Joe L, seems like we are getting close, but; still hung up on this one issue.  What's odd is the last series of commands either report no bad block, or have no problem writing 0 to the bad block.  Your the expert, and I am very thankful for your help.  Suff I'm, reading says the bad block 0 respone the the fix-fixable simply means you need to run rebuild-tree.  Which I'm doing, but; then it fails.  Trying again

Now that you are using putty, you can paste the commands you are typing into unRAID and their output here.  It might help.

 

What happens when you run the "dd" command?  What, if anything does it do to the reallocated sector count in the smart report.?

 

It makes it really tough when I cannot see the results, or know what they did...  Please cut from the putty window and past to this thread.

Joe L, O.K., I'll try this when I get home.  Thanks for not giving up on me.  From memory on the dd command it says it wrote something (does 4mg sound right), and everything looked O.K., I wish I could remember the exact result, but; from memory it's hard.  I know I concluded form looking at it, all seemed well.  Just like when I ran the debugreiserfs command on each block.  When I did this, reported that there were no badblocks.

 

Roger

Joe L, I don't know how to thank you enough.  Disk 8 is up.  It is running, virtually everything is there.  Yes, I have to rename some folders and move some stuff into the correct location, but, you helped me save this tower.  I am very grateful.  Thank you. I don't know how to mark this [sOLVED] BUT YOU DID IT.

 

Roger

Wow , what a story ! Joe you are rock star of this community 

Joe L, I don't know how to thank you enough.  Disk 8 is up.  It is running, virtually everything is there.  Yes, I have to rename some folders and move some stuff into the correct location, but, you helped me save this tower.  I am very grateful.  Thank you. I don't know how to mark this [sOLVED] BUT YOU DID IT.

 

Roger

Now... to add it to the array you will need to

stop the array

assign the drive as disk8

set a new disk configuration by logging in via telnet and typing

initconfig

then answer "Yes" to its prompt.

 

Then, when you start the array it will calculate parity on the entire array.

Once parity is completely calculated, Only then can you replace the failing disk8.  (remember, it is a drive that is failing)

 

 

Wow , what a story ! Joe you are rock star of this community 

I need a vacation.... 

Joe L, yes, is beyond a stud.  Hard to believe this all works. Wow, building parity now. Yes, you actually taught me the follow up steps which I remembered without looking.  Yow!  It's running parity now.  About 10% done.  We just need it to complete parity before something goes wrong, and woosh, out goes the disk8, and in goes a new one.

 

Not to be ungrateful, but; how exactly do I do that?  I'm guessing, but won't do until you tell me for sure, I stop the array.  Unassign Disk8, turn the machine off, install the new disk 8.  When it sees the new disk 8 it will ask to format it, say yes, then the parity drive will build it on it's own?  Probably wrong here.  I did have one actually fail, but;and I replaced it, but; I can't remember exactly what the steps were.

 

 

Thanks again.

 

Roger

Roger , WIKI is your best friend mate 

 

try this http://lime-technology.com/forum/index.php?topic=1553.msg10540#msg10540 (from rock star himself  )

or this http://lime-technology.com/wiki/index.php?title=Replacing_a_Data_Drive

try this http://lime-technology.com/forum/index.php?topic=1553.msg10540#msg10540 (from rock star himself  )

Can I be a groupy? 

try this http://lime-technology.com/forum/index.php?topic=1553.msg10540#msg10540 (from rock star himself  )

Can I be a groupy? 

an "unRAID Guru Groupy"

 

    "uGG"

 

 

an "unRAID Guru Groupy"

 

    "uGG"

 

 

I so want to be an uGG  

I've been reading as it developed...but - what a thread! and perseverance (who had more? - that is debatable)

I'm still not exactly sure what Roger did at the end (was it the dd command?) but it is fixed.  Great support Joe, great way to learn a little on your own Roger.  Congrats!

UGG

 

My view; Joe L, somehow figured out the correct series of commands to deal with the 19 bad blocks. There are lots of cites on the net that try and explain how these commands work in reisercschk but unfortunately, they don't address for a newbie like me how to phrase the command in linux well enough.  Joe, obviously, figured it out.  Not the first time, but; the second time.  GOOD ENOUGH FOR ME!

 

After dealing with the routine issues with the other 13 drives (routine for Joe, I only knew how to type my name), the bad blocks were a challenge and finding a way to correctly script the commands was the key. 

 

So, once he figured out how to script the commands, I think that was it.  Ultimately, I think two things brought the drive up, the non destructive writes to the the bad blocks, and then writing the zero's to them.  After that a rebuild-tree -S brought the drive up.

 

Roger

Joe L, I saw the post of your discussion on replacing a bad drive, excellent and it now is familiar.  As you said I'm the first one to run your series of commands to fix the bad blocks, is there a way to make that clearly available to other users?  My Subject line won't necessarily bring it up on a search, and I think the last set of three commands is the key to fixing a badblock situation:

 

The correct sequence should therefore be:

First.  badblocks -v -o /boot/badblocks_disk8 -b 4096 /dev/sdX1

(where sdX1 =  the four character device name of the disk8 reiserfs partition. )

Second.  reiserfsck --fix-fixable --badblocks /boot/badblocks_disk8 /dev/sdX1

(again, /dev/sdX1 = the first partition, where the file-system resides)

Followed by:

reiserfsck --rebuild-tree -S /dev/sdX1

 

IF THE AGOVE DIDN'T WORK, THEN:

You can try this... but be absolutely sure you get the commands correct.

 

badblocks -b 4096 -p 100 -s -v -n /dev/sdX1  52051000 52051000

 

Since "-n" uses a non-destructive read/write mode, it might force the re-allocation of the bad block.

Do this for each of the block addresses marked as bad.

 

You can test if it is subsequently readable by typing:

debugreiserfs -1 52051000 /dev/sdX1

 

If it is, great.  If not we can force write zeros to that block and hopefully force the SMART firmware to re-allocate it for you.

dd if=/dev/zero bs=4096 count=1 seek=52051000  of=/dev/sdX1

 

repeat with each of the blocks listed in the badblocks file.  Then

reiserfsck --check /dev/sdX1

 

followed by whatever it suggests next.

probably

reiserfsck --fix-fixable /dev/sdX1

or

reiserfsck --rebuild-tree -S /dev/sdX1

 

 

 

 

 

 

 

 

an "unRAID Guru Groupy"

 

    "uGG"

 

 

I so want to be an uGG  

 

Me too!!

 

I want to be a uGG!!

Joe L, I saw the post of your discussion on replacing a bad drive, excellent and it now is familiar.  As you said I'm the first one to run your series of commands to fix the bad blocks, is there a way to make that clearly available to other users?  My Subject line won't necessarily bring it up on a search, and I think the last set of three commands is the key to fixing a badblock situation:

 

The correct sequence should therefore be:

First.  badblocks -v -o /boot/badblocks_disk8 -b 4096 /dev/sdX1

(where sdX1 =  the four character device name of the disk8 reiserfs partition. )

Second.  reiserfsck --fix-fixable --badblocks /boot/badblocks_disk8 /dev/sdX1

(again, /dev/sdX1 = the first partition, where the file-system resides)

Followed by:

reiserfsck --rebuild-tree -S /dev/sdX1

 

IF THE AGOVE DIDN'T WORK, THEN:

You can try this... but be absolutely sure you get the commands correct.

 

badblocks -b 4096 -p 100 -s -v -n /dev/sdX1  52051000 52051000

 

Since "-n" uses a non-destructive read/write mode, it might force the re-allocation of the bad block.

Do this for each of the block addresses marked as bad.

 

You can test if it is subsequently readable by typing:

debugreiserfs -1 52051000 /dev/sdX1

 

If it is, great.  If not we can force write zeros to that block and hopefully force the SMART firmware to re-allocate it for you.

dd if=/dev/zero bs=4096 count=1 seek=52051000  of=/dev/sdX1

 

repeat with each of the blocks listed in the badblocks file.  Then

reiserfsck --check /dev/sdX1

 

followed by whatever it suggests next.

probably

reiserfsck --fix-fixable /dev/sdX1

or

reiserfsck --rebuild-tree -S /dev/sdX1

 

 

 

Has this been made a wiki article yet. Very good information.

I might add to it with commentary on ddrescue.

 

One point of anyone finding this in the future (/me waves from the past).

If you run badblocks in readonly mode and you find errors, your parity sync/generate is in jeopardy until you resolve the read errors.

 

As far as the DD writing zero's to force reallocation of the sector, The badlocks non destructive write test should be sufficient, since any write will remap a questionable sector in the SMART firmware.

 

However of you cannot remap the sector under any circmstance, then a last resort may be to use the ddrescue program to copy the disk to a known good disk. 

 

For example. I had a drive go bad. When reading the sectors forward, it would get stuck at a point and the drive would just go offline. so recovery was not possible.

 

With DD rescue, I was able to log where the drive failed. reboot, restart and continue from that point forward.

I did this several times over several reboots.

Then re-ran ddrescue in reverse mode where it read from the drives end to the start. This was in recovery mode.

It only tested the known bad sectors. But out of 5000 sectors it recovered all except 1. The one sector that started the whole mess.

 

Further reading here.

http://lime-technology.com/forum/index.php?topic=16734.0