On Sun, 2003-07-06 at 07:46, Peter Møller Neergaard wrote: > I have now been running Mandrake 9.1 with the 2.4.21-0.18mdk kernel > for about 3 weeks. At this point it is starting to be annoying that > this kernel locks up more often than even Micro$oft Windoze. > > The lock up will happen anything from 5 minutes to 10 hours of boot. > It must be the kernel locking up since there is no response to the > SysRq+Alt+... keys. > > This happens routinely, but irregular, so I have no idea how to track > it. I tried maximizing the information to syslog by choosing > > *.* /var/log/syslog > > in /etc/syslog.conf. A typical entry looks like this: > > Jul 6 14:42:18 pan spamd[5914]: identified spam (9.3/5.0) for turtle:501 in 0.4 > seconds, 3321 bytes. > Jul 6 14:45:00 pan CROND[5928]: (turtle) CMD (/usr/sbin/anacron -t > $HOME/bin/shell/cron/anacrontab) > Jul 6 14:45:00 pan anacron[5929]: Anacron 2.3 started on 2003-07-06 > Jul 6 14:45:00 pan anacron[5929]: Normal exit (0 jobs run) > Jul 6 14:47:12 pan syslogd 1.4.1: restart. > Jul 6 14:47:12 pan /etc/hotplug/net.agent: invoke ifplugd eth1 > Jul 6 14:47:12 pan ifplugd[1526]: Using interface eth1/00:02:2D:40:D0:92 > Jul 6 14:47:12 pan ifplugd[1526]: ETHTOOL_GLINK failed: Operation not supported > > which means that I have booted around 14:46:45. Thus there does not > appear to be any programs running just before the lock up. >
no log messages definitely says kernel -- most other stuff has a chance to squawk. However, 99.9999999% of the time the "other stuff" is X, which logs into a different file :-) Another indicator is that a crashed kernel will flash the keyboard lights. > So at this point I would like suggestions: > - how can I get more debug information from the kernel There's some info on kernel.org and www.tldp.org about getting a serial console onto a box to watch the kernel trap, but be aware that it's not likely to mean a damn thing to anyone but a kernel developer. > - should I change to a different kernel, e.g., vanilla 2.4.21. Or > should I consider one of the patched ones. > You've probably got a hardware problem. I saw James suggest memtest86, which is a good idea -- I'd also drop a Knoppix CD in or better yet boot with a LEAF disk and see what happens after a few minutes. If it barfs after running LEAF for ten minutes, it's either deeply incompatible with Linux in general or it has bad hardware problems. > Thanks > > /Peter -- Jack Coates Monkeynoodle: A Scientific Venture... http://www.monkeynoodle.org/resume.html
Want to buy your Pack or Services from MandrakeSoft? Go to http://www.mandrakestore.com