Setting a buffer_heads 'real_device'

Setting a buffer_heads 'real_device'

[Mario Vodisek]

[-- Attachment #1: Type: text/plain, Size: 611 bytes --]

Hi,

I'm writing a virtual disk driver consisting of serveral real disks.
Hence I've to map read/write requests for my virtual device to
underlying
real devices, like md does so for RAID-Arrays.

So my question is how a virtual device driver (like 'md') manages the
read/write
request associated with a virtual (md-) device to get data from a
specific real
device within a 'block_read/block/write()' invokation !
I know that the buffer head-structure contains an entry for a virtual
and a real
device, but i don't know when and in which proper way to set these
components.

Somebody to help me ??

Mario Vodisek

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: Visitenkarte für Mario Vodisek --]
[-- Type: text/x-vcard; charset=us-ascii; name="vodisek.vcf", Size: 363 bytes --]

begin:vcard 
n:Vodisek;Mario
tel;fax:---
tel;home:0160/632 96 92
tel;work:---
x-mozilla-html:FALSE
org:University of Paderborn;AG Meyer auf der Heide - Theoretical computing -
version:2.1
email;internet:vodisek@upb.de
title:research assistant
adr;quoted-printable:;;Liebfrauenweg 9=0D=0A	;33334 Gütersloh;;;Germany
x-mozilla-cpt:;-9984
fn:Mario Vodisek
end:vcard

Multiple disk failure - recover?

[Danilo Godec]

Hi!

I have (had) a three disk RAID5 array. One of the disks (sdb2) has failed,
however, under circumstances, two of the diskes were kicked out of the
array.

I tried re-assembling the array using 'mdadm --assemble --force /dev/md1
/dev/sda2 /dev/sdc2 /dev/sdb2', but it always takes the sdc2 & sdb2 as OK
disks and sda2 as failed.

Of course, when the reconstruction hits the bad area on sdb2, it all fails
again.

How can I tell mdadm that /dev/sda2 and /dev/sdc2 are the disks I want as
'good'!?

Thanks, D.

Re: Multiple disk failure - recover?

[Mike Black]

You don't say why two disks were kicked or when it happend so sdc may be out-of-whack a bit.  Sounds like sdc got kicked first.
So...I would do a bad-block scan on sdb2 so it can remap the bad section and re-build as you've been trying to do.

If you REALLY want to use sda and sdc just rebuild the array with --create and leave sdb out of it.  You shouldn't lose much data if
sdc is current.
----- Original Message -----
From: "Danilo Godec" <danci@agenda.si>
To: <linux-raid@vger.kernel.org>
Sent: Wednesday, July 10, 2002 9:10 AM
Subject: Multiple disk failure - recover?


> Hi!
>
> I have (had) a three disk RAID5 array. One of the disks (sdb2) has failed,
> however, under circumstances, two of the diskes were kicked out of the
> array.
>
> I tried re-assembling the array using 'mdadm --assemble --force /dev/md1
> /dev/sda2 /dev/sdc2 /dev/sdb2', but it always takes the sdc2 & sdb2 as OK
> disks and sda2 as failed.
>
> Of course, when the reconstruction hits the bad area on sdb2, it all fails
> again.
>
> How can I tell mdadm that /dev/sda2 and /dev/sdc2 are the disks I want as
> 'good'!?
>
> Thanks, D.
>
>
>
> -
> 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: Multiple disk failure - recover?

[Neil Brown]

On Wednesday July 10, mblack@csi-inc.com wrote:
> You don't say why two disks were kicked or when it happend so sdc may be out-of-whack a bit.  Sounds like sdc got kicked first.
> So...I would do a bad-block scan on sdb2 so it can remap the bad section and re-build as you've been trying to do.
> 
> If you REALLY want to use sda and sdc just rebuild the array with --create and leave sdb out of it.  You shouldn't lose much data if
> sdc is current.

Better still, do that --assemble and leave sdb out of it.
i.e.

 mdadm --assemble --force /dev/md1 /dev/sda2 /dev/sdc2

Then it will have to use a2 and c2.

NeilBrown

Re: Multiple disk failure - recover?

[Danilo Godec]

On Wed, 10 Jul 2002, Mike Black wrote:

> You don't say why two disks were kicked or when it happend so sdc may
> be out-of-whack a bit.  Sounds like sdc got kicked first.

I'm not 100% sure, but I think either of sda or sdb was the first
kicked... Dunno.

> So...I would do a bad-block scan on sdb2 so it can remap the bad
> section and re-build as you've been trying to do.

What tool would you use? I think that linux' badblocks operates on the FS
level... So it's useless on Software RAID5...

>
> If you REALLY want to use sda and sdc just rebuild the array with --create and leave sdb out of it.  You shouldn't lose much data if
> sdc is current.

I tried that and it simply ignores /dev/sda2...

I only need to recover as much data as possible. It's not that bad if I
loose yesterdays stuff.

   D.

Re: Multiple disk failure - recover?

[Danilo Godec]

On Wed, 10 Jul 2002, Neil Brown wrote:

>  mdadm --assemble --force /dev/md1 /dev/sda2 /dev/sdc2
>
> Then it will have to use a2 and c2.

I tried that, but it refused to do it (something about 2/3 drives
failed)!? Maybe my mdadm version is old (v1.0.0)?

   D.

Re: Multiple disk failure - recover?

[Neil Brown]

On Wednesday July 10, danci@agenda.si wrote:
> On Wed, 10 Jul 2002, Neil Brown wrote:
> 
> >  mdadm --assemble --force /dev/md1 /dev/sda2 /dev/sdc2
> >
> > Then it will have to use a2 and c2.
> 
> I tried that, but it refused to do it (something about 2/3 drives
> failed)!? Maybe my mdadm version is old (v1.0.0)?

Hmmm. Should work, but might need a "--run" as well...

Can you get me an exact error message, and maybe the output of
 mdadm --examine /dev/sd[abc]2

NeilBrown

Re: Multiple disk failure - recover?

[Danilo Godec]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 698 bytes --]

On Wed, 10 Jul 2002, Neil Brown wrote:

> On Wednesday July 10, danci@agenda.si wrote:
> > On Wed, 10 Jul 2002, Neil Brown wrote:
> >
> > >  mdadm --assemble --force /dev/md1 /dev/sda2 /dev/sdc2
> > >
> > > Then it will have to use a2 and c2.
> >
> > I tried that, but it refused to do it (something about 2/3 drives
> > failed)!? Maybe my mdadm version is old (v1.0.0)?
>
> Hmmm. Should work, but might need a "--run" as well...
>
> Can you get me an exact error message, and maybe the output of
>  mdadm --examine /dev/sd[abc]2
>
> NeilBrown
>

OK, it's all attached. One file contains results of mdadm --examine, the
other is what I get on the screen with mdadm --assemble (w/wo --run)...

  D.

[-- Attachment #2: Type: TEXT/PLAIN, Size: 1086 bytes --]

[root@lmq MDADM]# mdadm --assemble --force /dev/md1 /dev/sda2 /dev/sdc2
md: md1 stopped.
md: unbind<sdc2,0>
md: export_rdev(sdc2)
 [events: 00000058]
md: bind<sdc2,1>
mdadm: /dev/md1 assembled from 1 drive - not enough to start it (use --run
to insist).

[root@lmq MDADM]# mdadm --assemble --force /dev/md1 --run /dev/sda2
/dev/sdc2
md: md1 stopped.
md: unbind<sdc2,0>
md: export_rdev(sdc2)
 [events: 00000058]
md: bind<sdc2,1>
md: sdc2's event counter: 00000058
md1: removing former faulty sdb2!
md1: max total readahead window set to 496k
md1: 2 data-disks, max readahead per data-disk: 248k
raid5: device sdc2 operational as raid disk 1
raid5: not enough operational devices for md1 (2/3 failed)
RAID5 conf printout:
 --- rd:3 wd:1 fd:2
 disk 0, s:0, o:0, n:0 rd:0 us:1 dev:[dev 00:00]
 disk 1, s:0, o:1, n:1 rd:1 us:1 dev:sdc2
 disk 2, s:0, o:0, n:2 rd:2 us:1 dev:[dev 00:00]
raid5: failed to run raid set md1
md: pers->run() failed ...
md: md1 stopped.
md: unbind<sdc2,0>
md: export_rdev(sdc2)
mdadm: failed to RUN_ARRAY /dev/md1: Invalid argument

[-- Attachment #3: Type: TEXT/PLAIN, Size: 2748 bytes --]

/dev/sda2:
          Magic : a92b4efc
        Version : 00.90.00
           UUID : d1b50184:535ff73c:48dba5c1:e94fc70e
  Creation Time : Mon Jul 16 13:53:51 2001
     Raid Level : raid5
    Device Size : 17623232 (16.80 GiB 18.04 GB)
   Raid Devices : 3
  Total Devices : 4
Preferred Minor : 1

    Update Time : Wed Jul 10 15:18:37 2002
          State : dirty, no-errors
 Active Devices : 2
Working Devices : 3
 Failed Devices : 1
  Spare Devices : 1
       Checksum : 79f828e5 - correct
         Events : 0.82

         Layout : left-asymmetric
     Chunk Size : 32K

      Number   Major   Minor   RaidDevice State
this     3       8        2        3        /dev/sda2
   0     0       8       18        0      active sync   /dev/sdb2
   1     1       8       34        1      active sync   /dev/sdc2
   2     2       0        0        2      faulty removed
   3     3       8        2        3        /dev/sda2

/dev/sdb2:
          Magic : a92b4efc
        Version : 00.90.00
           UUID : d1b50184:535ff73c:48dba5c1:e94fc70e
  Creation Time : Mon Jul 16 13:53:51 2001
     Raid Level : raid5
    Device Size : 17623232 (16.80 GiB 18.04 GB)
   Raid Devices : 3
  Total Devices : 3
Preferred Minor : 1

    Update Time : Wed Jul 10 15:36:39 2002
          State : dirty, no-errors
 Active Devices : 2
Working Devices : 2
 Failed Devices : 1
  Spare Devices : 0
       Checksum : 79f82d2f - correct
         Events : 0.86

         Layout : left-asymmetric
     Chunk Size : 32K

      Number   Major   Minor   RaidDevice State
this     0       8       18        0      active sync   /dev/sdb2
   0     0       8       18        0      active sync   /dev/sdb2
   1     1       8       34        1      active sync   /dev/sdc2
   2     2       0        0        2      faulty removed

/dev/sdc2:
          Magic : a92b4efc
        Version : 00.90.00
           UUID : d1b50184:535ff73c:48dba5c1:e94fc70e
  Creation Time : Mon Jul 16 13:53:51 2001
     Raid Level : raid5
    Device Size : 17623232 (16.80 GiB 18.04 GB)
   Raid Devices : 3
  Total Devices : 3
Preferred Minor : 1

    Update Time : Wed Jul 10 16:08:35 2002
          State : clean, no-errors
 Active Devices : 1
Working Devices : 1
 Failed Devices : 2
  Spare Devices : 0
       Checksum : 79f834ba - correct
         Events : 0.88

         Layout : left-asymmetric
     Chunk Size : 32K

      Number   Major   Minor   RaidDevice State
this     1       8       34        1      active sync   /dev/sdc2
   0     0       8       18        0      faulty   /dev/sdb2
   1     1       8       34        1      active sync   /dev/sdc2
   2     2       0        0        2      faulty removed

Re: Multiple disk failure - recover?

[Neil Brown]

On Wednesday July 10, danci@agenda.si wrote:
> OK, it's all attached. One file contains results of mdadm --examine, the
> other is what I get on the screen with mdadm --assemble (w/wo
--run)...

Hmmm.  Yes --force isn't doing all that it should do. I'll have to
look into that.

You should be able to use create successfully.

If the original order was b2,c2,a2, which is what the --examine
suggests, and you want to not use b2, then

 mdadm --create /dev/md1 -l5 -c32 -n3 missing /dev/sdc2 /dev/sda2

should work.  It will not start a rebuild so you can mount it
read-only and if something was wrong, you can "mdadm --stop" and try
again with a different arrangement of drives.

If b2 and c2 are really the most recent, as it appears, then you might
want to use those:

  mdadm --create /dev/md1 -l5 -c32 -n3 /dev/sdb2 /dev/sdc2 missing

Again, this wont do a resync and as long as you don't touch the bad
block on b2, you should be able to access data.

NeilBrown