Hi,

Anyone here ever recompiled your Debian kernel for a SuperRAID server?When I got the server, it was setup with 2.4.18 but unfortunately without HighMem support.

So I added the option and recompiled. After running lilo on it, rebooting resulted in a Kernel panic(as reported by tech support). So had to get them to reboot back using the original kernel.

I used http://newbiedoc.sourceforge.net/system/ke...kernel-pkg.html as a guide.

edmundt,

Your going to have to talk to support and find out exactly what raid contorller is in the box. Then you need to compile that into the kernel NOT as a module. It has to be part of the kernel or you can't load anything from disk.

Frank

We use 3ware Escalade 7xxx series RAID cards. May I have your ticket number?

Hi Kevin,

Thanks, but finally managed to get it resolved by having tech support do a reboot for me each time I tried recompiling the kernel. Luckily, only took 2 reboots and some "hands & eyes" from them to identify the problem. Turns out it wasn't the RAID after all. It was that dang Intel Gigabit NIC(e1000) which required me to compile e1000 support into the kernel.

Strangely, the default install was loading the e1000 as a module. But try as I might, couldn't get my new kernels to work with it as a module. In the end, upgraded to the 2.4.24 kernel which has built-in support for it

My guess is that SM has a copy of e1000.o handy which they can dump into the /lib/modules/2.4.18 folder for each of their Debian install...oh yah, that and a change in modules.conf to add an alias for eth0 to e1000

edmundt,

Debian is a bit of a pain in that way. It's nice and "stable" but that also means that they are very slow to adopt "newer" technology. You have to use "newer" with a big rain of salt, as the e1000's have been out for quite a while.

I've had the exact same issue with Debian and e1000's.

Frank

Yeah, sometimes that word "stable" can be a mental crutch. Finally gave up on trying to have a 100% "stable" system and decided to use a newer kernel from the "testing" branch so that life can go on. Having said that, I'm now crazily apt-get'ing other goodies from testing....gcc 2.9.5 from stable branch is just too "stable" for me...gak!

Same goes for kernel-package. I kept wondering why "make-kpkg --add-patches" had no effect until I realized that most of the instructions on the web are assuming I'm using the newer version that didn't require me to explicitly set an environment variable to get it working *sigh*

But all in all, I kinda enjoy moving away from RedHat and trying out a new distro

beware when running testing on servers. The biggest problem with testing is that there is no security updates for testing. When there is a security patch, its first put into security updates on stable, then a new package is released for unstable. Its not until this new unstable package "graduates" to testing that testing gets the security patch. This can take weeks.

See http://www.backports.org/

Very handy site. Updated versions of a good number of packages, in .deb binary form, designed to be installed into Stable/Woody.

Just dpkg -i <backport package> and its ready.

Ferret, that's definitely a handy tip for a Debian newbie like me
Now, if only you had told me earlier....oh well...better late than never...hehehe

Is there a way to list out which installed packages are from "Testing"?

Not sure. I haven't used it THAT heavily. The main things I've pulled off backports was an updated procps package (The top command), and the 2.6.4 kernel source and headers and tools, etc.