.00 B)
c2-lspci.out.gz (1.34 kB)
c2-2.6.22.5.dmesg.out.gz (5.41 kB)
c2-2.6.22.5.config.gz (7.11 kB)
c2-2.6.17.13.dmesg.out.gz (5.60 kB)
Download all attachments

2007-08-28 23:51:22

by Michal Piotrowski

[permalink] [raw]
Subject: Re: Crash report 2.6.22.5

Hi Pete,

On 28/08/07, Pete Monroe <[email protected]> wrote:
> Hi,
>
> Sorry there's not more to go on here.
>
> A 32-bit firewall running the kernel LVS virtual server to fan out to
> a dozen webservers ran fine for a year using 2.6.17.13, but won't
> last more than four hours or so with 2.6.22.5. Another server,
> different hardware and vendor but same purpose, also crashed with
> 2.6.22.5 after a few hours. It had previously run 2.6.20.11. Nothing
> on the screen, nothing in the logs.
>
> I'm attaching zipped dmesg (both kernel versions),

Could you capture the bug with serial/netconsole etc.?

"Collecting kernel messages"
http://www.stardust.webpages.pl/files/handbook/handbook-en-0.3-rc1.pdf
for more info.

> .config and lspci
> -v output for one of the machines, a Dell Intel dual-Xeon box. The
> other machine is a dual Athlon box. Both use SCSI drives (the
> attached Dell uses MPT Fusion, the other one Adaptec.) Intel ethernet
> on both.
>
> I did enable the Slub allocator in 2.6.22.5, figuring that if it is
> going to be the default in 2.6.23 that it's probably solid in .22.5.
>
> PLMK if any more info would be useful.
>
> Thanks,
> Pete
>
>

Regards,
Michal

--
LOG
http://www.stardust.webpages.pl/log/

2007-08-29 00:07:15

by Pete Monroe

[permalink] [raw]
Subject: Re: Crash report 2.6.22.5

On 8/28/07, Michal Piotrowski <[email protected]> wrote:
> Hi Pete,
>
> On 28/08/07, Pete Monroe <[email protected]> wrote:
> > Hi,
> >
> > Sorry there's not more to go on here.
> >
> > A 32-bit firewall running the kernel LVS virtual server to fan out to
> > a dozen webservers ran fine for a year using 2.6.17.13, but won't
> > last more than four hours or so with 2.6.22.5. Another server,
> > different hardware and vendor but same purpose, also crashed with
> > 2.6.22.5 after a few hours. It had previously run 2.6.20.11. Nothing
> > on the screen, nothing in the logs.
> >
> > I'm attaching zipped dmesg (both kernel versions),
>
> Could you capture the bug with serial/netconsole etc.?

The servers are remote, production servers and it's a PITA when they
crash. But I'll see what I can do. Thanks for the pointer.

--
Pete

>
> "Collecting kernel messages"
> http://www.stardust.webpages.pl/files/handbook/handbook-en-0.3-rc1.pdf
> for more info.