Hello,

 

I had a failing 2TB disk.

I bought the same 8TB I had bought for my parity drive and put in place of the failing disk.

The clearing took a while (55hrs) and after that I had to power down the system because I lost the WebGUI.

The system went up without problem and all my disks were present.

I assigned my new 8TB disk (now cleared) in place of the failing one and started the array for it to rebuild the data and expand the array.

The rebuild took a while (8TB to write, 24hrs straight)

The rebuild is done but the disk appears as unmountable.

 

What did I do wrong ?

it is not the first time I have a failing drive and I thought I knew the proper way to do it.

 

I run Unraid 6.1.8 with a Pro version and an array of 24 disks

I am fairly new to the system

 

It probably means there was file system level corruption before the rebuild was started.  A rebuild does not fix such an issue, merely recreates on the new disk.  If you stop the array; restart it in Maintenance mode; and then run a file system check against the drive in question it will tell you if there is a problem and what needs to be done to resolve it.

Thank you for your advice, I am doing it right now.

 

Phase 1 - find and verify superblock...
couldn't verify primary superblock - not enough secondary superblocks with matching geometry !!!

attempting to find secondary superblock...
.....................with a lot of dots.........

The check is done and I don't seem to have any more hint on what to do.

 

Phase 1 - find and verify superblock...
couldn't verify primary superblock - not enough secondary superblocks with matching geometry !!!

attempting to find secondary superblock...
......a million dots....
unable to verify superblock, continuing...
..... a Billion dots.....
Exiting now.

 

Should I try to repair ?

Should I try to put back the old disk ?

except a smart alert, there was no problem.

But the smart alert went and the disk was marked by a cross.

 

What device name did you use (you did not say)?  Was the array in Maintenance mode?

The array was in maintenance mode as you told me to do so.

I just pressed the button [Check] in the Check Filesystem Status of the very disk.

 

The option seems to be -n

Was you old 2tb disk showing data or unmountable after it redballed?

 

If data was still accessible, was the new disk showing data when you started the rebuild or was it unmountable from the beginning?

 

One more thing, did you by any chance try to change filesystem before rebuilding?

My old disk was showing a cross, of course I could access the data by emulation.

 

I did not try to change the file system, I juste cleared the new disk in place of the old one (no slot left) and tried a rebuild.

 

Assume when you say cleared you mean precleared using the script/plugin.

 

In that case something went wrong during the rebuild, as xfs_repair is not finding a valid filesystem to repair.

 

Do you do periodic parity checks? Can you still read data from the old disk?

 

do you have a way to read XFS from Windows 7 ?

I don't have any more slot to plug this hdd in the server

 

I tried to launch the xfs_repair.

do you think it is premature ?

Should I cancel it ?

 

I tried to stop the array and put the old disk back to see if I could do a repair on it.

But the system told me that the replacement drive should be at least as big as the orginal.

The new disk seems to be already the one considered, and it it is 8TB.

 

Will I need to access the old drive to get the data back ?

You didn't answer a question, was the new disk unmountable from the beginning of the rebuild?

 

If yes, your parity probably wasn't 100% synced, in this case and if the old disk is readable it's probably your best option.

There's one more thing you can try, though I think it's only worth trying if the rebuild started with the disk mounted, and for some reason didn't finished correctly.

 

Stop array, unassign the disk (select no device), start array, if the emulated disk is still unmountable I think the old disk is your best bet, if it's mountable reassign the disk to rebuild again.

I had a failing 2TB disk.

My old disk was showing a cross, of course I could access the data by emulation.

Should I try to put back the old disk ?

except a smart alert, there was no problem.

But the smart alert went and the disk was marked by a cross.

When you say failing, what exactly was wrong with it, what were the symptoms of failure?

What was the SMART alert you saw?

When you say you could access the data from it, did it appear completely accessible, no issues at all with the data?

 

The option seems to be -n

The -n option indicates a trial run, a read-only check of the file system.  To actually make fixes, you have to remove the -n.  (You may have already figured that out.)

 

If the data on the old disk is fine, then you may want to add it back.  You will have to do a New Config and rebuild parity though, so the old disk needs to be good enough to not stop a parity rebuild.  But this option can be done any time, so you might as well continue experimenting with the new drive, as long as you want.

 

By the way, more slots would be a very good thing!  You can add a 2 port controller for around $10.

Yes, the new disk was unmountable from the beginning of the rebuild.

 

My parity was valid but maybe the corrupted filesystem was also in the parity.

 

The old disk can't be mounted in the array, as the array ask now for a 8TB disk...

 

Is there a way to read XFS drive in windows ?

 

I could try to fetch back my data.

I could format the new disk and then copy the data back on it through network.

Yes, the new disk was unmountable from the beginning of the rebuild.

 

In this case it was always going to stay unmountable at the end of it.

 

The old disk can't be mounted in the array, as the array ask now for a 8TB disk...

 

You can, but you have to a new config and sync parity, if the old disk is really bad it's possible it won't complete, it's difficult to say more without seeing at leat a SMART report.

 

Is there a way to read XFS drive in windows ?

 

I could try to fetch back my data.

I could format the new disk and then copy the data back on it through network.

 

I believe the best option for this is using a live linux cd/pen drive like Ubuntu but maybe someone else knows of a windows util.

Yes, the new disk was unmountable from the beginning of the rebuild.

 

My parity was valid but maybe the corrupted filesystem was also in the parity.

A common misconception about parity is that it has some idea of what data is any particular file.  Actually parity knows nothing about the data on a disk, and it does not even know what type of file system you have there.  All it knows is how to put a any particular sector back to how it thinks it should be.    Therefore if you have corruption before you start a rebuild it will always still be there afterwards.

A common misconception about parity is that it has some idea of what data is any particular file.  Actually parity knows nothing about the data on a disk, and it does not even know what type of file system you have there.  All it knows is how to put a any particular sector back to how it thinks it should be.    Therefore if you have corruption before you start a rebuild it will always still be there afterwards.

Exactly my point.

So I am trying to read the data on the old disk with TestDisk.

It should be able to read a XFS, but I don't know if it will be able to correct one.

It is right now analyzing it (45%)

 

The repair on the new disk (maintenance mode) is executing now (can't say how long it will take as the hard drive is HUGE.

I found the smart report for the old disk (in attach)

SAMSUNG_HD204UI_S2H7JD2ZA04744-20160315-2052.txt

Old disk does have pending sectors, so a new config is not a good option,  hopefully you'll be able to recover most data from it.

Old disk does have pending sectors, so a new config is not a good option,  hopefully you'll be able to recover most data from it.

What would be your way for doing that ?

Since you don't have extra ports and if you have one available, put the old disk in another computer, easiest way would probably be with an unraid trial key, if the disk mounts you should be able to copy most data to your server, I never tried it but some people have also used a Ubuntu live cd/pen, supposedly there are also windows utils that can read xfs formatted disks.

As I stated a few hours ago, I started the xfs_repair on the new drive (seagate 8TB) in the array.

I am surprised to see that the main page is stating a rebuild.

It will take a while and if the rebuild s the same as the first one, it will not change anything.

 

I tried to read the xfs old drive (samsung F4) on windows, Testdisk states a Bad partition and can't seem to be able to correct it.

 

I am not sure of what I am doing and I don't understand how could have I done it differently.

If a drive fails, I can only prep a new one and ask for a rebuild, which I did.

Can't Unraid handle the problem of corrupted filesystem ?

I only had a failure on the drive because of pending sectors, which is something everyone could experience, so It must be handled someway, but how ?

A rebuild would normally never start on its own.

 

You shouldn't be doing a rebuild and xfs_repair at the same time, both operations will slow down considerably especially because you trying to repair the emulated disk.

 

Testdisk main function is to recover damaged partitions, did you actually try to mount the old disk on unraid?

I didn't ask or press anything for the rebuild

I pressed the button for xfs_repair, that's all.

 

Should I stop the rebuild ?

 

As I told you, the array doesn't want the old disk as the new disk is 8TB and the OLD is 2TB.

I tried to plug it in and start the array and it clearly told me the new disk should be as big or bigger than the old one.

 

Do you mean mount the old disk out of the array ?

How can I do that ?

 

...put the old disk in another computer, easiest way would probably be with an unraid trial key, if the disk mounts you should be able to copy most data to your server