Unfortunately no one can be told what fun_plug is - you have to see it for yourself.
You are not logged in.
Update:
Looks like some of my directory woes are related to directories which were using unicode (chinese characters) as my itunes was set to "keep music folder organized" and pulled unicode from the mp3 extended data to setup a directory structure. It is still a mystery how the file system was still able to access and manipulate these directories for a while and then suddenly fail to do so.
Without firware 1.04 released it appears that unicode is not fully supported in 1.03. Could be a smoking gun. Regardless I will restage the box and ensure all my unicode is stricken from my files/directories and retry.
Offline
Aidan wrote:
Folks,
My 323 is running firmware version 1.02 with two Samsung 320Gb drives in RAID 1 configuration.
I recently had a drive failure. I replaced the drive and the DNS323 detected the new drive, formatted it and resynced the data. Everything going good, or so I thought.
I originally had a data directory in the root which contained two sub-directories for documents and templates. After the resync, the data directory is there, but the documents and templates sub-directories are missing. When I checked each drive individually in the 323, neither of them showed the missing sub-directories. Both drives are still showing the correct amount of used space - about 26Gb.
I tried to create the templates sub-directory and got an error back saying that I can't create a directory as one of the same name already exists.
I have installed Fonz's fun plug and telnetted into the 323. When I cd to the data directory and run ls, I get a listing back showing the two 'missing' sub-directories. When I try to cd into either of them I get an input/output error. I tried chmod to 777 on both but it came back with an error that the directory doesn't exist.
I've read through other posts in this forum and it appears that I'm not the only one who has this sort of problem. Can anyone shed any light on how to get my data back? I do have some of it backed up but there is still quite a lot that I would really like to get back.
At present I have removed one of the drives to ensure that I don't end up any worse off regarding my data. As such, since the DNS is running with just one drive in it, can I run fsck while I am telnetted into the drive?
Sorry if this seems stupid, but I am new to Linux and still very much finding my feet.
Thanks in advance
Aidan.
Hi Aidan, sorry the discussion got sidetracked and no one really actually answered your problem!
Some questions that will help:
- How did you know you had a drive failure?
- Are you sure you replaced the correct drive?
- Did you attempt to run e2fsck on the drive while telnetted in, this will attempt to fix file system errors.
- Is it possible that the the directories went missing when the drive failed, and were actually part of the failure. ie Are you sure that the directories were present before you inserted the replacement drive?
Mark
Offline
Hi Mark,
To Answer your questions:
1. On power up, there was a strange clicking noise - sounded like a head motor failure. Spoke to Samsung and got it replaced under warranty. Also there was no data being written to this drive - no activity on the LED even when the other drive was being written to.
2. Yes.
3. After posting this query, I did some further checking and found posts saying that you should not run FSCK while a drive is mounted. As far as I can make out, the drives are mounted on boot up so when I telnetted in, I presumed that I would not be able to run FSCK. So in short, no I didn't run it.
4. The directories were definitely present before I replaced the failed drive as I had the "My Documents" link in XP redirected to the DNS-323 documents directory and was able to read and write to the drive.
Cheers,
Aidan.
Offline
Aidan wrote:
Hi Mark,
To Answer your questions:
1. On power up, there was a strange clicking noise - sounded like a head motor failure. Spoke to Samsung and got it replaced under warranty. Also there was no data being written to this drive - no activity on the LED even when the other drive was being written to.
2. Yes.
3. After posting this query, I did some further checking and found posts saying that you should not run FSCK while a drive is mounted. As far as I can make out, the drives are mounted on boot up so when I telnetted in, I presumed that I would not be able to run FSCK. So in short, no I didn't run it.
4. The directories were definitely present before I replaced the failed drive as I had the "My Documents" link in XP redirected to the DNS-323 documents directory and was able to read and write to the drive.
Cheers,
Aidan.
OK cool. Interesting that you don't mention the DNS323 telling you a drive had failed; possibly it was a total failure so the DNS323 was unaware of the drive even being there. A few users have mentioned the device's 'bad drive detection' doesn't work very well, so your information seems to back this up too.
I've not actually tried this, but it should be possible to unmount the remaining drive and then run the file system check. From what I understand, the DNS323 system boots from a RAM disk and then mounts the drives. Once you've telnetted in it should be possible to run an unmount command, and then fsck. There is also another command e2fsck. Not sure if they're installed in the base system or if they need the fun_plug, but from a few posts it looks like it should be possible to try this if it ever happens again.
I've also found the device a bit flaky. I wish they would hurry up with the firmware update. I don't care about the media server etc, I just want reliable device functioning!
Mark
Offline
This thread is very informative.
Have just taken my DNS 323 out of the box and put in 2 x 500GB Samsung. Will have a fiddle and see if I can understand the limitations of this unit. There does seem some real concerns about the reliability of raid1 on this unit and whether it does provide redundancy.
The ability of the unit to detect a "failed" drive seems to be questionable. Can we define what a "failed " drive should be for raid1 and decide if the dns 323 is able to reliably detect this through its amber light and email function and indication of status on web GUI.
Comments on this forum would suggest that there is some question about the units ability to detect a failed drive and also the web interface is no reliable indication of the raid status.
So if the unit does not detect a failed HDD and there is no way of knowing the raid status, won't the unit happily chug along until the second drive falls over.
There is lots of confusion out there about the expectation of raid1 but it does not help with comments like:
Realistically we expect most people to put the DNS-323 into RAID 1, as this offers the most safety when it comes to backups
from review article
http://www.cnet.com.au/desktops/storage … 845,00.htm
It would seem reasonable to expect that a basic requirement of raid1 is that the unit can restore data to a new drive when only one drive fails. Can the DNS 323 do this thru its standard interface without having to resort to backing up data onto extra hardware and reformatting (as is proposed in the Dlink manual)?
I see 1.04 firmware is out - but does not seem to address any issues related to raid
DNS-323 Firmware version 1.04 Release Notes.
- Upgrade Samba from 2.x to 3x for Vista and Unicode Support
- FTP client language support for Western European, North European, Central European, Traditional Chinese, Simplified Chinese, and Cyrillic codepages.
- Add auto-refresh function for the iTunes and UPnP AV Server.
- Add set "Oplocks", "Map archive" options in the Network Access page.
- Add clear print queues function to clear any pending queues.
- Add Daylight Savings Time option to Setup Wizard.
- Add GoAhead WebServer logo to the login Webpage.
- Combine User / Group setting into one section.
- Separate the FTP Server Settings intto FTP Server Settings and FTP Account Settings.
- Changed Max. length of username and password for email alerts to 31 characters.
- Changed Max. length of username and password for Scheduled Downloads to 15 characters.
- Fixed failure to mount an existing HDD when it is pre-formatted as NTFS/FAT32 and is being inserted into the device.
- Fixed clock drift issue.
- Fixed issues with mapping drives in Vista when using usernames and passwords
The implementation of raid1 in this unit does not appear to meet reasonable expectations, therefore if the unit was bought with redundancy in mind, might need to look elsewhere. Need to go and have a tinker now.
Offline
Don't worry about using it for RAID, its not really built for that even though its a feature.
However, I've read you can set it up to automatically copy drive 1 to drive 2. I think its part of the existing web GUI. If its backup you're after then that is a better way to go than RAID. (There are also ways to do it via fun_plug).
As you say, RAID just addresses availability and is not a backup strategy. It should only be used to improve uptime of the device. Redundancy through RAID is not a backup strategy of course although it is necessary for protection of uptime.
It will be interesting to see if 1.04 has addressed any of the issues, hard to get a drive to actually "fail" at the right moment though. Ripping a drive out of the bay to simulate a failure makes it too easy for the device to see the mount point has gone, and raise an error that way. To simulate a real drive media failure I expect to be very difficult in practice.
Even plugging in a bad drive to start with, the device may not succeed in formatting it so might refuse to use it from the start.
Good luck though look forward to your analysis!
Offline
Whether or not use configure for RAID should be determined by the intended usage.
Start by asking yourself what will the impact of a failed drive be - there will always be some amount of time involved in the recovery, which ought to involve no more than the replacement of the failed disk and the restoration of data either from a backup or the original source. Now place a dollar value on the time required - does that figure justify the cost of the additional storage required to run RAID1?
ocular
I don't know which part of the D-Link manuak you were reading, but I find no reference to replacing a failed disk, nothing that speaks of having to backup & reformat, and it has been my experience that when a failed disk is replaced with a new or clean disk, the DNS-323 rebuilds the array with a minimum of fuss, and it can be done entirely through the standard interface. Note - new or clean disk - some forum users have reported problems with RAID rebuilds, which may or may not, be related to the use of replacement disks already containing data - something that I learned not to do, many, many years ago.
What defines a failed drive - a failed drive would be anything that jeopardizes the availability of your data, anything that reduces the redundancy below the 1+1 that RAID1 requires, including but not limited to missing drives (either removed or physically present & not detected by the unit due to electronic failure), drives containing corrupted disk formats or data, drives with abnormally high levels of errors, etc.
It has been my experience that removing a drive, which happens to be a fairly common test for redundancy (both with power on & off) could be done undetected with firmware prior to 1.03, with firmware 1.03 & later, the unit will detect the missing drive and send an email if configured to do so - you will not see the amber LED. To the best of my knowledge the unit will not detect corruption of either the disk format or the data and, at least with firmware prior to 1.04, did not utilize S.M.A.R.T if available - I have not yet done any tests in this regard with 1.04.
markchicobaby - I have found that you CAN install known failing disks in the unit and it will format them and allow you to use them without any indication of a problem other than taking a longer than normal time to complete the format - again this is pre firmware 1.04
Offline
There seems to be no indication of raid Status thru the Web GUI with firmware 1.03. Certainly with one drive of a Raid 1 Volume removed the WebGUI is no help.
There does seem to be some changes with raid under firmware 1.04
If the Raid 1 volume is broken then the web GUI Status tab under Hard Disk Info says "degraded" - to refresh you need to logout and then login again.
To get email notification to work after upgrade of firmware to 1.04 I had to unpartition each HDD reformat and rebuild array and add data.
I'll also notice that if you switch off and pull a drive and then reboot and then shut down and replace the same "old" drive there is automatic resyncing on rebooting - this didn't seem to happen with FW 1.03.
During resyncing the data can be still accessed.
On replacing with a new drive and logging into webgui begins reformatting of the new HDD and after about 5 mins rebuilding begins. Downtime is only the 5 minutes whilst reformatting. The dlink interface seems to behave in a more expected manner for Raid 1 now.
It is sad that there is no inherent SMART monitoring in the new Firmware.
Offline
Interesting thread, learned some stuff here. So... when will Dlink come up with a scheduled Sync function so users who cannot telnet in can still take advantage of speed of copying files from one drive to another rather than run Raid 1 to have data in both places. Come on Dlink!
Offline
fordem and mark, what do you recommend for this use case:
We have a DNS-323 and are using it as backup device. Various machines run scheduled backups using the DNS as the destination. The DNS stores multiple backups for each machine.
I'd like to use RAID1 for two purposes:
1) provide high availability/fault tolerance of the backed up data
2) provide an easy way to create off-site copies
Specifically, for #2, I wanted to be able to periodically "fail" a drive and take it to an off-site storage location. Initially I'd start with a handful of new drives and replace the "failed" drive with a blank. After I'd gone through the set of drives, I'd retrieve the first one from off-site and would want to now have the current data rebuilt onto this old drive.
So now this introduces the problem described above about trying to rebuild onto a non-new drive.
Does the GUI support the operations of this use case, e.g. forcing a rebuild onto a drive with old data?
Any other approaches to this use case?
Thanks
Offline
Personally, I don't see the need for high availability/fault tolerance for back up data, and as far as I concerned RAID1 in that application is a waste of space, it is however your space.
With regard to your need for off-site copies - let me start by saying this is not what RAID1 was intended for - however, to answer your question - no, the web admin interface does not allow you to choose which drive should be rebuilt, the only way I know to force that would be to use a PC to remove the data from the "destination" drive.
If you're using fw 1.04, you do have the option of not formatting the "new" drive at power up, which would let you use the CLI to force a rebuild as suggested by Jayas here
Code:
mdadm /dev/md0 -f /dev/sdb2 # signal as faulty
mdadm /dev/md0 -r /dev/sdb2 # remove from array
mdadm /dev/md0 -a /dev/sdb2 # add to the array
This assumes the second drive (sdb2) as being the one you wish to rebuild to.
Offline
takoma,
You might try BackupNetClone (http://backupnetclone.sourceforge.net/) which could automate your on- and off-site backup using the DNS-323.
Offline
Hi Takoma,
In terms of sequencing, you should type the first two commands with the original mirrored drives, then swap the nominated drive, then type the last command.
Jaya
Offline
Specifically, for #2, I wanted to be able to periodically "fail" a drive and take it to an off-site storage location. Initially I'd start with a handful of new drives and replace the "failed" drive with a blank. After I'd gone through the set of drives, I'd retrieve the first one from off-site and would want to now have the current data rebuilt onto this old drive.
I've been reading this forum for about a month, and I reckon that most posts about missing/lost data mention RAID 1. I'm too lazy to do a search to back that up ;-)
I think if you try to to do what you want to do you will either soon be posting a similar unhappy tale or needing to replace the skinny plastic front cover, drive removal lever, or picking fluff out of the SATA connector inside.
Would a couple of USB enclosures be a better solution? You could duplicate your backup copies on to them and the drives would be protected a little better while you transport them.
Good luck!
Last edited by sjmac (2008-03-10 15:45:00)
Offline
sjmac wrote:
Specifically, for #2, I wanted to be able to periodically "fail" a drive and take it to an off-site storage location. Initially I'd start with a handful of new drives and replace the "failed" drive with a blank. After I'd gone through the set of drives, I'd retrieve the first one from off-site and would want to now have the current data rebuilt onto this old drive.
I've been reading this forum for about a month, and I reckon that most posts about missing/lost data mention RAID 1. I'm too lazy to do a search to back that up ;-)
I think if you try to to do what you want to do you will either soon be posting a similar unhappy tale or needing to replace the skinny plastic front cover, drive removal lever, or picking fluff out of the SATA connector inside.
Would a couple of USB enclosures be a better solution? You could duplicate your backup copies on to them and the drives would be protected a little better while you transport them.
Good luck!
I tend to agree heartily - one caveat - a lot of the USB enclosures I've seen offer no protection whatsoever, and I'd be tempted to recommend something like the Quantum GoVault - also sold as the IBM GoVault, Tandberg RDX, and the Dell RD1000.
Offline
Or you could skip hauling physical media back and forth at all... I use BackupNetClone to do remote backup over a slow DSL connection, and with only 256Kbps I can still get 1GB worth of changes per night. It's automatic, sends email status, keeps snapshots of the data going back in time, and doesn't require fussing with moving around the disks...
Offline
sjmac wrote:
I've been reading this forum for about a month, and I reckon that most posts about missing/lost data mention RAID 1. I'm too lazy to do a search to back that up ;-)
It appears to me that loss of data is frequently caused by abrupt removal of power. I have been asked too many times Why can't I just turn off power? Sigh ...
I have had two instances of this recently. The systems were configured as RAID 1 and had the proverbial orange/white light as a result.
In both cases I was able to restore the systems by manually adding the 'faulty' drive back to the array. I cannot imagine data loss would have been avoided if they were not running RAID 1.
Jaya
Offline
That's an interesting insight.
BTW, how were the owners switching off the DNS323? Using the switch on the front, or flipping the wall outlet switch?
Offline
Hi Sjmac,
sjmac wrote:
BTW, how were the owners switching off the DNS323? Using the switch on the front, or flipping the wall outlet switch?
In one case, the owner simply plugged the cable off wall while installing the printer. In the other case it is not clear what exactly caused the power interruption but we know it happened because the device was off, and amber light came on sometime after it was turned on.
Using the switch on the front would not cause such problems. I have done this enough number of times to be satisfied.
Jaya
Offline
fordem wrote:
Personally, I don't see the need for high availability/fault tolerance for back up data, and as far as I concerned RAID1 in that application is a waste of space, it is however your space.
All, thank you very much for the informed discussion on the topic.
For me, the use of RAID1 on a backup device was to ensure availability of the "horizon" of backed up file versions for easy restore.
I do agree with various posts that perhaps using RAID as a "disk duplication" tool for creating offsite backup drives is strategy that needs rethinking ;-}
On the USB enclosure strategy, is there a scripted way that one can initiate a copy of the DNS contents to an externally connected USB drive?
Also, I do like the remote NAS method, but the disconnected drives does provide for another level of protection from virus or other activities that might effect the on-line NAS and remote copy.
Thanks
Offline