Wednesday, December 19, 2007

RAID

During the winter break, I decided to go back to doing work at CENS, and over the last two days, I've had a great deal of training with RAID systems, and I'll probably get some more over the break. For those of you who don't know, RAID stands for "Redundant Array of Inexpensive Disks". The basic purpose of RAIDs are to protect against disk failure. If you're running an array of disks, and one of the disks becomes corrupted, your data is still safe and your computer will keep running. However, you're urged to replace the hard drive, and once you do, everything will be backed up and well until another hard drive croaks.

There are several types of RAIDs. However, the type we did was a RAID1, which involves two hard drives, and one mirroring the other. That way, if one hard drive fails, the other will be an exact copy and can take over. The higher levels of RAIDs involve more complex means of restoring data such at parity and will make for an interesting discussion later on. But now, we're sticking to RAID.

The background of the project is a RAID1 on a Linux Server (A SUN server, I might add). The first part basically entailed getting used to the Linux RAID tool (mdadm) and finding out why one of the hard drives was removed from the RAID. It turned out, when you remove a hard drive, you have to manually re-add it after you put it back. And once you do that, it has to manually rebuild the mirror for an hour or so. So, removing hard drives to test out the RAID is something we quickly decided to do very rarely. The next time we did it was with the second step, getting GRUB to work on it.

The second thing turned out to be much easier than we expected. In order to get GRUB to boot off of a RAID, you just have to put a copy of GRUB on the Master Boot Sectors of BOTH hard drives. And they both have to reference (hd0), because if one is taken off, then the other will be called (hd0) because the first one isn't detected. So we didn't have to even differentiate that between the two GRUB menus. It was much easier than expected.

The third thing is a bit more private, and being done in progress. We're basically re-designing the hard drive to extreme modularity. There are entirely separate partitions of data some of which will be RAIDed, others which will not be. It's quite confusing, and I'll probably have more detail about that later on. But, I hope you enjoyed this update, and yes, I managed to put a picture in a post. Hooray for me!

No comments: