DSM-G600, DNS-3xx and NSA-220 Hack Forum

uplinkhack wrote:

Hi all,

I am currently running a RAID 1 configuration on my DNS-323.

If the NAS stops functioning, can i simply take out the HDD and plug it into my desktop which runs windows and read the data directly?

No - the NAS uses an ext2 or ext3 format which cannot be read by Windows - you may be able to get access through the use of a file system driver such as ext2ifs, however, some users have experienced difficulties using these methods.

Another question: How can i switch from RAID1 to JBOD without destroying the data in one of the HDD?

Thanks in advance!

To the best of my knowledge, you can't - you need to back up, make the change & restore.

A perhaps important point - JBOD - Just a Bunch Of Disks - is a term with different interpretations, and D-Link's interpretation is one where two disks are concatenated to form a single volume.  If your interpretation is the one where each disk remains a single entity (which D-Link calls standard volumes) then it is possible to go from RAID1 to standard volumes without destroying data.

If you have a linux system, it should be possible to connect the drives, mount the volume and gain access to the data - provided that you know how to do so.

When you make any change to the disk configuration in the DNS-323, it WILL prompt you to reformat the drives, and WILL destroy the data is you follow the prompts.  There is however, a trick, that will allow a RAID1 array to be changed to standard volumes with the formatting of one drive, leaving the data intact on the second.

With regard your mishap - there is a very high probability, that had you done the upgrade with the drives in place, you would not have ended up with a degraded array, however, the simple truth is any disruption to the function of a RAID array can cause it to become degraded with the possible loss of data.

Last edited by fordem (2010-06-09 23:14:16)

Remove either one of the two RAID1 disks and power the unit on, login and select TOOLS>RAID, set the RAID type to standard volumes and allow the unit to format the disk that is inside it.  Once this is complete, power the unit off and re-insert the disk that was removed and power it back up - the unit should at this point show you both disks as separate volumes.

Another somewhat similar question is whether it is possible to use RAID1 in a DNS-323 with 3 disks in a rotating fashion? 2 disks are in the DNS and once a week one of the disks are replaced with the 3rd one (and the 2nd taken to off-site backup storage).

If think the actual question is what happens if an up-to-date disk is placed next to an older version (that might lack some data) in a RAID1 system?

sulan wrote:

Another somewhat similar question is whether it is possible to use RAID1 in a DNS-323 with 3 disks in a rotating fashion? 2 disks are in the DNS and once a week one of the disks are replaced with the 3rd one (and the 2nd taken to off-site backup storage).

If think the actual question is what happens if an up-to-date disk is placed next to an older version (that might lack some data) in a RAID1 system?

From my experience handling the DNS-323, it seems that once a disk is removed from the drive bay, the RAID will go into degraded mode. The next time you insert a disk, it rebuilds the data (whether it rebuilds EVERYTHING or only the updated portions, i have no idea), meaning in practice your 3 disk rotation will work.

i don't understand. why would something that would work in theory fail if RAID works? what do you mean by redundancy but not backup?

uplinkhack wrote:

i don't understand. why would something that would work in theory fail if RAID works?

The reason the three disk rotation can fail in practice (specifically with the DNS-323) is that you are relying on Dlink's
utility scripts to correctly identify the "new" disk and issue the correct mdadm (http://en.wikipedia.org/wiki/Mdadm)
commands to reconstruct the RAID without losing any data.  Those RAID management scripts (not open source) do
not always do what you want them to do.  You can search this forum for numerous examples of users who lost data
when the RAID management scripts formatted the wrong disk.

uplinkhack wrote:

okay thanks for the clarification. In other words, DNS-323 is unreliable isn't it?

Well, it's far from perfect. But it does what it is supposed to do. disk rotating is not part of the intended usage.

sulan wrote:

oxygen wrote:

Well, it's far from perfect. But it does what it is supposed to do. disk rotating is not part of the intended usage.

No, but replacing a failing disk is and this is where RAID1 kicks in. So, as I see it the DNS-323 is unrealiable with regards to its implementation of RAID1. But except from that the little box seams to work just great!

Let me put it this way - as long as your data remains accessible in the event of a disk failure, the RAID1 implementation has served it's purpose - how you get the RAID array from a degraded state back to an optimum one is of no relevance - there was a time when a complete reinstallation of the OS was a requirement.

Yes - we have come to expect hot swappable disks, hot spares, automatic rebuilds, online capacity expansion, etc., etc., etc., but these are actually additional features and not required for a basic RAID implementation.

Less than a week ago I had a disk failure on an entry level server from one of the three big names in Windows based servers, the disk failed, the server alerted the administrator and at the end of the day I sent a tech to replace it, he powered the server down opened it up & removed & replaced the failed drive (a 250GB SATA) and powered it on, and expected it to start a rebuild, because that's what normally happens when he replaces a drive on a larger server.  When it didn't he called in and had to be talked through going in to the RAID Management Software utility and selecting the drive and starting a rebuild.

I'm talking about a piece of hardware designed for use in a small business environment or for small branch offices in a larger corporate climate - and that doesn't offer any of those additional features either.

If the DNS-323 allowed or required the user to select a drive and initiate a rebuild there might be fewer complaints about lost data, but would the users know how to deal with it?

Okay, I understand. Good points.

what about the thing about RAID being redundancy and not backup? seems the same to me..

that's a very clear explanation. thank you very much!