[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

RE: Beware of current RHEL4 kernel

On Wed, 2008-07-16 at 10:49 -0400, Clark, Patti wrote:
> We are running Oracle OEM on a Dell 1850 - a single purpose box.  That
> last kernel produced kernel panics within hours of each reboot.  We have
> gone back to the previous kernel release.  
> I have the new kernel on all of my other boxes without issue.  All of
> them are Dell platforms of varying flavors and applications.

In the end we had 3 systems out of 5 that crashed in less than 24 hours
(one did not make 6 hours).  All were at lead 2 CPU systems, and all
three were running Oracle.  Perhaps Oracle is a prime tickler for this
bug, however, Redhat support says the problem call happen on any system
with more than two threaded applications and a high load.  Several
people have reported it on MySQL servers and with clamd.

I'm not convinced it doesn't sometimes cause strange application
failures even if the kernel doesn't panic.  Even the systems that didn't
panic had some very strange behavior before we reverted.

Redhat support officially says I should just run 2.6.9-67.0.15 for now
as "...there are some know issues with this kernel and engineering is
aware of the issue and are currently working on it."  Gee thanks for
keeping it out there so that your customers can continue to be at risk.

Also, there's a reproducer in the Bugzilla that will pretty much panic
any system running 2.6.9-67.0.20.  That seems like a security issue to
me as it'd be pretty easy to DoS any system running that kernel with the
very simple program in the Bugzilla.  I really don't understand how
Redhat cannot pull a kernel with such known broken behavior, or at a
minimum, rush out the fixed kernel.  It also calls in to question
Redhat's QA testing.  How could a kernel with such a major issue ever
get past even minimal QA testing?


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]