Re: XEON SMP KERNEL 2.4.20 tree
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi again,
How long have those machines been online?
$ uptime
11:14:51 up 8 days, 21:46, 1 user, load average: 0.00, 0.03, 0.00
$ uname -a
Linux data 2.4.21-rc6 #1 SMP Tue Jun 3 12:46:45 CEST 2003 i686 unknown
There is a discussion about the 2.4.21rc kernels on the SE Linux
mailing list.
Machines running 2.4.21rc kernels seem to have a kernel memory leak
which
makes them unusably slow. Opinion is divided on whether this is a
kernel
issue or a SE Linux issue. I don't believe that it is SE Linux at
fault as
the SE code we are using is a straight port from 2.4.20 (and should run
equally well as on 2.4.20). However other people have seen some
evidence to
suggest that the problem may only occur on SE Linux machines (I don't
run any
non-SE machines at the moment for testing).
With the 2.4.20 kernel, this machine froze up to 2 times per day. As
you can see, its now
operational for more than a week and doesn't have any problems. About
the performance, I cannot make an informed comment, since this is our
NFS-Server and the machine performs as well
as it did before. So .. for me, it does do as well as before.
Ok .. just for a quick comparison I compiled a stock 2.4.20 kernel on 2
machines (one with 2.4.19, one with 2.4.21rc6 kernels running). The
machines are xSeries 340 DP with 1 GHz CPUs, 1 GIG of RAM and 36 GIG
SCSI HDs, onboard AIC controller.
$ time make -j3 bzImage
Machine running 2.4.19
real 2m18.895s
user 4m2.760s
sys 0m16.810s
Machine running 2.4.21rc6
real 2m18.842s
user 4m12.920s
sys 0m16.370s
So .. again .. I cannot say that the performance has changed in any way.
- - Cheers, Peter
- --
Dipl.-Ing. Peter Burgstaller
Technical Director
@ all information network & services gmbh
email: peter.burgstaller@ains.at
phone: +43 662 452335
fax : +43 662 452335 90
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.0 (Darwin)
iEYEARECAAYFAj7oW64ACgkQezyUhHKdNXRCRACeKTgWhHV8fp66BAyYjzGAjuGL
ZyIAnA/ENUIgCZ73eT1V2iXMDKtN0H03
=dQML
-----END PGP SIGNATURE-----
Reply to: