Re: random system crashes

[Matthew Hodgson <matthew@arasphere.net>]

On Mon, 14 Oct 2002, Bill Rugolsky Jr. wrote:
> On Mon, Oct 14, 2002 at 08:42:13PM +0100, Matthew Hodgson wrote:
> > What did happen to Ingo?  He was remarkably helpful and active on the
> > lists right up to his last post on the 2nd of May - and then seems to have
> > disappeared off the surface of the planet.  Does anyone know if he's
> > okay?
>
> I think the short answer to that is:  busy reworking major components
> of the kernel for ultra-high-performance and functionality, as he did
> in 2.3.x.   This time it's scheduling and threading:  irq performance,
> O(1) scheduling, the kernel side of Native POSIX Threads Library (with
> glibc maintainer Ulrich Drepper doing most of the userland side)
> including thread-local storage, threading and signal semantics, futex
> and MM fixes and improvements, workqueues, etc.  [Ingo has an uncanny
> knack for adding functionality, improving performance, and reducing
> code size all at once.] Expect him to be busy until the next stable
> release ships.

Fair enough - I guess I was just worried that he might have washed his
hands of the project in between TUX apparently being renamed "Red Hat
Content Accelerator 2.2" and rumblings about the Red Hat patent
application controversy ( http://lwn.net/Articles/1251/ ) which seem to
highly feature both TUX and Ingo's name.

> A little benign neglect has been known to cause somebody else to step up
> to the plate, read the code, and do a bit of maintenance ...

Well, if Dell feels like lending me an identical specced server to my
site's production box, I'd happily try to step up to the plate and get my
hands dirty - at least to attempt to diagnose this wretched hanging bug.

For what it's worth, the problem appears not to be readily replicable on a
quad Pentium Pro Compaq Proliant 6500, but that's hardly a surprise given
the differences in architecture and the fact that so many other people
seem to be happily running TUX on various boxes without any stability
issues.

TUX certainly runs in a class of its own whilst it's functional - but this
bug renders it utterly useless both for me and seemingly many other people
- it'd be fantastic for it to be fixed.  As it stands I simply can't risk
recommending TUX for any production server whatsoever, just in case it
suffers from The Bug.

Given how similar Marco and Alex's hardware setup appears to be, perhaps a
hint at the root of the problem might be isolated by a straightforward
detailed comparison of the hardware, drivers, distribution, loaded
modules, or even the actual TUX configuration itself...?

M.