-2

I've got a couple of disks lying around which are small in capacity and old enough to be considered spinning spinning rust, my gut instinct is to not trust them, but the damn things still work. They're a collection of 320Gb and 500Gb disks I've got running in a server, that also has some newer (more reliable), larger disks (2Tb+).

I'd like to use the smaller, unreliable disks to create a some sort of usable Linux filesystem that I can treat as "scratch space" - for low value data that I don't particularly care about losing. Ideally, a filesystem that simply looses files if a disk is lost - I don't want the hassle of rebuilding a RAID array when I lose disks.

mfinni
  • 36,144
  • 4
  • 53
  • 86
xconspirisist
  • 157
  • 8
  • 6
    *"I know this water has dirt, mud and animal feces in it, and will likely make me really, really sick with explosive diarrhea... but I'm thirsty and don't care about the consequences... What type of straw should I use?*" – ewwhite Nov 02 '14 at 15:48
  • @ewwhite You misread the question, he doesn't care if he gets sick with explosive 'diarrhea'. – captcha Nov 02 '14 at 21:13
  • If this question is not appropriate to SF maybe it worth to check over there : https://www.reddit.com/r/filesystems/comments/4caoln/what_filesystem_would_you_use_on_a_faulty/ – rzr Mar 28 '16 at 21:25

1 Answers1

3

I can think of one option off the top of my head, and another one turned up while looking.

First and foremost you generally do not want this on a production system. I used to do something similar on a drive I wanted to kill, but that was for personal use.

Secondly, since drive reliability usually follows a bell curve, I'd use SMART to decide which drives are closer to dying, not arbitrary things like age and capacity.

The first option that comes to mind is greyhole - which I've always considered odd. It uses samba to allow you to use multiple drives on a single mountpoint. It also has the option of redundancy should you want it. If a drive dies, you can just replace it, or remove it from the pool, your remount the remaining drives standalone. It does however suck in some scenarios by the developer's own admission.

Another alternative I came across is MHDDFSwhich does the same sort of jbod concatanation as a FUSE filesystem. Its apparently a tiny bit slower than greyhole, dosen't let you set redundant copies, but might work better with lots of small files.

In both these cases, underlying file systems matter less, and you can always recover data per drive.

Journeyman Geek
  • 6,977
  • 3
  • 32
  • 50