Welcome to the MacNN Forums.

Tyler McAdams · Feb 15, 2006, 10:37 PM

If you lose a disk on a raid set that uses OS X Concatenation are the data files on the other disks recoverable by simply remounting them?

In other words, if you had a bunch of mp3's on the disks, and since these disks are not striped it would seem to reason that the mp3's would behave as regular file formats on the disk and thus recoverable by simply mounting the disk or the remaining disk in the concatenation set as a different volume.

Detrius · Feb 17, 2006, 12:11 AM

Originally Posted by Tyler McAdams

If you lose a disk on a raid set that uses OS X Concatenation are the data files on the other disks recoverable by simply remounting them?

In other words, if you had a bunch of mp3's on the disks, and since these disks are not striped it would seem to reason that the mp3's would behave as regular file formats on the disk and thus recoverable by simply mounting the disk or the remaining disk in the concatenation set as a different volume.

Are you talking about mirroring? If so, you wouldn't need to "recover" anything. One of the disks in the mirror failed, but the other survived, so your mirror is still valid. You replace the one that went bad and restore the mirror to the new drive.

OreoCookie · Feb 17, 2006, 04:48 AM

No, I think he's talking about what is called jbod in the pc world: just a bunch of disks: when one disk ends, the OS switches to the next one. You don't get any performance improvements over a single drive.

I have read that you can recover files in the way you propose, at least in principle. Although I have never used a concatenated volume, much less recovered from it. But to be honest, I wouldn't be surprised if all your data were lost.

Tyler McAdams · Feb 17, 2006, 09:58 AM

Originally Posted by OreoCookie

No, I think he's talking about what is called jbod in the pc world: just a bunch of disks: when one disk ends, the OS switches to the next one. You don't get any performance improvements over a single drive.

I have read that you can recover files in the way you propose, at least in principle. Although I have never used a concatenated volume, much less recovered from it. But to be honest, I wouldn't be surprised if all your data were lost.

That's what I'm talking about. It's a new option under the OS X software raid selections. It differs from a striped set in that the files are not split up between the disks in order to gain a performance enhancement. Instead, it just joins the disks together in to one big volume. (Similar to how a LVM scheme would "grow" a disk by adding a new volume to a predefined set)

In theory it would be less useful I feel if you could not recover data by remounting the remaining volumes from a failed set... if one or two volumes in, say a 5 disk set failed for some reason. In a srtiped set you would be screwed, but in a concatenated set it would stand to reason you could recover the data simply because it's not split up between the disks in the set but only written to a single disk. When that disk gets full it would move to the next disk.

It would be nice to have some sort of "built in recovery tool" for recovering a concatenated set in the event that such an error occoured. I would imagine more advanced disk managment systems such as VxVM allow for this situation. (But I see nothing about that in the disk utility so I'm guessing you could either "jimmy rig" it or your screwed)

Detrius · Feb 18, 2006, 01:45 PM

I had always heard that you are screwed, even with JBOD.

Without a directory structure, there's no way of telling where the files are in order to recover them. So, without a directory structure, it would be no different from when Norton Utilities hosed your volume, so you had to do a recovery, and only a few files survived.

So, in theory, if there were some sort of directory structure on each drive, it would be plausible that data would be recoverable.

littlejohn · Feb 19, 2006, 05:24 AM

I'm hisyorically a PC user... but I believe the raid technology and algorithms are the same no matter which OS you use. If you JBOD array dies, you're stuffed, unless you're keen on sending your drives to professional data recovery mob.
The crux of it is, that you've lost your partition information and therefore your dives can no longer be mounted. Thats the reason no one configures RAID in JBOD, it just increases the likelyhood if failure with no real gain.
Just curious... why would you want to setup your RAID in JBOD??

OreoCookie · Feb 19, 2006, 07:10 AM

Originally Posted by littlejohn

Just curious... why would you want to setup your RAID in JBOD??

It was used to concatenate disks for larger RAID5 arrays. But we're talking 30+ disks here. Note the date of the article, so I don't think anybody still uses it this way. At my university, we have configured a RAID5 on our SunFire.

Anybody who uses a RAID (or pseudo-RAID) mode which increases the risk of failure should either have a good backup strategy or data which is not valuable to him/her on this volume.

littlejohn · Feb 19, 2006, 07:27 AM

Originally Posted by OreoCookie

It was used to concatenate disks for larger RAID5 arrays. But we're talking 30+ disks here. Note the date of the article, so I don't think anybody still uses it this way. At my university, we have configured a RAID5 on our SunFire.

Anybody who uses a RAID (or pseudo-RAID) mode which increases the risk of failure should either have a good backup strategy or data which is not valuable to him/her on this volume.

When you concatenate raid 5 it's not longer JBOD.. it's Raid 50 or 5+0.. ie you can afford to have one drive fail from each RAID 5 subset. so if one drive fails, you still have redunancy. There is no redunancy in JBOD so if on drive fails, the whole array fails...
There a lots of scenarios where admins would use RAID 50 especially in larger SAN arrarys but doing JBOD is a poor mans way of using getting a larger volume from a bunch of smaller drivers.. but the risk is not worth it IMO..

OreoCookie · Feb 19, 2006, 08:44 AM

Originally Posted by littlejohn

When you concatenate raid 5 it's not longer JBOD.. it's Raid 50 or 5+0.. ie you can afford to have one drive fail from each RAID 5 subset. so if one drive fails, you still have redunancy. There is no redunancy in JBOD so if on drive fails, the whole array fails...
There a lots of scenarios where admins would use RAID 50 especially in larger SAN arrarys but doing JBOD is a poor mans way of using getting a larger volume from a bunch of smaller drivers.. but the risk is not worth it IMO..

No, you misunderstand. Just take a look at the link I've given you.

In one example, they jbod' 30 disks in 10 triples and then create a RAID5 on those 10 triples. That's not a RAID50 (I am familiar with the different RAID levels and their combinations). The redundancy comes into play because one triple may fail. As I have said, afaik this is not used anymore, or at least I haven't heard of it.

But again, you asked if somebody has ever used it professionally, and the answer is apparently yes.

littlejohn · Feb 19, 2006, 05:47 PM

Originally Posted by OreoCookie

No, you misunderstand. Just take a look at the link I've given you.

In one example, they jbod' 30 disks in 10 triples and then create a RAID5 on those 10 triples. That's not a RAID50 (I am familiar with the different RAID levels and their combinations). The redundancy comes into play because one triple may fail. As I have said, afaik this is not used anymore, or at least I haven't heard of it.

But again, you asked if somebody has ever used it professionally, and the answer is apparently yes.

Ahh,, yes.. my bad.. must learn to read link more carefully.. I my orginal Q was aimed at the thread starter..