Steve, having tried again with the SMP kernel on a worker node, I now
recall what the problem was. It wasn't really a problem with the
+update.lilokernel line. But with the SMP kernel, I got kernel startup
messages like "IRQ probe failed", and the kernel could not see /dev/hda at
all.
This time I used my /sbin/lilo -C /etc/lilo.conf.BHAM method
and included both smp and non-smp lines in /etc/lilo.conf.BHAM
so I've been able to recover OK and reboot that worker into
uni-processor mode, without starting from scratch, as I would have
had to if I'd used the
+update.lilokernel /boot/vmlinuz-2.2.19-6.2.12smp
line in site-cfg, as of course that supports the one kernel at boot.
Using a differently named lilo.conf avoids the problem of LCFG
over-writing lilo.conf and running /sbin/lilo.
Maybe I shouldn't be too surprised that 6.2 SMP kernel doesn't work
on dual P4 Xeons, though! (7.3 SMP works fine on my non-grid nodes).
Roll on 7.3!
Lawrence.
|