Remove a materially failed device from a Btrfs "single-raid" using partitions

Remove a materially failed device from a Btrfs "single-raid" using partitions

[Vincent]

Hello,

I am on Ubuntu Server 13.04 with Linux 3.8.

I've created a "single-raid" using /dev/sd{a,b,c,d}{1,3}. One of my hard 
drives has failed, I mean it's materially dead.

:~$ sudo btrfs filesystem show
Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
     Total devices 5 FS bytes used 226.90GB
     devid    4 size 37.27GB used 31.01GB path /dev/sdd1
     devid    3 size 37.27GB used 31.01GB path /dev/sdc1
     devid    2 size 37.31GB used 31.00GB path /dev/sdb1
     devid    1 size 139.73GB used 132.02GB path /dev/sda3
     *** Some devices missing


Many tutorials I found about it never mention the simple deletion of a 
non-remountable disk in case of a "single-raid" (where the datas doesn't 
matter, I've used the only "d=single" option, insinuating "m=mirrored").

I've read this page http://www.howtoforge.com/a-beginners-guide-to-btrfs 
until "8 Adding/Deleting Hard Drives To/From A btrfs File System" 
section. But this page want to make me mount the drive, but it's dead.

When my Btrfs partition is not mounted and when I do:
:~$ sudo btrfs device delete missing
btrfs device delete: too few arguments
or
:~$ sudo btrfs device delete missing /media/single-raid/
Nothing happen.

If I try to mount the failed device, and remove /dev/sde1 from the 
mountpoint, my console doesn't respond anymore.

I've also read the official documentation 
https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Removing_devices 
using degraded mode: mount -o degraded /dev/sda3 /media/single-raid/

The fstab line is however: /dev/sda3 
/media/single-raid/                         btrfs 
device=/dev/sda3,device=/dev/sdb1,device=/dev/sdc1,device=/dev/sdd1 
0       2

Then perform :~$ sudo btrfs filesystem show
Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
     Total devices 5 FS bytes used 226.30GB
     devid    4 size 37.27GB used 31.01GB path /dev/sdd1
     devid    3 size 37.27GB used 31.01GB path /dev/sdc1
     devid    2 size 37.31GB used 31.00GB path /dev/sdb1
     devid    1 size 139.73GB used 132.02GB path /dev/sda3
**** Some devices missing*

I don't understand why in degraded mode I can't remove the failed 
device. Could you help me please?

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Harald Glatt]

On Mon, May 13, 2013 at 4:15 PM, Vincent <vincent@influence-pc.fr> wrote:
> Hello,
>
> I am on Ubuntu Server 13.04 with Linux 3.8.
>
> I've created a "single-raid" using /dev/sd{a,b,c,d}{1,3}. One of my hard
> drives has failed, I mean it's materially dead.
>
> :~$ sudo btrfs filesystem show
> Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
>     Total devices 5 FS bytes used 226.90GB
>     devid    4 size 37.27GB used 31.01GB path /dev/sdd1
>     devid    3 size 37.27GB used 31.01GB path /dev/sdc1
>     devid    2 size 37.31GB used 31.00GB path /dev/sdb1
>     devid    1 size 139.73GB used 132.02GB path /dev/sda3
>     *** Some devices missing
>
>
> Many tutorials I found about it never mention the simple deletion of a
> non-remountable disk in case of a "single-raid" (where the datas doesn't
> matter, I've used the only "d=single" option, insinuating "m=mirrored").
>
> I've read this page http://www.howtoforge.com/a-beginners-guide-to-btrfs
> until "8 Adding/Deleting Hard Drives To/From A btrfs File System" section.
> But this page want to make me mount the drive, but it's dead.
>
> When my Btrfs partition is not mounted and when I do:
> :~$ sudo btrfs device delete missing
> btrfs device delete: too few arguments
> or
> :~$ sudo btrfs device delete missing /media/single-raid/
> Nothing happen.
>
> If I try to mount the failed device, and remove /dev/sde1 from the
> mountpoint, my console doesn't respond anymore.
>
> I've also read the official documentation
> https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Removing_devices
> using degraded mode: mount -o degraded /dev/sda3 /media/single-raid/
>
> The fstab line is however: /dev/sda3 /media/single-raid/
> btrfs device=/dev/sda3,device=/dev/sdb1,device=/dev/sdc1,device=/dev/sdd1 0
> 2
>
> Then perform :~$ sudo btrfs filesystem show
> Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
>     Total devices 5 FS bytes used 226.30GB
>     devid    4 size 37.27GB used 31.01GB path /dev/sdd1
>     devid    3 size 37.27GB used 31.01GB path /dev/sdc1
>     devid    2 size 37.31GB used 31.00GB path /dev/sdb1
>     devid    1 size 139.73GB used 132.02GB path /dev/sda3
> **** Some devices missing*
>
> I don't understand why in degraded mode I can't remove the failed device.
> Could you help me please?
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

If you have used d=single, it means that the data that was on the
drive that failed is now gone. I think btrfs refuses to remove devices
if it means data loss, but I could be wrong here..

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Vincent]

Le 13/05/2013 16:29, Harald Glatt a écrit :
> On Mon, May 13, 2013 at 4:15 PM, Vincent <vincent@influence-pc.fr> wrote:
>> Hello,
>>
>> I am on Ubuntu Server 13.04 with Linux 3.8.
>>
>> I've created a "single-raid" using /dev/sd{a,b,c,d}{1,3}. One of my hard
>> drives has failed, I mean it's materially dead.
>>
>> :~$ sudo btrfs filesystem show
>> Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
>>      Total devices 5 FS bytes used 226.90GB
>>      devid    4 size 37.27GB used 31.01GB path /dev/sdd1
>>      devid    3 size 37.27GB used 31.01GB path /dev/sdc1
>>      devid    2 size 37.31GB used 31.00GB path /dev/sdb1
>>      devid    1 size 139.73GB used 132.02GB path /dev/sda3
>>      *** Some devices missing
>>
>>
>> Many tutorials I found about it never mention the simple deletion of a
>> non-remountable disk in case of a "single-raid" (where the datas doesn't
>> matter, I've used the only "d=single" option, insinuating "m=mirrored").
>>
>> I've read this page http://www.howtoforge.com/a-beginners-guide-to-btrfs
>> until "8 Adding/Deleting Hard Drives To/From A btrfs File System" section.
>> But this page want to make me mount the drive, but it's dead.
>>
>> When my Btrfs partition is not mounted and when I do:
>> :~$ sudo btrfs device delete missing
>> btrfs device delete: too few arguments
>> or
>> :~$ sudo btrfs device delete missing /media/single-raid/
>> Nothing happen.
>>
>> If I try to mount the failed device, and remove /dev/sde1 from the
>> mountpoint, my console doesn't respond anymore.
>>
>> I've also read the official documentation
>> https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Removing_devices
>> using degraded mode: mount -o degraded /dev/sda3 /media/single-raid/
>>
>> The fstab line is however: /dev/sda3 /media/single-raid/
>> btrfs device=/dev/sda3,device=/dev/sdb1,device=/dev/sdc1,device=/dev/sdd1 0
>> 2
>>
>> Then perform :~$ sudo btrfs filesystem show
>> Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
>>      Total devices 5 FS bytes used 226.30GB
>>      devid    4 size 37.27GB used 31.01GB path /dev/sdd1
>>      devid    3 size 37.27GB used 31.01GB path /dev/sdc1
>>      devid    2 size 37.31GB used 31.00GB path /dev/sdb1
>>      devid    1 size 139.73GB used 132.02GB path /dev/sda3
>> **** Some devices missing*
>>
>> I don't understand why in degraded mode I can't remove the failed device.
>> Could you help me please?
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> If you have used d=single, it means that the data that was on the
> drive that failed is now gone. I think btrfs refuses to remove devices
> if it means data loss, but I could be wrong here..

I've no problem with data loss, I just want to have a kind of sharing 
area, but data on other drive are always here, why couldn't I retrieve 
them alive?

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Harald Glatt]

On Mon, May 13, 2013 at 4:33 PM, Vincent <vincent@influence-pc.fr> wrote:
> Le 13/05/2013 16:29, Harald Glatt a écrit :
>
>> On Mon, May 13, 2013 at 4:15 PM, Vincent <vincent@influence-pc.fr> wrote:
>>>
>>> Hello,
>>>
>>> I am on Ubuntu Server 13.04 with Linux 3.8.
>>>
>>> I've created a "single-raid" using /dev/sd{a,b,c,d}{1,3}. One of my hard
>>> drives has failed, I mean it's materially dead.
>>>
>>> :~$ sudo btrfs filesystem show
>>> Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
>>>      Total devices 5 FS bytes used 226.90GB
>>>      devid    4 size 37.27GB used 31.01GB path /dev/sdd1
>>>      devid    3 size 37.27GB used 31.01GB path /dev/sdc1
>>>      devid    2 size 37.31GB used 31.00GB path /dev/sdb1
>>>      devid    1 size 139.73GB used 132.02GB path /dev/sda3
>>>      *** Some devices missing
>>>
>>>
>>> Many tutorials I found about it never mention the simple deletion of a
>>> non-remountable disk in case of a "single-raid" (where the datas doesn't
>>> matter, I've used the only "d=single" option, insinuating "m=mirrored").
>>>
>>> I've read this page http://www.howtoforge.com/a-beginners-guide-to-btrfs
>>> until "8 Adding/Deleting Hard Drives To/From A btrfs File System"
>>> section.
>>> But this page want to make me mount the drive, but it's dead.
>>>
>>> When my Btrfs partition is not mounted and when I do:
>>> :~$ sudo btrfs device delete missing
>>> btrfs device delete: too few arguments
>>> or
>>> :~$ sudo btrfs device delete missing /media/single-raid/
>>> Nothing happen.
>>>
>>> If I try to mount the failed device, and remove /dev/sde1 from the
>>> mountpoint, my console doesn't respond anymore.
>>>
>>> I've also read the official documentation
>>>
>>> https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Removing_devices
>>> using degraded mode: mount -o degraded /dev/sda3 /media/single-raid/
>>>
>>> The fstab line is however: /dev/sda3 /media/single-raid/
>>> btrfs device=/dev/sda3,device=/dev/sdb1,device=/dev/sdc1,device=/dev/sdd1
>>> 0
>>> 2
>>>
>>> Then perform :~$ sudo btrfs filesystem show
>>> Label: none  uuid: 40886f51-8c9b-4be1-8721-83bf5653d2a0
>>>      Total devices 5 FS bytes used 226.30GB
>>>      devid    4 size 37.27GB used 31.01GB path /dev/sdd1
>>>      devid    3 size 37.27GB used 31.01GB path /dev/sdc1
>>>      devid    2 size 37.31GB used 31.00GB path /dev/sdb1
>>>      devid    1 size 139.73GB used 132.02GB path /dev/sda3
>>> **** Some devices missing*
>>>
>>> I don't understand why in degraded mode I can't remove the failed device.
>>> Could you help me please?
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>>> the body of a message to majordomo@vger.kernel.org
>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>> If you have used d=single, it means that the data that was on the
>> drive that failed is now gone. I think btrfs refuses to remove devices
>> if it means data loss, but I could be wrong here..
>
>
> I've no problem with data loss, I just want to have a kind of sharing area,
> but data on other drive are always here, why couldn't I retrieve them alive?
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

I think it's a safety mechanism, you'll have to wait until a dev
responds... Otherwise you can always rsync the data into a new
filesystem and replace the old one if you can't wait.

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Vincent]

Shit, I've followed the official documentation 
https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Replacing_failed_devices

With, as explain:
mkfs.btrfs -d single /dev/sda3 /dev/sdb1 /dev/sdc1 /dev/sdd1

mount -o degraded /dev/sda3 /media/single-raid/
btrfs device delete missing /media/single-raid/

And now all my datas on ALL drive has been deleted. How could I retrieve 
datas on my Btrfs formatted partition now?!

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Chris Murphy]

On May 13, 2013, at 9:30 AM, Vincent <vincent@influence-pc.fr> wrote:

> Shit, I've followed the official documentation https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Replacing_failed_devices
> 
> With, as explain:
> mkfs.btrfs -d single /dev/sda3 /dev/sdb1 /dev/sdc1 /dev/sdd1

You did this originally, not just a couple hours ago, right? Because if you did it again, of course you've deleted all of your data by creating a new file system.
> 
> mount -o degraded /dev/sda3 /media/single-raid/

This should work, and you should be able to rsync or copy the data off the file system onto a new one, except for data that's on the failed drive. I'm not sure what sort of errors you get when copying data that the file system knows isn't available.

> btrfs device delete missing /media/single-raid/

I don't expect this to apply to either -d single or -d raid0. It doesn't actually have any benefit or meaning. I'd say at best, my expectation is you could copy out the surviving data, in degraded mode, out of the btrfs file system onto a new file system. I do not expect it's possible to replace the dead drive and go happily along with just the missing data being lost. You'll have to create a new file system at some point.


> And now all my datas on ALL drive has been deleted. How could I retrieve datas on my Btrfs formatted partition now?!

If you really formatted it, I'd expect this. If that was an example of the original mkfs, I'd expect degraded should allow access to data on working drives. Maybe read/write is possible, but personally I'd prefer the file system in such a degraded state to be read only. And get the surviving data off.

But I am not a Btrfs developer.


Chris Murphy

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Vincent]

Why is it more easier to extend a Brfs partition rather than to reduce 
it, while the single raid is supposed to be made ​​for that?

Re: Remove a materially failed device from a Btrfs "single-raid" using partitions

[Chris Murphy]

On May 14, 2013, at 9:30 AM, Vincent <vincent@influence-pc.fr> wrote:

> Why is it more easier to extend a Brfs partition rather than to reduce it, while the single raid is supposed to be made ​​for that?

Are you referring to normal operation or your degraded case? In normal operation, yes you can reduce it within the limits of the physical devices.

Expecting anything other than getting the surviving data off a -d single volume, I think, is unreasonable. If Btrfs even lets you extract the surviving data off the -d single volume, insofar as I'm aware that's better than anything else presently offers.

Chris Murphy