Setting up a reliable server
2005-07-11 13:43:45
Usually when I set up a new linux computer system, my goal is to make the cheapest thing possible, and set it up as simply as it can be. The main idea is that if the hard dies or something, it's easy to just replace the dead machine with an identical clone in a short period of time. Of course, administrative scripts, recipes, and BACKUPS are key.
I am starting to put together a server for my wife's business. For various reasons, the cheap throwaway strategy won't work in this case. Part of it is the business cost of any downtime, part of it is the fact that there is no on-site IT staff (i.e. me).
So I'm actually putting together a highly available, reliable server. I'll document how it comes together here.
For starters, I bought a semi-fancy Dell 2800 series tower server. It has dual CPUs, dual power supplies, dual network interfaces, 2GB RAM. It has a split SCSI backplane with hardware RAID 1. I've had it configured so there are three logical drives, each of which are made out of two physical disks.
Logical disk 0 is the operating system. I'm starting with Arch Linux, which is a nice distribution that uses a debian-style package management system called "pacman". It's nice because it's all command-line based.
Logical disk 1 is the application software and database disk pair. It's 75 GB logical disk, which should be 10X what we actually need.
Logical disk 2 is a 300GB storage unit. The application we will be installing has document management and imaging capabilities, and this is where that data will be kept.
Finally, I have a dual backup strategy planned. First, there is a big, fancy, painfully expensive digital tape system that as a 200GB capacity. Then, I have a cheapo server that I am building, which is a $399 low-end dell server with 4 big SATA disks. This will give me over 1TB of network disk for over-the-network backups at a total cost of less than $800.