JBOD failure recovery

[budric]

Hi,
I was thinking of buying an enclosure that supports JBOD for backup. I was wondering what would happen if one drive in the enclosure fails. Do I loose all the data or can whatever is on the other drive still be recovered? Maybe if I defragment the volume to get things to line up more.

 

[chookman]

JBOD... stands for Just a Bunch Of Disks. They are basically like stapling 2 or more disks together to make one bigger one.

If you loose 1 disk in the array then whatever is on that disk is lost, however, the data that is on the other should still be able to be obtained.

Have a read here http://en.wikipedia.org/wiki/JBOD#Concatenation_.28JBOD_or_SPAN.29

Yay 1000 posts

 

[budric]

I have read that and other descriptions of JBOD. I wanted to know for sure if someone has dealt with a failure and was able (or not) to recover some of their data.

For example the wiki page says:
"The Mac OS X 10.4 implementation — called a "Concatenated Disk Set" — does not leave the user with any usable data on the remaining drives if one drive fails in a concatenated disk set"

This is for Mac OS, however it doesn't say whether this applies to JBOD implementations or not. For example would the file allocation table (or equivalent for the file system) be stored on each drive or only on the first drive? Then if the first drive fails you no longer have any information about files. And probably won't be able to recover.

These are the concrete details I was hoping to sort out. It would be more convenient to deal with 1 volume, however I'd rather have a backup where if one drive fails, I wouldn't loose everything.

 

[chookman]

Youd be better off with a Mirror or using an external disk. Mirror is not backup, however, if one drive fails data is easily recoverable.

 

[KWHobbit]

Hi there,
I just bought a D-Link DNS-343 and 3 TG Maxtor SATA drives. I had exactly the same question: What happens if one drive in a JBOD configuration fails? I called D-Link and one of their engineers told me that you should be able to recover data from the other (intact) drives. I setup the DNS-343 with 2 1TB drives in JBOD mode and copied a bunch of data to the volume (about 70 GB only). Then I switched the machine off, removed the second drive (as if it was a complete failure), turned the machine back on and waited to see whether any volume reconstruction or simply a Table of Content reconstruction would be done, but nothing. The machine simply reported that no volume can be found (as if you were doing a new configuration). I switched the machine back off, inserted another empty drive (I had bought 3) to see whether the volume or the TOC would be restored (at least partially) if another drive is detected such that the original volume (by size) could be recreated. Again, the DNS-343 reported that no volume could be found. Then, the same process again, but I reinserting the original 2. drive (the one that was originally used to create the 2-drive volume), and voila - the volume could be found again with all the 70 GB.

You can certainly do all kinds of tests - this one took some 30 Minutes and was quite instructive. I will run all drives in standard mode (so I get max capacity) and run backups externally on my old IDE drives (not what I wanted to do but hey...). On the other hand, I would give the engineer at D-Link only credit for wanting me to buy the drive (I asked before I bought it), but for either his ethics or his knowledge - well, I do not easily believe what people tell you - see, see...

Anyway, I assume that JBOD is standardized enough that this outcome would be universal (independent of manufacturer) and therefore hope my findings help you.

If anyone knows whether there are other ways to recover data given the above scenario, please let me know. If there is no way, then JBOD may as well be thrown out since RAID0 has the added benefit that the parallel disk access due to striping may speed up any volume access.

BTW: I just also ran some transfer rate tests (does it give away that I am an engineer, too). Running off a HP TC4400 with a Gigabit adapter, over a Gigabit switch (about a 100 ft of CAT5 cable in our house) to the DNS-343 and this is what I got.
Download: 75 Mbps (or 9.375 MB/sec)
Upload: 69 Mbps (or 8.625 MB/sec)
Not the world either, perhaps some tweaking may help but overall not a good transfer rate. If anyone could direct towards possible bottlenecks, any help/idea is appreciated. Thanks.

Cheers,
Klaus

 

[Black92RS]

Hello, sorry for my bad english.

I think linksys engineer meant that you could recover the data stored on the hard disk that still are perfectly readable, with recovery software and conecting it separatelly and nativelly in your comupter, not in the BYOD NAS.

If defragmentation is low little data of the available drives would be lost.

Raid 0 would meant total lost of data no matter defragmentation, as data is striped between two drives no data will be complete.

I have not tested this myself but is what I understand from the principle of JBOD working, I also think that in diferent implementations each drive has its own MBR with the data stored on each drive but that would meant that no file can be spaned between then. Only if you had enought space available in one of the drives a file could be copied, otherwise each drive must have all the MBR information of each drive and that would be non standard I think so very propietary and in usable if both hard drive and controller-hardware occurs.

Hope to been of some help, and shed some light.

 

[Guest]

I use an external USB hard drive (Freecom Hard Drive XL 2TB USB 2.0*) to backup files. This has an array of two 1TB drives connected by JBOD**.

Recently this device has stopped working so I took it apart to what data I could recover.

On inspection EASEUS Data Recovery Wizard Professional*** finds both disks and seemingly all files.

It only begins to become unstuck however when the files are recovered and many of the recovered files cannot be opened****.

I will be trying to recreate the JBOD array in the next few days and will report how succesful this is as an alternative.



Links
-----

* Freecom Hard Drive XL 2TB USB 2.0 http://www.freecom.com/ecproduct_detail.asp?ID=3888&CatID=&sCatID=

** Frage zu Hard Drive XL 2TB USB 2.0 http://forum.freecompromo.com/viewtopic.php?p=25529&sid=49c9e63f4795a6aa73aecd127987a7d9

*** Data Recovery Wizard Professional V4.3.6 http://www.easeus.com/datarecoverywizardpro/

**** I have recovered a file, but I cannot open it http://www.easeus-deletedrecovery.com/docs/faq/I-have-recovered-a-file-but-I-cannot-open-it.html

 

[fatice]

Hi there,
I just bought a D-Link DNS-343 and 3 TG Maxtor SATA drives. I had exactly the same question: What happens if one drive in a JBOD configuration fails? I called D-Link and one of their engineers told me that you should be able to recover data from the other (intact) drives. I setup the DNS-343 with 2 1TB drives in JBOD mode and copied a bunch of data to the volume (about 70 GB only). Then I switched the machine off, removed the second drive (as if it was a complete failure), turned the machine back on and waited to see whether any volume reconstruction or simply a Table of Content reconstruction would be done, but nothing. The machine simply reported that no volume can be found (as if you were doing a new configuration). I switched the machine back off, inserted another empty drive (I had bought 3) to see whether the volume or the TOC would be restored (at least partially) if another drive is detected such that the original volume (by size) could be recreated. Again, the DNS-343 reported that no volume could be found. Then, the same process again, but I reinserting the original 2. drive (the one that was originally used to create the 2-drive volume), and voila - the volume could be found again with all the 70 GB.

You can certainly do all kinds of tests - this one took some 30 Minutes and was quite instructive. I will run all drives in standard mode (so I get max capacity) and run backups externally on my old IDE drives (not what I wanted to do but hey...). On the other hand, I would give the engineer at D-Link only credit for wanting me to buy the drive (I asked before I bought it), but for either his ethics or his knowledge - well, I do not easily believe what people tell you - see, see...

Anyway, I assume that JBOD is standardized enough that this outcome would be universal (independent of manufacturer) and therefore hope my findings help you.

If anyone knows whether there are other ways to recover data given the above scenario, please let me know. If there is no way, then JBOD may as well be thrown out since RAID0 has the added benefit that the parallel disk access due to striping may speed up any volume access.

BTW: I just also ran some transfer rate tests (does it give away that I am an engineer, too). Running off a HP TC4400 with a Gigabit adapter, over a Gigabit switch (about a 100 ft of CAT5 cable in our house) to the DNS-343 and this is what I got.
Download: 75 Mbps (or 9.375 MB/sec)
Upload: 69 Mbps (or 8.625 MB/sec)
Not the world either, perhaps some tweaking may help but overall not a good transfer rate. If anyone could direct towards possible bottlenecks, any help/idea is appreciated. Thanks.

Cheers,
Klaus

Hi

Thanks for the testings. If I understand it correctly, tests are done for JBOD as:
Original setup with 70Gb data on JBOD with disk 0 and disk 1.

Test1: Take out disk 1,leaving disk0, JBOD fails.
Test2: disk 0 + disk 2, JBOD fails
Test3: disk 0 + disk 1, JBOD come back alive.

The common fact here is disk0 and we assume all the data&TOC is written into disk0 first then disk1. What if the logical JBOD volume starts writing from disk1 first ?

May I suggest to run the following tests:
Test4: disk3+disk1 ? (use a new disk to replace disk0)
Test5: disk1+disk0 (swap the slot for disk0 and disk1)

Cheers
fatice

 

[michael.stenlund]

Hi Budric,
Jbod is more of an 'experimental' setup. Please don't use it. Its used for fun and testing of drives.
Instead, buy an external (in my case SATA) docking bay, place each individual disk into it one at a time and copy the drive to a RAID10 setup. A raid 10 (raid 1 + 0) has parity+speed. This is what I'm doing. I am running 1.5 terabyte on 4x 750gigabyte RAID10 disks on a NAS, copying over the old drives to a folder on this drive called 'Archive of old HDDs' and voila! Parity + speed. When one drive failes on the RAID 10 NAS, you simply go out to the store the same day, buy a new one and swap it in. You do that while the system is running. No power down needed. If you're like me, you have an extra 750GB new drive laying at home in the static wrap in the drawer of the room. Enjoy.