Re: [gentoo-server] Testing and Installing New Kernel Remotely

[Dan Podeanu]

Hello Karl,

My personal experience (with lilo, you'll see why its important) says
that best practice is to install the new kernel with a different bootloader
label, leave the default to the working kernel and then run "lilo -R
<label of the new kernel>". This basicly means that only the next boot
will be done using the new kernel, subsequent boots will revert to the
default (that is, the working kernel).

This way, if it doesn't boot you can just get the colo staff to reboot your
server and you're up with the old, working kernel.

Until now I've never managed to leave a server completely unusable,
even while doing crazy stuff such as installing gentoo on the swap
partition of a linux running suse ;)

Of course, other ideas are more than welcome ;)

Cheers,
Dan.

> ----- Original Message -----
> From: "Karl Zander" <gentoolist@××××××××××××.com>
> To: <gentoo-server@l.g.o>
> Sent: Monday, February 09, 2004 12:59 AM
> Subject: [gentoo-server] Testing and Installing New Kernel Remotely
>
>
> >
> > Server will be at a colo site and run Gentoo.  I am tying to think
through
> > how to handle kernel updates.  Compiling is fine.  But after copying the
> > new kernel to /boot and configuring Grub, do I just re-boot and hope for
> > the best?    When I am at the console I can choose the kernel to load.
I
> > can then test the new kernel before setting it as the default in Grub.
If
> > the new kernel has a problem, I  restart the server  and choose a known
> > working kernel.  So far the kernels I have done have worked fine.  But I
> > have to think that sometime in the future I will manage to mess up my
> > configuration.  Mistakes happen.  Its not like I will recompile the
kernel
> > every week.  But from time to time I will want to take advantage of
fixes
> > or security patches.
> >
> > Still new to Gentoo so was wondering if there are any best practices for
> this.
> >
> > -Karl
> >
> >
>