Stuart Herbert

I’ve been asked how I’m getting on with my personal project to build my new fileserver. I figured that, instead of just replying to James directly, I might as well add an entry to my blog’s category on this.

To be honest, I haven’t touched the fileserver project since my last blog update. The fileserver’s sat under my desk, powered down, until I have a bit more free time to devote to it. I’ve been sidetracked (from Gentoo work as well ) with providing consultancy and some programming to help a firm migrate one of their in-house systems to a Gentoo-based solution. That project should be coming to a successful conclusion in the next few weeks, and that should allow me time to get back to my fileserver work.

The next steps for me are to

1. Upgrade to kernel 2.6.15, and try the SATA chipset driver now included. At the moment, the fileserver uses a driver that I modified to work with later 2.6 kernels. It’s not that I don’t trust my data to my own programming skills, but, well frankly, I don’t I do kernel hacking so rarely that it’d be all too easy for me to have made some mistake that I wouldn’t spot until all my data disappeared.
2. Use the dual-xeon box as a compile host for the fileserver. The EPIA 1300SP is fast enough to make an adequate desktop box, but it’s not well-suited for large compilations.
3. Rebuild the box from scratch, using Xen 3.0 and Modular-X. Although the box is going to be a fileserver first and foremost, most of the time it’s going to be sat around with the CPU idling. I also have the small matter of the 200GB RAID-1 array dedicated to the O/S that I’m never going to fill just with Samba. So I’m going to evaluate Xen 3 on the box, and assuming that the performance isn’t an issue, I’ll eventually add other roles for the server, such adding my IMAP server.
4. Evaluate and select a package for managing backups. I’m not going to write a backup solution myself; there must be plenty of suitable open packages out there. I need to pick the one that will best fit the fileserver project; preferably one that I can integrate into a web-based management console.

I still need to sit down, play with the HAL bindings for Python, and put together some utilities to automatically assess and re-configure drives when they are hot-swapped. I’m repeatedly reminded that the kernel’s SATA hot-swap code isn’t 100% solid yet, so I’m assuming I’ll have to look into whether anything can be done about that at some point. The box will need a web-based management interface, and some solid server monitoring and alerting tools (because out of sight, out of mind leads to dead fileservers and lost data). I’ll need to add support for additional backups to DVD, and to USB external drives. I’d like to add support for my digital cameras, to save the images having to travel over the LAN just to be downloaded to the fileserver.

There’s a lot to do before I’ll actually be able to store any real data on this puppy, and scant time to spend on it.