Greetings all.

An incident today has made me started thinking about kernels and the like.

Sometimes if you gracelessly reboot a server whether it's due to power failure, or someone hitting the big white switch, or a rare bug causing a kernel panic, or some hardware cruftiness - the system might not come up again. The e2fsck program might discover serious enough filesystem inconsistencies that the system drops into 'disk maintenance mode' requiring fsck to be run manually.

There is a solution to this - journalled filesystems, in this case, ext3 (the Linux systems currently run ext2). The thing is the RedHat 7.1 kernel doesn't support ext3 as it stands.

The trouble with servers in remote data centers is if the thing croaks for whatever reason, and gets into this state, you have to get someone on the console, tell them your root password, and have them fix it. Major pain. Especially if this happens when you're away!

To cut to the chase, what I want to do is this:
- build a new kernel, probably 2.4.16 - with ext3 support as a static part of the kernel (not a module) so all filesystems are ext3
- also compile in iptables support and remove ipchains altogether as it's obsolete.
- Install the kernel.
- Reboot. Pray.
- upgrade my filesystems to ext3 (not really a big deal - all you're doing is adding journaling to ext2)

I'm willing to be a guinea-pig for this process, and supply a working >2.4.16 kernel with these features to other RackShackers who have the Duron setup. Howerver, I /may/ need some support from RackShack when doing this. There is a well known phenomenon - the probability of a newly compiled kernel booting is inversely proportional to your distance from and/or ease of access to the server's console!

I would delay this until RackShack are in the new datacenter, and wait for a time when tech support is quiet. If it all goes horribly pear-shaped though, I may need to guide the support dude a little to get the old kernel rebooted. Preferably, I'd like to find out what was displayed on the console if it all dies horribly.

I've built plenty of kernels from source in the past (my Debian systems are all upgraded to 2.4.13, which is the last one I got) so I'm by no means a stranger to the process...and all my kernels have booted so far. However, it would be reassuring to know for this little project that I could talk to someone at the datacenter itself if things don't work out as they should.

Just make sure you keep the old kernel in lilo and we'll get you back up in case.