That or just turn off ahci. Since the issue does not occur when the drives are set to an IDE mode.

Thanks.  Do I have to shut down the array to issue the write cache command?

 

No

Stock unRAID OS does not issue the IDENTIFY DEVICE command, and I'm fairly certain (without scouring all the code) that the linux disk driver doesn't do this on it's own either, except perhaps at startup.  I don't know about 3rd party add-ons that may be using smartmontools or hdparm. (Joe?)  Probably this is nothing to worry about, but if someone can find if Samsung ever publishes hard drive firmware updates, that would be good to monitor.

 

It occurs to me a workaround to this problem can be put into the linux disk driver.  If the driver wants to issue an IDENTIFY DEVICE command it should first wait for all pending commands to finish first, issue a FLUSH CACHE, wait for it to finish, then go ahead and issue IDENTIFY DEVICE.  I don't know if anyone is working on this though... Probably it's better for Samsung to fix their bug.

Good to know,  thanks for the response.

Stock unRAID OS does not issue the IDENTIFY DEVICE command, and I'm fairly certain (without scouring all the code) that the linux disk driver doesn't do this on it's own either, except perhaps at startup.  I don't know about 3rd party add-ons that may be using smartmontools or hdparm. (Joe?)  Probably this is nothing to worry about, but if someone can find if Samsung ever publishes hard drive firmware updates, that would be good to monitor.

 

It occurs to me a workaround to this problem can be put into the linux disk driver.  If the driver wants to issue an IDENTIFY DEVICE command it should first wait for all pending commands to finish first, issue a FLUSH CACHE, wait for it to finish, then go ahead and issue IDENTIFY DEVICE.  I don't know if anyone is working on this though... Probably it's better for Samsung to fix their bug.

 

Tom,

 

Could you make a fix for the F4?

Can someone suggest which of the problem hdds to get that has the least problematic problem?

Can someone suggest which of the problem hdds to get that has the least problematic problem?

 

Hitachi 7200RPM drives work fine, the WD drives work fine so long as you put the jumper on, Seagate LP may need a firmware update (but that is not to hard to do).

 

The samsung drives need to be avoided as far as I am concerned as anything that can cause silent data corruption need not have a place in a data server like unRAID.

Yup,  I'll be buying WD for my next drive I think.

The EARS and the Seagate LP have potential problems in how long they will work for due to their load/unload cycle problems.

Could you make a fix for the F4?

 

Until Samsung come out with new firmware, the fix is for the user to disable write caching on the drives.

 

Perhaps some addon community creator can create a simple addon script that performs the task for all Samsung F4 drives. Now how does one identify the drive is a Samsung F4 with troubled firmware version without using the IDENTIFY DEVICE command? This seems like a catch 22 situation. To prevent the bug, you potentially trigger the bug. Perhaps the addon would be as simple of allowing disabling of the write cache on a drive by drive basis, thus letting it be up to the user to decide and avoiding potentially triggering the bug.

 

Frankly, I'd prefer for Limetech to spend their time on finalizing 4.6 and coming out with 5.0 beta 3 than mucking about for potentially an endless amount of time inside the raw Linux device drivers just because of mostly worthless drives like the Samsung F4. Let Samsung fix their mess.

The EARS and the Seagate LP have potential problems in how long they will work for due to their load/unload cycle problems.

 

I have heard of this potential problem with the WD drives but not with the seagate drives.

 

I think the load/unload cycles problem is blown out of proportion, but I don't have any green drives in my server.

The Seagate has only 50,000 unload/load cycles whilst all other drives are 300,000+!

The EARS and the Seagate LP have potential problems in how long they will work for due to their load/unload cycle problems.

 

I have heard of this potential problem with the WD drives but not with the seagate drives.

 

I think the load/unload cycles problem is blown out of proportion, but I don't have any green drives in my server.

 

I must be rather fortunate as out of my 3 WD Green drives, they have at most a 1156 Load_Cycle_Count in 12023 Power_On_Hours. That's on the drive that used to be my parity drive and when I used 15 minute idle spin-down timer.

 

I have no idea what my Seagate LP is as it doesn't report attribute 193, Load_Cycle_Count at all in the SMART report.

The following is an email to Samsung UK hdd section. [email protected]        [email protected]

 

I have emailed them about the problem and their reply was

 

"Dear customer,

 

In reference to your firmware issue, So far Samsung have not release  a new firmware for your unit.

 

We suggest to contact Samsung 1-800-726-7864 for further assistance."

 

I think we should bombard them with our concerns about the F4!

I have a reply from Samsung!!!!!

 

Our HQ engineering are aware of this problem and a solution will be ready very soon. I believe we will have a new firmware

 

and will let you when it will be available for download from our website.

 

Regards,

Chas.

 

QA manager - SSEL

http://www.samsung.com/global/business/hdd/productmodel.do?group=72&type=94&subtype=98&model_cd=552&dType=G&mType=SW&tab=down&ppmi=1219

 

He told me that there should be one there by the end of this week. But will tell you when he tells me.

I really hope this gets solved from Samsung by the end of this week.  It will ease some of our stress.

 

Although other sources say not

http://hardforum.com/showpost.php?p=1036531314&postcount=26

I am building a new server and I have 7 brand new untouched F4. I have one Hitachi 7k2000 that will act as parity.

 

How do I put them into the server to avoid any problems?

 

turn off write caching before you write any data to them :-

 

hdparm -W 0 /dev/sda  (replace sda with your disk)

 

wait for a new firmware from Samsung.

 

 

I have 3 of these in my new build (one that will ultimately be for parity, and then 2 more for data). I'm about half way through transferring all my media files over to the server (from external hard drives...I've been copying, though, so I still have all the original files). I haven't set up the parity drive (since the config guide says the initial transfer will be faster without parity set up). Is it too late for this fix to do anything? Do I need to start over (i.e. re-preclear, then turn off write caching, then re-transfer files)? Will a firmware fix solve the issue for everything or just for data added to the drive(s) after the firmware update (I don't know if that question makes any sense)?

 

 

I am building a new server and I have 7 brand new untouched F4. I have one Hitachi 7k2000 that will act as parity.

 

How do I put them into the server to avoid any problems?

 

turn off write caching before you write any data to them :-

 

hdparm -W 0 /dev/sda  (replace sda with your disk)

 

wait for a new firmware from Samsung.

 

 

I have 3 of these in my new build (one that will ultimately be for parity, and then 2 more for data). I'm about half way through transferring all my media files over to the server (from external hard drives...I've been copying, though, so I still have all the original files). I haven't set up the parity drive (since the config guide says the initial transfer will be faster without parity set up). Is it too late for this fix to do anything? Do I need to start over (i.e. re-preclear, then turn off write caching, then re-transfer files)? Will a firmware fix solve the issue for everything or just for data added to the drive(s) after the firmware update (I don't know if that question makes any sense)?

Run the commands to disable write-caching.

 

Odds are you'll be fine.

you can run md5 checksums to verify your existing files against their original copies.  That will be faster than transferring them all again.

 

Tom has said that unRAID did not issue the IDENTITY DEVICE command.  I do know the unMENU add-on uses smartctl and hdparm to get disk status and temperature.  I know they have been  described as issuing the IDENTIFY DEVICE command for some operations.  If you were running unMENU while transferring the files I would get the checksums on the files and verify their correctness.

 

You can get the MD5 checksum of a file by typing

md5sum filename

since the md5 process has to read the entire file this will take a while for larger files.

 

You can get md5 checksums for all your files on a given disk by typing

find /mnt/disk1 -type f -exec md5sum {} \; | tee /boot/md5_disk1_checksums.txt

When running the mdchecksum what am I looking for when comparing it to the original file size?

When running the mdchecksum what am I looking for when comparing it to the original file size?

You are looking for an identical MD5 checksum calculated on the original file.  (There are window's utilities to calculate the md5 checksum too)

 

Comparing the file sizes will not work since the size will likely be identical with the type of corruption that the F4 might have.

 

Here is one example program: http://www.winmd5.com/

(I've never used it just found it with a google search for "windows md5 checker")

Can that utility check a library of files or only 1 file at a time? (i.e. TVShow_Season 1, vs. Episode 1, Episode 2, etc.).

 

Similarly, can the command you listed before (md5sum filename) check an entire directory (again, for example, all of TVShow1, which includes multiple folders, in which there are tv episode files)?

 

The reason I ask (and the reason I can't just do the "check an entire disk" option) is because, while most of my media files have come off of 2 external hard drives, I have been reripping some of my shows and those files are actually on my desktop, not the externals.

Can that utility check a library of files or only 1 file at a time? (i.e. TVShow_Season 1, vs. Episode 1, Episode 2, etc.).

 

Similarly, can the command you listed before (md5sum filename) check an entire directory (again, for example, all of TVShow1, which includes multiple folders, in which there are tv episode files)?

 

The reason I ask (and the reason I can't just do the "check an entire disk" option) is because, while most of my media files have come off of 2 external hard drives, I have been reripping some of my shows and those files are actually on my desktop, not the externals.

You can type

md5sum directory/*

 

or

 

cd directory

md5sum *

DGault

 

Do you run unMENU?

If Yes - Did you view the unMENU web page while transferring files?

 

Peter

 

unMenu is installed, although I haven't really used it at all. I've been following the Configuration tutorial found here:

http://lime-technology.com/wiki/index.php?title=Configuration_Tutorial

 

However, the web page was not open during file transfer. I haven't opened it since it was originally called for in that tutorial.