MythTV storage issues: Update

I posted this entry back in November explaining my thinking behind what I planned to use for storage on the MythTV box. This post is about my latest, updated, thinking on the matter.

As you may recall, my goal was to start out with a 1 terabyte RAID5 array handled by LVM. The RAID5 is for redundancy in case of drive failure, and the LVM was for ease of future management and growth.

Since then, I've decided that it would be wise to start smaller and then grow when necessary. Also, I've been doing some more reading and have changed my thinking somewhat about how it should be configured.

I've seen it suggested a few times to make smaller RAID arrays and group them together using LVM. For example (and a good example at that, since this is actually what I've decided to do), if I bought 2 300 gig drives, the first thought may be to make them into one 300 gig RAID1 array. Then, if you wanted to grow your storage space, you'd have to buy 2 new drives and make another RAID1 array and then add it to the logical volume. The suggestion that I've been reading is to split it up into, say, 3 100 gig RAID1 arrays (grouped using LVM, of course). It took me a while to realize the benefit in this. The benefit is flexibility. With smaller RAID arrays, it is easier to shuffle them around in the future.

Maybe just laying out what I'm planning in detail will help.

For now, I'm going to buy 2 300 gig drives (hda and hdb). I will partition each into 3 100 gig partitions (hdx0, hdx1, hdx2) which are then configured as 3 100 gig RAID1 arrays (md0, md1, md2). I've attempted to illustrate this in text below:

hda0 + hdb0 = md0
hda1 + hdb1 = md1
hda2 + hdb2 = md1

The total space I'll have available with this setup is 300 gigs.

Then, in the future, when I need more space, I can just buy one new drive and reshuffle. So, let's say I buy a 500 gig drive (hdc). I would then partition it into 5 100 gig partitions (hdc0, hdc1, etc.). Then what I would do, is (basically) break the existing RAID1 arrays by removing the partitions on hdb from the arrays. This wouldn't result in data loss because the data is mirrored on both drives, so losing one drive doesn't cause loss.

The next step would be to rebuild those RAID arrays using 3 of the new partitions on hdc. Basically, at this point it would like this:

hda0 + hdc0 = md0
hda1 + hdc1 = md1
hda2 + hdc2 = md2

That leaves me with 3 partitions on hdb (one of the original 300 gig drives) and 2 partitions on hdc (new 500 gig drive). So, I'd just make 2 new arrays like this:

hdb0 + hdc3 = md3
hdb1 + hdc4 = md4

After extending the logical volume onto the new arrays, I'd then have 500 gigs of usuable, RAIDed, space. I'd also have a 100 gig partition left unused (hdb2).

I can continue doing this as I need more space. After doing this twice, though, I'll have to start retiring drives. This shouldn't be any harder than what I've discussed above, though. The only difference is that when I break the RAID arrays, the drive that was removed from the arrays would also be physically removed and replaced with the bigger drive.