RAID-6 with 3 missing disks

RAID-6 with 3 missing disks

[Piergiorgio Sartor]

Hi all,

I know this was probably already discussed, but
maybe I need some refresh.

I've a 10 HDDs RAID-6 which, due to mishap (disks
were disconnected accidentaly), has now 3 missing
devices and cannot be assembled.
The data should be OK, since no writes were occurring
during the accident, so putting them together again
should work.

As far as I know, one option is to create, with
"mdadm -C" the array again, giving the disks in
the proper order.

Since all HDDs are readable, I guess "mdadm -E"
should return the role of each device.
Is this correct for the creation order?

Second question is about the "Data Offset", since
this array was created with an older version of
"mdadm" and the data offset is very close to the
superblock.
As far as I know, new mdadm creates the data a
bit far aways.
Is there any way to specifiy the proper offset?

Finally, is there an alternative to "mdadm -C"
or it is the only option?
Forcing assembly does not work, but maybe there
is another way to tell mdadm to really assemby
the array, taking into account the superblock
information, which are all readable.

Thanks a lot in advance,

bye,

-- 

piergiorgio

Re: RAID-6 with 3 missing disks

[Rudy Zijlstra]

Hi

I would start with mdadm assemble --force

Do not use create unless all else has failed

Cheers

Rudy
---
Verstuurd met mijn BlackBerry van Vodafone

-----Original Message-----
From: Piergiorgio Sartor <piergiorgio.sartor@nexgo.de>
Sender: linux-raid-owner@vger.kernel.org
Date:	Sat, 4 May 2013 18:08:04 
To: <linux-raid@vger.kernel.org>
Subject: RAID-6 with 3 missing disks

Hi all,

I know this was probably already discussed, but
maybe I need some refresh.

I've a 10 HDDs RAID-6 which, due to mishap (disks
were disconnected accidentaly), has now 3 missing
devices and cannot be assembled.
The data should be OK, since no writes were occurring
during the accident, so putting them together again
should work.

As far as I know, one option is to create, with
"mdadm -C" the array again, giving the disks in
the proper order.

Since all HDDs are readable, I guess "mdadm -E"
should return the role of each device.
Is this correct for the creation order?

Second question is about the "Data Offset", since
this array was created with an older version of
"mdadm" and the data offset is very close to the
superblock.
As far as I know, new mdadm creates the data a
bit far aways.
Is there any way to specifiy the proper offset?

Finally, is there an alternative to "mdadm -C"
or it is the only option?
Forcing assembly does not work, but maybe there
is another way to tell mdadm to really assemby
the array, taking into account the superblock
information, which are all readable.

Thanks a lot in advance,

bye,

-- 

piergiorgio
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: RAID-6 with 3 missing disks

[Piergiorgio Sartor]

Hi Rudy,

thanks for the answer, but as mentioned at the end,
"--force" assemby does not work.
Reason is, 7 disks complains 3 are missing and the
3 missing are assembed, since their superblock does
not report errors.
Of course, 3 disks are not enough to assembly the
array, forced or not.

bye,

pg

On Sat, May 04, 2013 at 04:22:39PM +0000, Rudy Zijlstra wrote:
> Hi
> 
> I would start with mdadm assemble --force
> 
> Do not use create unless all else has failed
> 
> Cheers
> 
> Rudy
> ---
> Verstuurd met mijn BlackBerry van Vodafone
> 
> -----Original Message-----
> From: Piergiorgio Sartor <piergiorgio.sartor@nexgo.de>
> Sender: linux-raid-owner@vger.kernel.org
> Date:	Sat, 4 May 2013 18:08:04 
> To: <linux-raid@vger.kernel.org>
> Subject: RAID-6 with 3 missing disks
> 
> Hi all,
> 
> I know this was probably already discussed, but
> maybe I need some refresh.
> 
> I've a 10 HDDs RAID-6 which, due to mishap (disks
> were disconnected accidentaly), has now 3 missing
> devices and cannot be assembled.
> The data should be OK, since no writes were occurring
> during the accident, so putting them together again
> should work.
> 
> As far as I know, one option is to create, with
> "mdadm -C" the array again, giving the disks in
> the proper order.
> 
> Since all HDDs are readable, I guess "mdadm -E"
> should return the role of each device.
> Is this correct for the creation order?
> 
> Second question is about the "Data Offset", since
> this array was created with an older version of
> "mdadm" and the data offset is very close to the
> superblock.
> As far as I know, new mdadm creates the data a
> bit far aways.
> Is there any way to specifiy the proper offset?
> 
> Finally, is there an alternative to "mdadm -C"
> or it is the only option?
> Forcing assembly does not work, but maybe there
> is another way to tell mdadm to really assemby
> the array, taking into account the superblock
> information, which are all readable.
> 
> Thanks a lot in advance,
> 
> bye,
> 
> -- 
> 
> piergiorgio
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 

piergiorgio

Re: RAID-6 with 3 missing disks

[Piergiorgio Sartor]

Hi all again,

I solved the problem, maybe useful to know how.

I used the fully working disks (the 3 disconnected)
as reference and I recreated the configuration by
modifying the superblock information of the other
disks (including checksum).

The array was then working and the data was there,
at least to the extent I could check (all files
stored have MD5, but I did not check all).

Maybe one suggestion for "mdadm" future feature.
It would be nice to be able to perform such
operations, like forcing an assembly, also under
extreme condition.
With all the warnings of the case, of course.

bye,

pg

On Sat, May 04, 2013 at 06:08:04PM +0200, Piergiorgio Sartor wrote:
> Hi all,
> 
> I know this was probably already discussed, but
> maybe I need some refresh.
> 
> I've a 10 HDDs RAID-6 which, due to mishap (disks
> were disconnected accidentaly), has now 3 missing
> devices and cannot be assembled.
> The data should be OK, since no writes were occurring
> during the accident, so putting them together again
> should work.
> 
> As far as I know, one option is to create, with
> "mdadm -C" the array again, giving the disks in
> the proper order.
> 
> Since all HDDs are readable, I guess "mdadm -E"
> should return the role of each device.
> Is this correct for the creation order?
> 
> Second question is about the "Data Offset", since
> this array was created with an older version of
> "mdadm" and the data offset is very close to the
> superblock.
> As far as I know, new mdadm creates the data a
> bit far aways.
> Is there any way to specifiy the proper offset?
> 
> Finally, is there an alternative to "mdadm -C"
> or it is the only option?
> Forcing assembly does not work, but maybe there
> is another way to tell mdadm to really assemby
> the array, taking into account the superblock
> information, which are all readable.
> 
> Thanks a lot in advance,
> 
> bye,
> 
> -- 
> 
> piergiorgio
> --
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 

piergiorgio

Re: RAID-6 with 3 missing disks

[NeilBrown]

[-- Attachment #1: Type: text/plain, Size: 2726 bytes --]

On Sat, 4 May 2013 18:30:06 +0200 Piergiorgio Sartor
<piergiorgio.sartor@nexgo.de> wrote:

> Hi Rudy,
> 
> thanks for the answer, but as mentioned at the end,
> "--force" assemby does not work.
> Reason is, 7 disks complains 3 are missing and the
> 3 missing are assembed, since their superblock does
> not report errors.
> Of course, 3 disks are not enough to assembly the
> array, forced or not.

Details please.  "--examine" output of every device would be a good start.
Output for "mdadm --assemble --force --verbose ....."
would help too.

NeilBrown


> 
> bye,
> 
> pg
> 
> On Sat, May 04, 2013 at 04:22:39PM +0000, Rudy Zijlstra wrote:
> > Hi
> > 
> > I would start with mdadm assemble --force
> > 
> > Do not use create unless all else has failed
> > 
> > Cheers
> > 
> > Rudy
> > ---
> > Verstuurd met mijn BlackBerry van Vodafone
> > 
> > -----Original Message-----
> > From: Piergiorgio Sartor <piergiorgio.sartor@nexgo.de>
> > Sender: linux-raid-owner@vger.kernel.org
> > Date:	Sat, 4 May 2013 18:08:04 
> > To: <linux-raid@vger.kernel.org>
> > Subject: RAID-6 with 3 missing disks
> > 
> > Hi all,
> > 
> > I know this was probably already discussed, but
> > maybe I need some refresh.
> > 
> > I've a 10 HDDs RAID-6 which, due to mishap (disks
> > were disconnected accidentaly), has now 3 missing
> > devices and cannot be assembled.
> > The data should be OK, since no writes were occurring
> > during the accident, so putting them together again
> > should work.
> > 
> > As far as I know, one option is to create, with
> > "mdadm -C" the array again, giving the disks in
> > the proper order.
> > 
> > Since all HDDs are readable, I guess "mdadm -E"
> > should return the role of each device.
> > Is this correct for the creation order?
> > 
> > Second question is about the "Data Offset", since
> > this array was created with an older version of
> > "mdadm" and the data offset is very close to the
> > superblock.
> > As far as I know, new mdadm creates the data a
> > bit far aways.
> > Is there any way to specifiy the proper offset?
> > 
> > Finally, is there an alternative to "mdadm -C"
> > or it is the only option?
> > Forcing assembly does not work, but maybe there
> > is another way to tell mdadm to really assemby
> > the array, taking into account the superblock
> > information, which are all readable.
> > 
> > Thanks a lot in advance,
> > 
> > bye,
> > 
> > -- 
> > 
> > piergiorgio
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: RAID-6 with 3 missing disks

[Piergiorgio Sartor]

Hi Neil,

On Sun, May 05, 2013 at 12:58:27PM +1000, NeilBrown wrote:
> On Sat, 4 May 2013 18:30:06 +0200 Piergiorgio Sartor
> <piergiorgio.sartor@nexgo.de> wrote:
> 
> > Hi Rudy,
> > 
> > thanks for the answer, but as mentioned at the end,
> > "--force" assemby does not work.
> > Reason is, 7 disks complains 3 are missing and the
> > 3 missing are assembed, since their superblock does
> > not report errors.
> > Of course, 3 disks are not enough to assembly the
> > array, forced or not.
> 
> Details please.  "--examine" output of every device would be a good start.
> Output for "mdadm --assemble --force --verbose ....."
> would help too.

thanks for the answer.
As mentioned in a sequent post, I fixed it by hex editing
the superblock and restoring, in the 7 disks, the other 3
missing.

Anyway, I did not considered the "--verbose" option, thanks
for the hint.

Thanks again,

bye,

pg

> 
> NeilBrown
> 
> 
> > 
> > bye,
> > 
> > pg
> > 
> > On Sat, May 04, 2013 at 04:22:39PM +0000, Rudy Zijlstra wrote:
> > > Hi
> > > 
> > > I would start with mdadm assemble --force
> > > 
> > > Do not use create unless all else has failed
> > > 
> > > Cheers
> > > 
> > > Rudy
> > > ---
> > > Verstuurd met mijn BlackBerry van Vodafone
> > > 
> > > -----Original Message-----
> > > From: Piergiorgio Sartor <piergiorgio.sartor@nexgo.de>
> > > Sender: linux-raid-owner@vger.kernel.org
> > > Date:	Sat, 4 May 2013 18:08:04 
> > > To: <linux-raid@vger.kernel.org>
> > > Subject: RAID-6 with 3 missing disks
> > > 
> > > Hi all,
> > > 
> > > I know this was probably already discussed, but
> > > maybe I need some refresh.
> > > 
> > > I've a 10 HDDs RAID-6 which, due to mishap (disks
> > > were disconnected accidentaly), has now 3 missing
> > > devices and cannot be assembled.
> > > The data should be OK, since no writes were occurring
> > > during the accident, so putting them together again
> > > should work.
> > > 
> > > As far as I know, one option is to create, with
> > > "mdadm -C" the array again, giving the disks in
> > > the proper order.
> > > 
> > > Since all HDDs are readable, I guess "mdadm -E"
> > > should return the role of each device.
> > > Is this correct for the creation order?
> > > 
> > > Second question is about the "Data Offset", since
> > > this array was created with an older version of
> > > "mdadm" and the data offset is very close to the
> > > superblock.
> > > As far as I know, new mdadm creates the data a
> > > bit far aways.
> > > Is there any way to specifiy the proper offset?
> > > 
> > > Finally, is there an alternative to "mdadm -C"
> > > or it is the only option?
> > > Forcing assembly does not work, but maybe there
> > > is another way to tell mdadm to really assemby
> > > the array, taking into account the superblock
> > > information, which are all readable.
> > > 
> > > Thanks a lot in advance,
> > > 
> > > bye,
> > > 
> > > -- 
> > > 
> > > piergiorgio
> > > --
> > > To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> > > the body of a message to majordomo@vger.kernel.org
> > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> 



-- 

piergiorgio