Re: U3 broken kernel

[Tom Sightler <ttsig tuxyturvy com>]

On Thu, 2004-12-23 at 12:04 -0500, Brian Long wrote:
> With regards to Bugzilla 132639, we just release U3 internally on
> December 6 without knowing about this bug.  We cannot release U4 without
> complete testing of various applications, but we're considering running
> the U4 kernel on our U3 systems.

So you're willing to change the kernel without testing various
applications but not the rest of the system.  This feels odd to me.  I
supposed that's probably because we have a lot of systems that run
Oracle and the kernel is far more important that most of the supporting
cast.

With SUSE Enterprise 8 just changing the kernel between -215 and -251
caused an unstable system with Oracle, while the actual -215 and -251
kernels were fine.

I've had similar experiences with Redhat.  Systems that ran fine with
2.4.21-4 (The original RHEL3 kernel I think) went swap wacky with
2.4.21-9.  2.4.21-15.x kernels seemed to mostly fix that problem, but
then 2.4.21-20 went wacky on simple things like big rsyncs backups, or
just simultaneous copies.

In other words, I'd think you would need to test your applications just
as much during a kernel change as a full update.

> Any thoughts about using the U4 kernel on a system that's primarily U3?

I don't think there would be any issue here if that is what you want to
do.  I've been running the U4-beta kernel (2.4.21-25.EL) on several
systems that are otherwise U3 and had no real issues.

Later,
Tom