[plug] kernel 2.6.3 production-ready?

[Denis Brown]

Dear Tony,

On Thu, 4 Mar 2004, Tony Breeds wrote:

> On Thu, Mar 04, 2004 at 12:11:20AM +0800, Denis Brown wrote:
> > Dear PLUG list members,
> >
> > Is kernel 2.6.3 worth betting the farm on, stability-wise, to get my hands
> > on ACLs?   AFAICS, ACLs are natively in the 2.6 tree.   Alernatively I can
> > patch a 2.4 kernel with the stuff from bestbits.
>
> If this is /all/ you want then the 2.4 kernel may be the better option
> for you.  Having said that I run 2.6 on 2 production servers and I have
> no problems at all.  I went from 2.4.22 -> 2.6.1 (haven't scheduled the
> reboot time for the upgrade to .3) and I noticed things seemed more
> snappy.
>
Thanks for the comments.  "Snappy" would be good but I might wait for a
few more releases in the 2.6 tree.   I might even be tempted to look at
LVM then, too :-)

> One of the slightly annoying problems with the upgrade to 2.6 is the
> need for a new/different set of mudule tools.  To the best of my
> knowledge they're not packaged in Debian stable (the only choice for a
> UWA server ;P)
>
Agreed :-)

> and they (module-init-tools) tend to create package
> managemtn problems for modutils.   This is by no means a showstopper
> problem but it another reason you may not want to go down the 2.6 path
> (yet).
>
I haven't looked but something may be on backports?

> In short I think 2.6 is production ready for fairly std. server BUT
> distribution packageing may make it just a teansy-bit too painful to
> switch to now.
>
Like most people I guess, I have met the Debian "we'll release it when
we're darned good and ready to" syndrome and have so far worked around it
thus:   critical system stuff I leave well alone, to be managed by
apt-get.   Applications such as Apache, PHP, PostgreSQL I compile from
source.   So far it has worked pretty well.   The only challenge I have
had (am having, more accurately) is with some heavyweight neuroanatomy
imaging software[1].   For that, it looks as though I will have to compile
my own gcc3.3.3 and support tools, possibly also the latest glibc.
Debian's gcc-2.95.4 (and gcc-3.0.4) and friends has been reported to give
problems which may explain the anomalies I'm seeing.   A few palpitatons
there, but working my way through it.

[1]  a combination of graphics and analytical packages glued together
with Tcl/Tk; uses named pipes to communicate between the various
applications.   And other than documentation at the operator level (which
is brilliant) the programme doco is really "in the source."   You can
probably imagine the debugging effort required to determine where
interesting things like segfaults arise :-)   And of course it "compiles
and runs pefectly on RH and SGI"   Aaaaarrrrggghhhhh...

Cheers,
Denis