Big trouble during reassemble a Raid5

Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello, 

i'm sorry to ask this questions but the raid 5 with 4 disk is in big trouble during re-assemble. 2 disks are out of order. 
I have change a disk of the raid 5 (sde) to growing the raid. 
But a second disk (sdc) have too many bad sector during the re-assemble, and shutdown the re-assemble. 
"mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" 
I have try to correct bad sectors with badblocks, but it's finished by no more spare sectors and the disk still have some bad sector. 
badblocks -b 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000 1140169336 
1140169400 
1140169401 
1140169402 
1140169403 
1140169404 
1140169405 
1140169406 
1140169407 
1140169416 
1140169417 
1140169418 
1140169419 
1140169420 
1140169421 
1140169422 
1140169423 


For information the mdadm examine return : 
cat mdadm-exam.txt 
/dev/sdb: 
MBR Magic : aa55 
Partition[0] : 1953523120 sectors at 2048 (type fd) 
/dev/sdc: 
MBR Magic : aa55 
Partition[0] : 1953520002 sectors at 63 (type fd) 
/dev/sdd: 
MBR Magic : aa55 
Partition[0] : 1953520002 sectors at 63 (type fd) 
/dev/sde: 
MBR Magic : aa55 
Partition[0] : 4294965247 sectors at 2048 (type fd) 

I have 2 way to solve the issue. 
The first, is to have special command to pass bad sector during re-assemble as "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" 
The second is change the disk sde with the old good one, but some datas have been changed on the raid since i have remove it. But these datas are not important. It's only logs and history activity. 

What can i do to recover a maximum datas without too many risk? 

Thank's in advance 
Best Regards 
---------------------------------- 
Sylvain Depuille (in trouble) 

sylvain.depuille@laposte.net 

Re: Big trouble during reassemble a Raid5

[John Stoffel]

Sylvain,

I would recommend that you buy a replacement disk for the one
throwing errors and then run dd_rescue to copy as much data from the
dying disk to the replacement.  Then, and only then, do you try to
reassemble the array with the --force option.

That disk is dying, and dying quickly.

Can you also post the output of 

mdadm -E /dev/sd[bcde]1 for each disk, even the dying one, so we can
look at the counts and give you some more advice.  Also, the output of
the mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 would also be
good.  The more info the better.

Good luck!
John


sylvain> i'm sorry to ask this questions but the raid 5 with 4 disk is
sylvain> in big trouble during re-assemble. 2 disks are out of order.
sylvain> I have change a disk of the raid 5 (sde) to growing the raid.
sylvain> But a second disk (sdc) have too many bad sector during the
sylvain> re-assemble, and shutdown the re-assemble.  "mdadm --assemble
sylvain> --force /dev/md2 /dev/sd[bcde]1" I have try to correct bad
sylvain> sectors with badblocks, but it's finished by no more spare
sylvain> sectors and the disk still have some bad sector.  badblocks
sylvain> -b 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000
sylvain> 1140169336 1140169400 1140169401 1140169402 1140169403
sylvain> 1140169404 1140169405 1140169406 1140169407 1140169416
sylvain> 1140169417 1140169418 1140169419 1140169420 1140169421
sylvain> 1140169422 1140169423


sylvain> For information the mdadm examine return : cat mdadm-exam.txt
sylvain> /dev/sdb: MBR Magic : aa55 Partition[0] : 1953523120 sectors
sylvain> at 2048 (type fd) /dev/sdc: MBR Magic : aa55 Partition[0] :
sylvain> 1953520002 sectors at 63 (type fd) /dev/sdd: MBR Magic : aa55
sylvain> Partition[0] : 1953520002 sectors at 63 (type fd) /dev/sde:
sylvain> MBR Magic : aa55 Partition[0] : 4294965247 sectors at 2048
sylvain> (type fd)

sylvain> I have 2 way to solve the issue.  The first, is to have
sylvain> special command to pass bad sector during re-assemble as
sylvain> "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" The second
sylvain> is change the disk sde with the old good one, but some datas
sylvain> have been changed on the raid since i have remove it. But
sylvain> these datas are not important. It's only logs and history
sylvain> activity.

sylvain> What can i do to recover a maximum datas without too many
sylvain> risk?

sylvain> Thank's in advance Best Regards
sylvain> ---------------------------------- Sylvain Depuille (in
sylvain> trouble)

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

Re: Re : Re: Big trouble during reassemble a Raid5

[John Stoffel]

sylvain> Hi john, thanks for your answer!  I have change a 1TB disk to
sylvain> growing the raid with 3TB disk.  if i can re-insert the old
sylvain> 1TB disk in place of 3TB disk, only some log and history are
sylvain> corrupted.  i think that is the best way to relaunch the raid
sylvain> without data loss.  But i dont known how change the timestamp
sylvain> of the one raid disk.  Have you a magic command to change a
sylvain> timestamp of a raid partition, and how known the timestamp of
sylvain> the other disk of the raid?  After' raid relaunch, i can
sylvain> change the burn disk by a 3TB new one.  To do the ddrescue, i
sylvain> have a 2TB disk spare! Its not the same geometry, is it
sylvain> possible?  thanks in advance for your help

Sylvain,

Always glad to help here.  I'm going to try and understand what you
wrote and do my best to reply.  

Is the 1Tb disk the bad disk?  And if you re-insert it and re-start
the RAID5 array, you only have some minor lost files?  If so, I would
probably just copy all the data off the RAID5 onto the single 3Tb disk
as a quick and dirty backup, then I'd use 'dd_rescue' to copy the bad
1Tb disk onto the new 2Tb disk.

All you would have to do is make a partition on the 2tb disk which is
the same size (or a little bigger) than the partition on the 1tb disk,
then copy the partition over like this:

  ddrescue /dev/sd[BAD DISK LETTER HERE]1 /dev/sd[2TB disk letter]1 \ 
  /tmp/rescue.log

So say the bad disk is sdc, and the good 2tb is sdf, you would do:

  ddrescue /dev/sdc1 /dev/sdf1 /tmp/rescue.log

and let it go.  Then you would assemble the array using the NEW 2tb
disk.  Ideally you would remove the bad 1tb disk from the system when
trying to do this.

But you really do need send us the output of the following commands:

  cat /proc/mdstat 
  cat /proc/partitions
  mdadm --detail /dev/md#  

    do the above for the RADI5 array.

  mdadm --examine /dev/sd#1

    for each disk in the RAID5 array.

And we can give you better advice.  

Good luck!


sylvain> ---------------------------------- Sylvain Depuille
sylvain> sylvain.depuille@laposte.net ----- Mail d'origine ----- De:
sylvain> John Stoffel <john@stoffel.org> À: sylvain depuille
sylvain> <sylvain.depuille@laposte.net> Cc: linux-raid@vger.kernel.org
sylvain> Envoyé: Mon, 29 Dec 2014 19:32:04 +0100 (CET) Objet: Re: Big
sylvain> trouble during reassemble a Raid5

sylvain> Sylvain, I would recommend that you buy a replacement disk
sylvain> for the one throwing errors and then run dd_rescue to copy as
sylvain> much data from the dying disk to the replacement. Then, and
sylvain> only then, do you try to reassemble the array with the
sylvain> --force option.  That disk is dying, and dying quickly.  Can
sylvain> you also post the output of mdadm -E /dev/sd[bcde]1 for each
sylvain> disk, even the dying one, so we can look at the counts and
sylvain> give you some more advice. Also, the output of the mdadm
sylvain> --assemble --force /dev/md2 /dev/sd[bcde]1 would also be
sylvain> good. The more info the better.  Good luck!  John

sylvain> i'm sorry to ask this questions but the raid 5 with 4 disk is
sylvain> in big trouble during re-assemble. 2 disks are out of order.
sylvain> I have change a disk of the raid 5 (sde) to growing the raid.
sylvain> But a second disk (sdc) have too many bad sector during the
sylvain> re-assemble, and shutdown the re-assemble. "mdadm --assemble
sylvain> --force /dev/md2 /dev/sd[bcde]1" I have try to correct bad
sylvain> sectors with badblocks, but it's finished by no more spare
sylvain> sectors and the disk still have some bad sector. badblocks -b
sylvain> 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000 1140169336
sylvain> 1140169400 1140169401 1140169402 1140169403 1140169404
sylvain> 1140169405 1140169406 1140169407 1140169416 1140169417
sylvain> 1140169418 1140169419 1140169420 1140169421 1140169422
sylvain> 1140169423

sylvain> For information the mdadm examine return : cat mdadm-exam.txt
sylvain> /dev/sdb: MBR Magic : aa55 Partition[0] : 1953523120 sectors
sylvain> at 2048 (type fd) /dev/sdc: MBR Magic : aa55 Partition[0] :
sylvain> 1953520002 sectors at 63 (type fd) /dev/sdd: MBR Magic : aa55
sylvain> Partition[0] : 1953520002 sectors at 63 (type fd) /dev/sde:
sylvain> MBR Magic : aa55 Partition[0] : 4294965247 sectors at 2048
sylvain> (type fd) I have 2 way to solve the issue. The first, is to
sylvain> have special command to pass bad sector during re-assemble as
sylvain> "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" The second
sylvain> is change the disk sde with the old good one, but some datas
sylvain> have been changed on the raid since i have remove it. But
sylvain> these datas are not important. It's only logs and history
sylvain> activity.  What can i do to recover a maximum datas without
sylvain> too many risk?  Thank's in advance Best Regards
sylvain> ---------------------------------- Sylvain Depuille (in
sylvain> trouble) sylvain.depuille@laposte.net -- To unsubscribe from
sylvain> this list: send the line "unsubscribe linux-raid" in the body
sylvain> of a message to majordomo@vger.kernel.org More majordomo info
sylvain> at http://vger.kernel.org/majordomo-info.html

--
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: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hi John,

I'm sorry! I'm French and my English is poor!
Yesterday, i wrote the mail on a tablett, but my mail's provider doesn't offer plain text option on mobile site :-(

I'm back on a PC under Linux :-) (not crash this one, but not on the same site, connected by ssh to the broke system).
For the history, i have installed PC of many friends under Linux (no more Virus problem and Windows Blue Screen).

Yesterday, i have try to you explain what is the history of the issue.
The Raid was fully operational before i change the disk sde by a 3TB disk to grow the Raid, by changing all disks of the raid. 
But when i have do the re-assemble the disk sdc goes burny :-(

I have keep the old sde 1TB disk in state.

The result of your asked command :

 cat /proc/mdstat
Personalities : [raid1] [raid6] [raid5] [raid4] 
md2 : inactive sdc1[2](S) sdd1[3](S) sde1[4](S) sdb1[5](S)
      5077760089 blocks super 1.2

major minor  #blocks  name

   8        0  125034840 sda
   8        1     131072 sda1
   8        2   62914560 sda2
   8        3   61865984 sda3
   8       32  976762584 sdc
   8       33  976760001 sdc1
   8       48  976762584 sdd
   8       49  976760001 sdd1
   8       16  976762584 sdb
   8       17  976761560 sdb1
   8       64 2930266584 sde
   8       65 2147482623 sde1

mdadm --detail /dev/md2
/dev/md2:
        Version : 1.2
     Raid Level : raid0
  Total Devices : 4
    Persistence : Superblock is persistent

          State : inactive

           Name : le-bohec:2  (local to host le-bohec)
           UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
         Events : 167456

    Number   Major   Minor   RaidDevice

       -       8       17        -        /dev/sdb1
       -       8       33        -        /dev/sdc1
       -       8       49        -        /dev/sdd1
       -       8       65        -        /dev/sde1


mdadm -E /dev/sd[bcde]1
/dev/sdb1:
          Magic : a92b4efc
        Version : 1.2
    Feature Map : 0x1
     Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
           Name : le-bohec:2  (local to host le-bohec)
  Creation Time : Tue Apr  9 17:56:19 2013
     Raid Level : raid5
   Raid Devices : 4

 Avail Dev Size : 1953521072 (931.51 GiB 1000.20 GB)
     Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
  Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
    Data Offset : 2048 sectors
   Super Offset : 8 sectors
   Unused Space : before=1960 sectors, after=3504 sectors
          State : clean
    Device UUID : 8506e09c:b87a44ed:7b4ee314:777ce89c

Internal Bitmap : 8 sectors from superblock
    Update Time : Sat Dec 27 22:08:34 2014
  Bad Block Log : 512 entries available at offset 72 sectors
       Checksum : bad62d22 - correct
         Events : 167456

         Layout : left-symmetric
     Chunk Size : 512K

   Device Role : Active device 0
   Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing)
/dev/sdc1:
          Magic : a92b4efc
        Version : 1.2
    Feature Map : 0x1
     Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
           Name : le-bohec:2  (local to host le-bohec)
  Creation Time : Tue Apr  9 17:56:19 2013
     Raid Level : raid5
   Raid Devices : 4

 Avail Dev Size : 1953517954 (931.51 GiB 1000.20 GB)
     Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
  Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
    Data Offset : 2048 sectors
   Super Offset : 8 sectors
   Unused Space : before=1968 sectors, after=386 sectors
          State : clean
    Device UUID : 44002aad:d3e17729:a93854eb:4139972e

Internal Bitmap : 8 sectors from superblock
    Update Time : Sat Dec 27 22:08:22 2014
       Checksum : 6f69285d - correct
         Events : 167431

         Layout : left-symmetric
     Chunk Size : 512K

   Device Role : Active device 1
   Array State : AAAA ('A' == active, '.' == missing, 'R' == replacing)
/dev/sdd1:
          Magic : a92b4efc
        Version : 1.2
    Feature Map : 0x1
     Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
           Name : le-bohec:2  (local to host le-bohec)
  Creation Time : Tue Apr  9 17:56:19 2013
     Raid Level : raid5
   Raid Devices : 4

 Avail Dev Size : 1953517954 (931.51 GiB 1000.20 GB)
     Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
  Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
    Data Offset : 2048 sectors
   Super Offset : 8 sectors
   Unused Space : before=1968 sectors, after=386 sectors
          State : clean
    Device UUID : 5cff6f7f:ea6b89b6:28e4e8b3:7a2b5a7f

Internal Bitmap : 8 sectors from superblock
    Update Time : Sat Dec 27 22:08:34 2014
       Checksum : e81a4f63 - correct
         Events : 167456

         Layout : left-symmetric
     Chunk Size : 512K

   Device Role : Active device 2
   Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing)
/dev/sde1:
          Magic : a92b4efc
        Version : 1.2
    Feature Map : 0x9
     Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
           Name : le-bohec:2  (local to host le-bohec)
  Creation Time : Tue Apr  9 17:56:19 2013
     Raid Level : raid5
   Raid Devices : 4

 Avail Dev Size : 4294963199 (2048.00 GiB 2199.02 GB)
     Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
  Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
    Data Offset : 2048 sectors
   Super Offset : 8 sectors
   Unused Space : before=1960 sectors, after=2341445631 sectors
          State : clean
    Device UUID : 0ebce28d:1a792d55:76a86538:12cc94dd

Internal Bitmap : 8 sectors from superblock
    Update Time : Sat Dec 27 22:08:34 2014
  Bad Block Log : 512 entries available at offset 72 sectors - bad blocks present.
       Checksum : 3801cfa - correct
         Events : 167456

         Layout : left-symmetric
     Chunk Size : 512K

   Device Role : spare
   Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing)


Now, i have installed ddrescue on the system, and i go to my friends to add the 2TB disk in the tower, and launch the command.

Many thank's for your help and your patience.
Best Regards
Sylvain Depuille

----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
Envoyé: Lundi 29 Décembre 2014 21:36:23 
Objet: Re: Re : Re: Big trouble during reassemble a Raid5 


sylvain> Hi john, thanks for your answer! I have change a 1TB disk to 
sylvain> growing the raid with 3TB disk. if i can re-insert the old 
sylvain> 1TB disk in place of 3TB disk, only some log and history are 
sylvain> corrupted. i think that is the best way to relaunch the raid 
sylvain> without data loss. But i dont known how change the timestamp 
sylvain> of the one raid disk. Have you a magic command to change a 
sylvain> timestamp of a raid partition, and how known the timestamp of 
sylvain> the other disk of the raid? After' raid relaunch, i can 
sylvain> change the burn disk by a 3TB new one. To do the ddrescue, i 
sylvain> have a 2TB disk spare! Its not the same geometry, is it 
sylvain> possible? thanks in advance for your help 

Sylvain, 

Always glad to help here. I'm going to try and understand what you 
wrote and do my best to reply. 

Is the 1Tb disk the bad disk? And if you re-insert it and re-start 
the RAID5 array, you only have some minor lost files? If so, I would 
probably just copy all the data off the RAID5 onto the single 3Tb disk 
as a quick and dirty backup, then I'd use 'dd_rescue' to copy the bad 
1Tb disk onto the new 2Tb disk. 

All you would have to do is make a partition on the 2tb disk which is 
the same size (or a little bigger) than the partition on the 1tb disk, 
then copy the partition over like this: 

ddrescue /dev/sd[BAD DISK LETTER HERE]1 /dev/sd[2TB disk letter]1 \ 
/tmp/rescue.log 

So say the bad disk is sdc, and the good 2tb is sdf, you would do: 

ddrescue /dev/sdc1 /dev/sdf1 /tmp/rescue.log 

and let it go. Then you would assemble the array using the NEW 2tb 
disk. Ideally you would remove the bad 1tb disk from the system when 
trying to do this. 

But you really do need send us the output of the following commands: 

cat /proc/mdstat 
cat /proc/partitions 
mdadm --detail /dev/md# 

do the above for the RADI5 array. 

mdadm --examine /dev/sd#1 

for each disk in the RAID5 array. 

And we can give you better advice. 

Good luck! 


sylvain> ---------------------------------- Sylvain Depuille 
sylvain> sylvain.depuille@laposte.net ----- Mail d'origine ----- De: 
sylvain> John Stoffel <john@stoffel.org> À: sylvain depuille 
sylvain> <sylvain.depuille@laposte.net> Cc: linux-raid@vger.kernel.org 
sylvain> Envoyé: Mon, 29 Dec 2014 19:32:04 +0100 (CET) Objet: Re: Big 
sylvain> trouble during reassemble a Raid5 

sylvain> Sylvain, I would recommend that you buy a replacement disk 
sylvain> for the one throwing errors and then run dd_rescue to copy as 
sylvain> much data from the dying disk to the replacement. Then, and 
sylvain> only then, do you try to reassemble the array with the 
sylvain> --force option. That disk is dying, and dying quickly. Can 
sylvain> you also post the output of mdadm -E /dev/sd[bcde]1 for each 
sylvain> disk, even the dying one, so we can look at the counts and 
sylvain> give you some more advice. Also, the output of the mdadm 
sylvain> --assemble --force /dev/md2 /dev/sd[bcde]1 would also be 
sylvain> good. The more info the better. Good luck! John 

sylvain> i'm sorry to ask this questions but the raid 5 with 4 disk is 
sylvain> in big trouble during re-assemble. 2 disks are out of order. 
sylvain> I have change a disk of the raid 5 (sde) to growing the raid. 
sylvain> But a second disk (sdc) have too many bad sector during the 
sylvain> re-assemble, and shutdown the re-assemble. "mdadm --assemble 
sylvain> --force /dev/md2 /dev/sd[bcde]1" I have try to correct bad 
sylvain> sectors with badblocks, but it's finished by no more spare 
sylvain> sectors and the disk still have some bad sector. badblocks -b 
sylvain> 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000 1140169336 
sylvain> 1140169400 1140169401 1140169402 1140169403 1140169404 
sylvain> 1140169405 1140169406 1140169407 1140169416 1140169417 
sylvain> 1140169418 1140169419 1140169420 1140169421 1140169422 
sylvain> 1140169423 

sylvain> For information the mdadm examine return : cat mdadm-exam.txt 
sylvain> /dev/sdb: MBR Magic : aa55 Partition[0] : 1953523120 sectors 
sylvain> at 2048 (type fd) /dev/sdc: MBR Magic : aa55 Partition[0] : 
sylvain> 1953520002 sectors at 63 (type fd) /dev/sdd: MBR Magic : aa55 
sylvain> Partition[0] : 1953520002 sectors at 63 (type fd) /dev/sde: 
sylvain> MBR Magic : aa55 Partition[0] : 4294965247 sectors at 2048 
sylvain> (type fd) I have 2 way to solve the issue. The first, is to 
sylvain> have special command to pass bad sector during re-assemble as 
sylvain> "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" The second 
sylvain> is change the disk sde with the old good one, but some datas 
sylvain> have been changed on the raid since i have remove it. But 
sylvain> these datas are not important. It's only logs and history 
sylvain> activity. What can i do to recover a maximum datas without 
sylvain> too many risk? Thank's in advance Best Regards 
sylvain> ---------------------------------- Sylvain Depuille (in 
sylvain> trouble) sylvain.depuille@laposte.net -- To unsubscribe from 
sylvain> this list: send the line "unsubscribe linux-raid" in the body 
sylvain> of a message to majordomo@vger.kernel.org More majordomo info 
sylvain> at http://vger.kernel.org/majordomo-info.html 


--
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: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello John, 

the ddrescue command is finished. 
The log file :
# Rescue Logfile. Created by GNU ddrescue version 1.18.1
# Command line: ddrescue --force /dev/sdc1 /dev/sdf1 rescue.log
# Start time:   2014-12-30 11:28:57
# Current time: 2014-12-30 13:40:29
# Copying non-tried blocks... Pass 1 (forwards)
# current_pos  current_status
0x89EE5A0000     ?
#      pos        size  status
0x00000000  0x87EB34F000  +
0x87EB34F000  0x00001000  *
0x87EB350000  0x00001000  +
0x87EB351000  0x0001F000  *
0x87EB370000  0x0015B000  +
0x87EB4CB000  0x00005000  *
0x87EB4D0000  0x0BA39000  +
0x87F6F09000  0x00007000  *
0x87F6F10000  0x1F76A0000  +
0x89EE5B0000  0x5EF2580400  ?

Now, can i change the sdc by the sdf disk?

Thank's in advance for your help.
Best Regards

Sylvain Depuille
----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
Envoyé: Lundi 29 Décembre 2014 21:36:23 
Objet: Re: Re : Re: Big trouble during reassemble a Raid5 


sylvain> Hi john, thanks for your answer! I have change a 1TB disk to 
sylvain> growing the raid with 3TB disk. if i can re-insert the old 
sylvain> 1TB disk in place of 3TB disk, only some log and history are 
sylvain> corrupted. i think that is the best way to relaunch the raid 
sylvain> without data loss. But i dont known how change the timestamp 
sylvain> of the one raid disk. Have you a magic command to change a 
sylvain> timestamp of a raid partition, and how known the timestamp of 
sylvain> the other disk of the raid? After' raid relaunch, i can 
sylvain> change the burn disk by a 3TB new one. To do the ddrescue, i 
sylvain> have a 2TB disk spare! Its not the same geometry, is it 
sylvain> possible? thanks in advance for your help 

Sylvain, 

Always glad to help here. I'm going to try and understand what you 
wrote and do my best to reply. 

Is the 1Tb disk the bad disk? And if you re-insert it and re-start 
the RAID5 array, you only have some minor lost files? If so, I would 
probably just copy all the data off the RAID5 onto the single 3Tb disk 
as a quick and dirty backup, then I'd use 'dd_rescue' to copy the bad 
1Tb disk onto the new 2Tb disk. 

All you would have to do is make a partition on the 2tb disk which is 
the same size (or a little bigger) than the partition on the 1tb disk, 
then copy the partition over like this: 

ddrescue /dev/sd[BAD DISK LETTER HERE]1 /dev/sd[2TB disk letter]1 \ 
/tmp/rescue.log 

So say the bad disk is sdc, and the good 2tb is sdf, you would do: 

ddrescue /dev/sdc1 /dev/sdf1 /tmp/rescue.log 

and let it go. Then you would assemble the array using the NEW 2tb 
disk. Ideally you would remove the bad 1tb disk from the system when 
trying to do this. 

But you really do need send us the output of the following commands: 

cat /proc/mdstat 
cat /proc/partitions 
mdadm --detail /dev/md# 

do the above for the RADI5 array. 

mdadm --examine /dev/sd#1 

for each disk in the RAID5 array. 

And we can give you better advice. 

Good luck! 


sylvain> ---------------------------------- Sylvain Depuille 
sylvain> sylvain.depuille@laposte.net ----- Mail d'origine ----- De: 
sylvain> John Stoffel <john@stoffel.org> À: sylvain depuille 
sylvain> <sylvain.depuille@laposte.net> Cc: linux-raid@vger.kernel.org 
sylvain> Envoyé: Mon, 29 Dec 2014 19:32:04 +0100 (CET) Objet: Re: Big 
sylvain> trouble during reassemble a Raid5 

sylvain> Sylvain, I would recommend that you buy a replacement disk 
sylvain> for the one throwing errors and then run dd_rescue to copy as 
sylvain> much data from the dying disk to the replacement. Then, and 
sylvain> only then, do you try to reassemble the array with the 
sylvain> --force option. That disk is dying, and dying quickly. Can 
sylvain> you also post the output of mdadm -E /dev/sd[bcde]1 for each 
sylvain> disk, even the dying one, so we can look at the counts and 
sylvain> give you some more advice. Also, the output of the mdadm 
sylvain> --assemble --force /dev/md2 /dev/sd[bcde]1 would also be 
sylvain> good. The more info the better. Good luck! John 

sylvain> i'm sorry to ask this questions but the raid 5 with 4 disk is 
sylvain> in big trouble during re-assemble. 2 disks are out of order. 
sylvain> I have change a disk of the raid 5 (sde) to growing the raid. 
sylvain> But a second disk (sdc) have too many bad sector during the 
sylvain> re-assemble, and shutdown the re-assemble. "mdadm --assemble 
sylvain> --force /dev/md2 /dev/sd[bcde]1" I have try to correct bad 
sylvain> sectors with badblocks, but it's finished by no more spare 
sylvain> sectors and the disk still have some bad sector. badblocks -b 
sylvain> 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000 1140169336 
sylvain> 1140169400 1140169401 1140169402 1140169403 1140169404 
sylvain> 1140169405 1140169406 1140169407 1140169416 1140169417 
sylvain> 1140169418 1140169419 1140169420 1140169421 1140169422 
sylvain> 1140169423 

sylvain> For information the mdadm examine return : cat mdadm-exam.txt 
sylvain> /dev/sdb: MBR Magic : aa55 Partition[0] : 1953523120 sectors 
sylvain> at 2048 (type fd) /dev/sdc: MBR Magic : aa55 Partition[0] : 
sylvain> 1953520002 sectors at 63 (type fd) /dev/sdd: MBR Magic : aa55 
sylvain> Partition[0] : 1953520002 sectors at 63 (type fd) /dev/sde: 
sylvain> MBR Magic : aa55 Partition[0] : 4294965247 sectors at 2048 
sylvain> (type fd) I have 2 way to solve the issue. The first, is to 
sylvain> have special command to pass bad sector during re-assemble as 
sylvain> "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" The second 
sylvain> is change the disk sde with the old good one, but some datas 
sylvain> have been changed on the raid since i have remove it. But 
sylvain> these datas are not important. It's only logs and history 
sylvain> activity. What can i do to recover a maximum datas without 
sylvain> too many risk? Thank's in advance Best Regards 
sylvain> ---------------------------------- Sylvain Depuille (in 
sylvain> trouble) sylvain.depuille@laposte.net -- To unsubscribe from 
sylvain> this list: send the line "unsubscribe linux-raid" in the body 
sylvain> of a message to majordomo@vger.kernel.org More majordomo info 
sylvain> at http://vger.kernel.org/majordomo-info.html 


--
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: Re: Re : Re: Big trouble during reassemble a Raid5

[John Stoffel]

>>>>> "sylvain" == sylvain depuille <sylvain.depuille@laposte.net> writes:

sylvain> Hi John,
sylvain> I'm sorry! I'm French and my English is poor!

No problems, your english is much better than I could ever do in
French!  

sylvain> Yesterday, i have try to you explain what is the history of
sylvain> the issue.  The Raid was fully operational before i change
sylvain> the disk sde by a 3TB disk to grow the Raid, by changing all
sylvain> disks of the raid.  But when i have do the re-assemble the
sylvain> disk sdc goes burny :-(

sylvain> I have keep the old sde 1TB disk in state.

sylvain> The result of your asked command :

sylvain> cat /proc/mdstat
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> md2 : inactive sdc1[2](S) sdd1[3](S) sde1[4](S) sdb1[5](S)
sylvain> 5077760089 blocks super 1.2

Ok, this looks like the raid won't assemble, which is probably fine
for now. 

sylvain> major minor  #blocks  name

sylvain> 8        0  125034840 sda
sylvain> 8        1     131072 sda1
sylvain> 8        2   62914560 sda2
sylvain> 8        3   61865984 sda3
sylvain> 8       32  976762584 sdc
sylvain> 8       33  976760001 sdc1
sylvain> 8       48  976762584 sdd
sylvain> 8       49  976760001 sdd1
sylvain> 8       16  976762584 sdb
sylvain> 8       17  976761560 sdb1
sylvain> 8       64 2930266584 sde
sylvain> 8       65 2147482623 sde1

Did the reshape or replacement of /dev/sde 1tb work fine to the 3tb
disk?  It's not clear to me, and I want to be safe.  But I think you
should be able to do the --assemble --force command once the dd_rescue
has finished from the bad disk onto the new disk.  

Then I would also seriously think about moving to RAID6 for your data
as well.  It's cheap insurance when you have lots of important data
and such to keep.

sylvain> mdadm --detail /dev/md2
sylvain> /dev/md2:
sylvain> Version : 1.2
sylvain> Raid Level : raid0
sylvain> Total Devices : 4
sylvain> Persistence : Superblock is persistent

sylvain> State : inactive

sylvain> Name : le-bohec:2  (local to host le-bohec)
sylvain> UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
sylvain> Events : 167456

sylvain> Number   Major   Minor   RaidDevice

sylvain> -       8       17        -        /dev/sdb1
sylvain> -       8       33        -        /dev/sdc1
sylvain> -       8       49        -        /dev/sdd1
sylvain> -       8       65        -        /dev/sde1


sylvain> mdadm -E /dev/sd[bcde]1
sylvain> /dev/sdb1:
sylvain> Magic : a92b4efc
sylvain> Version : 1.2
sylvain> Feature Map : 0x1
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
sylvain> Name : le-bohec:2  (local to host le-bohec)
sylvain> Creation Time : Tue Apr  9 17:56:19 2013
sylvain> Raid Level : raid5
sylvain> Raid Devices : 4

sylvain> Avail Dev Size : 1953521072 (931.51 GiB 1000.20 GB)
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
sylvain> Data Offset : 2048 sectors
sylvain> Super Offset : 8 sectors
sylvain> Unused Space : before=1960 sectors, after=3504 sectors
sylvain> State : clean
sylvain> Device UUID : 8506e09c:b87a44ed:7b4ee314:777ce89c

sylvain> Internal Bitmap : 8 sectors from superblock
sylvain> Update Time : Sat Dec 27 22:08:34 2014
sylvain> Bad Block Log : 512 entries available at offset 72 sectors
sylvain> Checksum : bad62d22 - correct
sylvain> Events : 167456

sylvain> Layout : left-symmetric
sylvain> Chunk Size : 512K

sylvain> Device Role : Active device 0
sylvain> Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing)
sylvain> /dev/sdc1:
sylvain> Magic : a92b4efc
sylvain> Version : 1.2
sylvain> Feature Map : 0x1
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
sylvain> Name : le-bohec:2  (local to host le-bohec)
sylvain> Creation Time : Tue Apr  9 17:56:19 2013
sylvain> Raid Level : raid5
sylvain> Raid Devices : 4

sylvain> Avail Dev Size : 1953517954 (931.51 GiB 1000.20 GB)
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
sylvain> Data Offset : 2048 sectors
sylvain> Super Offset : 8 sectors
sylvain> Unused Space : before=1968 sectors, after=386 sectors
sylvain> State : clean
sylvain> Device UUID : 44002aad:d3e17729:a93854eb:4139972e

sylvain> Internal Bitmap : 8 sectors from superblock
sylvain> Update Time : Sat Dec 27 22:08:22 2014
sylvain> Checksum : 6f69285d - correct
sylvain> Events : 167431

sylvain> Layout : left-symmetric
sylvain> Chunk Size : 512K

sylvain> Device Role : Active device 1
sylvain> Array State : AAAA ('A' == active, '.' == missing, 'R' == replacing)
sylvain> /dev/sdd1:
sylvain> Magic : a92b4efc
sylvain> Version : 1.2
sylvain> Feature Map : 0x1
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
sylvain> Name : le-bohec:2  (local to host le-bohec)
sylvain> Creation Time : Tue Apr  9 17:56:19 2013
sylvain> Raid Level : raid5
sylvain> Raid Devices : 4

sylvain> Avail Dev Size : 1953517954 (931.51 GiB 1000.20 GB)
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
sylvain> Data Offset : 2048 sectors
sylvain> Super Offset : 8 sectors
sylvain> Unused Space : before=1968 sectors, after=386 sectors
sylvain> State : clean
sylvain> Device UUID : 5cff6f7f:ea6b89b6:28e4e8b3:7a2b5a7f

sylvain> Internal Bitmap : 8 sectors from superblock
sylvain> Update Time : Sat Dec 27 22:08:34 2014
sylvain> Checksum : e81a4f63 - correct
sylvain> Events : 167456

sylvain> Layout : left-symmetric
sylvain> Chunk Size : 512K

sylvain> Device Role : Active device 2
sylvain> Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing)
sylvain> /dev/sde1:
sylvain> Magic : a92b4efc
sylvain> Version : 1.2
sylvain> Feature Map : 0x9
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e
sylvain> Name : le-bohec:2  (local to host le-bohec)
sylvain> Creation Time : Tue Apr  9 17:56:19 2013
sylvain> Raid Level : raid5
sylvain> Raid Devices : 4

sylvain> Avail Dev Size : 4294963199 (2048.00 GiB 2199.02 GB)
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB)
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB)
sylvain> Data Offset : 2048 sectors
sylvain> Super Offset : 8 sectors
sylvain> Unused Space : before=1960 sectors, after=2341445631 sectors
sylvain> State : clean
sylvain> Device UUID : 0ebce28d:1a792d55:76a86538:12cc94dd

sylvain> Internal Bitmap : 8 sectors from superblock
sylvain> Update Time : Sat Dec 27 22:08:34 2014
sylvain> Bad Block Log : 512 entries available at offset 72 sectors - bad blocks present.
sylvain> Checksum : 3801cfa - correct
sylvain> Events : 167456

sylvain> Layout : left-symmetric
sylvain> Chunk Size : 512K

sylvain> Device Role : spare
sylvain> Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing)


sylvain> Now, i have installed ddrescue on the system, and i go to my
sylvain> friends to add the 2TB disk in the tower, and launch the
sylvain> command.




sylvain> Many thank's for your help and your patience.
sylvain> Best Regards
sylvain> Sylvain Depuille

sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "sylvain depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
sylvain> Envoyé: Lundi 29 Décembre 2014 21:36:23 
sylvain> Objet: Re: Re : Re: Big trouble during reassemble a Raid5 


sylvain> Hi john, thanks for your answer! I have change a 1TB disk to 
sylvain> growing the raid with 3TB disk. if i can re-insert the old 
sylvain> 1TB disk in place of 3TB disk, only some log and history are 
sylvain> corrupted. i think that is the best way to relaunch the raid 
sylvain> without data loss. But i dont known how change the timestamp 
sylvain> of the one raid disk. Have you a magic command to change a 
sylvain> timestamp of a raid partition, and how known the timestamp of 
sylvain> the other disk of the raid? After' raid relaunch, i can 
sylvain> change the burn disk by a 3TB new one. To do the ddrescue, i 
sylvain> have a 2TB disk spare! Its not the same geometry, is it 
sylvain> possible? thanks in advance for your help 

sylvain> Sylvain, 

sylvain> Always glad to help here. I'm going to try and understand what you 
sylvain> wrote and do my best to reply. 

sylvain> Is the 1Tb disk the bad disk? And if you re-insert it and re-start 
sylvain> the RAID5 array, you only have some minor lost files? If so, I would 
sylvain> probably just copy all the data off the RAID5 onto the single 3Tb disk 
sylvain> as a quick and dirty backup, then I'd use 'dd_rescue' to copy the bad 
sylvain> 1Tb disk onto the new 2Tb disk. 

sylvain> All you would have to do is make a partition on the 2tb disk which is 
sylvain> the same size (or a little bigger) than the partition on the 1tb disk, 
sylvain> then copy the partition over like this: 

sylvain> ddrescue /dev/sd[BAD DISK LETTER HERE]1 /dev/sd[2TB disk letter]1 \ 
sylvain> /tmp/rescue.log 

sylvain> So say the bad disk is sdc, and the good 2tb is sdf, you would do: 

sylvain> ddrescue /dev/sdc1 /dev/sdf1 /tmp/rescue.log 

sylvain> and let it go. Then you would assemble the array using the NEW 2tb 
sylvain> disk. Ideally you would remove the bad 1tb disk from the system when 
sylvain> trying to do this. 

sylvain> But you really do need send us the output of the following commands: 

sylvain> cat /proc/mdstat 
sylvain> cat /proc/partitions 
sylvain> mdadm --detail /dev/md# 

sylvain> do the above for the RADI5 array. 

sylvain> mdadm --examine /dev/sd#1 

sylvain> for each disk in the RAID5 array. 

sylvain> And we can give you better advice. 

sylvain> Good luck! 


sylvain> ---------------------------------- Sylvain Depuille 
sylvain> sylvain.depuille@laposte.net ----- Mail d'origine ----- De: 
sylvain> John Stoffel <john@stoffel.org> À: sylvain depuille 
sylvain> <sylvain.depuille@laposte.net> Cc: linux-raid@vger.kernel.org 
sylvain> Envoyé: Mon, 29 Dec 2014 19:32:04 +0100 (CET) Objet: Re: Big 
sylvain> trouble during reassemble a Raid5 

sylvain> Sylvain, I would recommend that you buy a replacement disk 
sylvain> for the one throwing errors and then run dd_rescue to copy as 
sylvain> much data from the dying disk to the replacement. Then, and 
sylvain> only then, do you try to reassemble the array with the 
sylvain> --force option. That disk is dying, and dying quickly. Can 
sylvain> you also post the output of mdadm -E /dev/sd[bcde]1 for each 
sylvain> disk, even the dying one, so we can look at the counts and 
sylvain> give you some more advice. Also, the output of the mdadm 
sylvain> --assemble --force /dev/md2 /dev/sd[bcde]1 would also be 
sylvain> good. The more info the better. Good luck! John 

sylvain> i'm sorry to ask this questions but the raid 5 with 4 disk is 
sylvain> in big trouble during re-assemble. 2 disks are out of order. 
sylvain> I have change a disk of the raid 5 (sde) to growing the raid. 
sylvain> But a second disk (sdc) have too many bad sector during the 
sylvain> re-assemble, and shutdown the re-assemble. "mdadm --assemble 
sylvain> --force /dev/md2 /dev/sd[bcde]1" I have try to correct bad 
sylvain> sectors with badblocks, but it's finished by no more spare 
sylvain> sectors and the disk still have some bad sector. badblocks -b 
sylvain> 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000 1140169336 
sylvain> 1140169400 1140169401 1140169402 1140169403 1140169404 
sylvain> 1140169405 1140169406 1140169407 1140169416 1140169417 
sylvain> 1140169418 1140169419 1140169420 1140169421 1140169422 
sylvain> 1140169423 

sylvain> For information the mdadm examine return : cat mdadm-exam.txt 
sylvain> /dev/sdb: MBR Magic : aa55 Partition[0] : 1953523120 sectors 
sylvain> at 2048 (type fd) /dev/sdc: MBR Magic : aa55 Partition[0] : 
sylvain> 1953520002 sectors at 63 (type fd) /dev/sdd: MBR Magic : aa55 
sylvain> Partition[0] : 1953520002 sectors at 63 (type fd) /dev/sde: 
sylvain> MBR Magic : aa55 Partition[0] : 4294965247 sectors at 2048 
sylvain> (type fd) I have 2 way to solve the issue. The first, is to 
sylvain> have special command to pass bad sector during re-assemble as 
sylvain> "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" The second 
sylvain> is change the disk sde with the old good one, but some datas 
sylvain> have been changed on the raid since i have remove it. But 
sylvain> these datas are not important. It's only logs and history 
sylvain> activity. What can i do to recover a maximum datas without 
sylvain> too many risk? Thank's in advance Best Regards 
sylvain> ---------------------------------- Sylvain Depuille (in 
sylvain> trouble) sylvain.depuille@laposte.net -- To unsubscribe from 
sylvain> this list: send the line "unsubscribe linux-raid" in the body 
sylvain> of a message to majordomo@vger.kernel.org More majordomo info 
sylvain> at http://vger.kernel.org/majordomo-info.html 

--
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: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello,

It's my fault too, because the mailing list doesn't accept HTML format.
And my provider doesn't offer plain text option with the mobile site :-(

Back to the issue :

I have stop the raid.
mdadm --stop /dev/md2
mdadm: stopped /dev/md2

cat /proc/mdstat
Personalities : [raid1] [raid6] [raid5] [raid4] 
unused devices: <none>

And when i try to re-assemble, the command is aborting :
mdadm --assemble --force /dev/md2 /dev/sd[bcde]1
mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array.


The command wan't re-acquire the sdc1 device :
cat /proc/mdstat
Personalities : [raid1] [raid6] [raid5] [raid4] 
md2 : inactive sde1[4](S) sdd1[2](S) sdb1[5](S)
      4101001112 blocks super 1.2
       
unused devices: <none>


Many thank's for your help.
Best Regards
Sylvain Depuille

----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org> 
Envoyé: Mercredi 31 Décembre 2014 15:14:44 
Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


sylvain> I have remove the 1TB burny disk, and replace it by the 2TB 
sylvain> ddrescued of the burny disk. 

Great, 

sylvain> But i can't re-assemble the raid. 
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 
sylvain> mdadm: /dev/sdb1 is busy - skipping 
sylvain> mdadm: /dev/sdd1 is busy - skipping 
sylvain> mdadm: /dev/sde1 is busy - skipping 
sylvain> mdadm: Merging with already-assembled /dev/md/2 
sylvain> mdadm: /dev/md/2 assembled from 2 drives and 1 spare - not enough to start the array. 

I think you first need to stop the array, to make sure all the devices 
aren't in use. Have you looked through the archives of this list for 
previous examples? 

So you should be able to do: 

> mdadm --stop md2 
> mdadm --assemble --force /dev/md2 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1 

and send the output. You should also be replying to the mailing list 
as well, which I just noticed you haven't. My fault too! 


sylvain> cat /proc/mdstat : 
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> md2 : inactive sdd1[2](S) sde1[4](S) sdb1[5](S) 
sylvain> 4101001112 blocks super 1.2 

sylvain> unused devices: <none> 

sylvain> The result of command mdadm -E /dev/sd[bcde]1 > mdadm-E-new.log is 
sylvain> /dev/sdb1: 
sylvain> Magic : a92b4efc 
sylvain> Version : 1.2 
sylvain> Feature Map : 0x1 
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e 
sylvain> Name : le-bohec:2 (local to host le-bohec) 
sylvain> Creation Time : Tue Apr 9 17:56:19 2013 
sylvain> Raid Level : raid5 
sylvain> Raid Devices : 4 

sylvain> Avail Dev Size : 1953521072 (931.51 GiB 1000.20 GB) 
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB) 
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB) 
sylvain> Data Offset : 2048 sectors 
sylvain> Super Offset : 8 sectors 
sylvain> Unused Space : before=1960 sectors, after=3504 sectors 
sylvain> State : clean 
sylvain> Device UUID : 8506e09c:b87a44ed:7b4ee314:777ce89c 

sylvain> Internal Bitmap : 8 sectors from superblock 
sylvain> Update Time : Sat Dec 27 22:08:34 2014 
sylvain> Bad Block Log : 512 entries available at offset 72 sectors 
sylvain> Checksum : bad52d25 - correct 
sylvain> Events : 167456 

sylvain> Layout : left-symmetric 
sylvain> Chunk Size : 512K 

sylvain> Device Role : Active device 0 
sylvain> Array State : AAA. ('A' == active, '.' == missing, 'R' == replacing) 
sylvain> /dev/sdc1: 
sylvain> Magic : a92b4efc 
sylvain> Version : 1.2 
sylvain> Feature Map : 0x1 
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e 
sylvain> Name : le-bohec:2 (local to host le-bohec) 
sylvain> Creation Time : Tue Apr 9 17:56:19 2013 
sylvain> Raid Level : raid5 
sylvain> Raid Devices : 4 

sylvain> Avail Dev Size : 1953517954 (931.51 GiB 1000.20 GB) 
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB) 
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB) 
sylvain> Data Offset : 2048 sectors 
sylvain> Super Offset : 8 sectors 
sylvain> Unused Space : before=1968 sectors, after=386 sectors 
sylvain> State : clean 
sylvain> Device UUID : 44002aad:d3e17729:a93854eb:4139972e 

sylvain> Internal Bitmap : 8 sectors from superblock 
sylvain> Update Time : Sat Dec 27 22:08:22 2014 
sylvain> Checksum : 6f69285d - correct 
sylvain> Events : 167431 

sylvain> Layout : left-symmetric 
sylvain> Chunk Size : 512K 

sylvain> Device Role : Active device 1 
sylvain> Array State : AAAA ('A' == active, '.' == missing, 'R' == replacing) 
sylvain> /dev/sdd1: 
sylvain> Magic : a92b4efc 
sylvain> Version : 1.2 
sylvain> Feature Map : 0x1 
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e 
sylvain> Name : le-bohec:2 (local to host le-bohec) 
sylvain> Creation Time : Tue Apr 9 17:56:19 2013 
sylvain> Raid Level : raid5 
sylvain> Raid Devices : 4 

sylvain> Avail Dev Size : 1953517954 (931.51 GiB 1000.20 GB) 
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB) 
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB) 
sylvain> Data Offset : 2048 sectors 
sylvain> Super Offset : 8 sectors 
sylvain> Unused Space : before=1968 sectors, after=1953507504 sectors 
sylvain> State : clean 
sylvain> Device UUID : 44002aad:d3e17729:a93854eb:4139972e 

sylvain> Internal Bitmap : 8 sectors from superblock 
sylvain> Update Time : Sat Dec 27 22:08:22 2014 
sylvain> Checksum : 6f692876 - correct 
sylvain> Events : 167456 

sylvain> Layout : left-symmetric 
sylvain> Chunk Size : 512K 

sylvain> Device Role : Active device 1 
sylvain> Array State : AAAA ('A' == active, '.' == missing, 'R' == replacing) 
sylvain> /dev/sde1: 
sylvain> Magic : a92b4efc 
sylvain> Version : 1.2 
sylvain> Feature Map : 0x9 
sylvain> Array UUID : 2a1440cd:762a90fb:e3bd2f4d:617acb0e 
sylvain> Name : le-bohec:2 (local to host le-bohec) 
sylvain> Creation Time : Tue Apr 9 17:56:19 2013 
sylvain> Raid Level : raid5 
sylvain> Raid Devices : 4 

sylvain> Avail Dev Size : 4294963199 (2048.00 GiB 2199.02 GB) 
sylvain> Array Size : 2930276352 (2794.53 GiB 3000.60 GB) 
sylvain> Used Dev Size : 1953517568 (931.51 GiB 1000.20 GB) 
sylvain> Data Offset : 2048 sectors 
sylvain> Super Offset : 8 sectors 
sylvain> Unused Space : before=1960 sectors, after=2341445631 sectors 
sylvain> State : clean 
sylvain> Device UUID : 0ebce28d:1a792d55:76a86538:12cc94dd 

sylvain> Internal Bitmap : 8 sectors from superblock 
sylvain> Update Time : Sat Dec 27 22:08:34 2014 
sylvain> Bad Block Log : 512 entries available at offset 72 sectors - bad blocks present. 
sylvain> Checksum : 3801cfa - correct 
sylvain> Events : 167456 

sylvain> Layout : left-symmetric 
sylvain> Chunk Size : 512K 

sylvain> Device Role : spare 
sylvain> Array State : A.A. ('A' == active, '.' == missing, 'R' == replacing) 


sylvain> Could you help me to re-assemble safely the raid array? 

sylvain> Thank's in advance 
sylvain> Best Regards 
sylvain> Sylvain Depuille (always in trouble). 

sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "Sylvain Depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org> 
sylvain> Envoyé: Mardi 30 Décembre 2014 22:07:04 
sylvain> Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


Sylvain> I'm front of the pc! The Pass1 of The ddrescue is Not 
Sylvain> finished! Sorry for The false News! 

sylvain> No problem. Let it finish before you make any other attempts to 
sylvain> re-assemble the array. 

Sylvain> Of the 5 pass tale the same Time, the command Will finish in 20 or 24 jours. 

Sylvain> Thank's in advance 
Sylvain> Best Regards 


Sylvain> Envoyé depuis un telephone portable 

>>> Le 29 déc. 2014 à 21:36, John Stoffel <john@stoffel.org> a écrit : 
>>> 
>>> 
sylvain> Hi john, thanks for your answer! I have change a 1TB disk to 
sylvain> growing the raid with 3TB disk. if i can re-insert the old 
sylvain> 1TB disk in place of 3TB disk, only some log and history are 
sylvain> corrupted. i think that is the best way to relaunch the raid 
sylvain> without data loss. But i dont known how change the timestamp 
sylvain> of the one raid disk. Have you a magic command to change a 
sylvain> timestamp of a raid partition, and how known the timestamp of 
sylvain> the other disk of the raid? After' raid relaunch, i can 
sylvain> change the burn disk by a 3TB new one. To do the ddrescue, i 
sylvain> have a 2TB disk spare! Its not the same geometry, is it 
sylvain> possible? thanks in advance for your help 
>>> 
>>> Sylvain, 
>>> 
>>> Always glad to help here. I'm going to try and understand what you 
>>> wrote and do my best to reply. 
>>> 
>>> Is the 1Tb disk the bad disk? And if you re-insert it and re-start 
>>> the RAID5 array, you only have some minor lost files? If so, I would 
>>> probably just copy all the data off the RAID5 onto the single 3Tb disk 
>>> as a quick and dirty backup, then I'd use 'dd_rescue' to copy the bad 
>>> 1Tb disk onto the new 2Tb disk. 
>>> 
>>> All you would have to do is make a partition on the 2tb disk which is 
>>> the same size (or a little bigger) than the partition on the 1tb disk, 
>>> then copy the partition over like this: 
>>> 
>>> ddrescue /dev/sd[BAD DISK LETTER HERE]1 /dev/sd[2TB disk letter]1 \ 
>>> /tmp/rescue.log 
>>> 
>>> So say the bad disk is sdc, and the good 2tb is sdf, you would do: 
>>> 
>>> ddrescue /dev/sdc1 /dev/sdf1 /tmp/rescue.log 
>>> 
>>> and let it go. Then you would assemble the array using the NEW 2tb 
>>> disk. Ideally you would remove the bad 1tb disk from the system when 
>>> trying to do this. 
>>> 
>>> But you really do need send us the output of the following commands: 
>>> 
>>> cat /proc/mdstat 
>>> cat /proc/partitions 
>>> mdadm --detail /dev/md# 
>>> 
>>> do the above for the RADI5 array. 
>>> 
>>> mdadm --examine /dev/sd#1 
>>> 
>>> for each disk in the RAID5 array. 
>>> 
>>> And we can give you better advice. 
>>> 
>>> Good luck! 
>>> 
>>> 
sylvain> ---------------------------------- Sylvain Depuille 
sylvain> sylvain.depuille@laposte.net ----- Mail d'origine ----- De: 
sylvain> John Stoffel <john@stoffel.org> À: sylvain depuille 
sylvain> <sylvain.depuille@laposte.net> Cc: linux-raid@vger.kernel.org 
sylvain> Envoyé: Mon, 29 Dec 2014 19:32:04 +0100 (CET) Objet: Re: Big 
sylvain> trouble during reassemble a Raid5 
>>> 
sylvain> Sylvain, I would recommend that you buy a replacement disk 
sylvain> for the one throwing errors and then run dd_rescue to copy as 
sylvain> much data from the dying disk to the replacement. Then, and 
sylvain> only then, do you try to reassemble the array with the 
sylvain> --force option. That disk is dying, and dying quickly. Can 
sylvain> you also post the output of mdadm -E /dev/sd[bcde]1 for each 
sylvain> disk, even the dying one, so we can look at the counts and 
sylvain> give you some more advice. Also, the output of the mdadm 
sylvain> --assemble --force /dev/md2 /dev/sd[bcde]1 would also be 
sylvain> good. The more info the better. Good luck! John 
>>> 
sylvain> i'm sorry to ask this questions but the raid 5 with 4 disk is 
sylvain> in big trouble during re-assemble. 2 disks are out of order. 
sylvain> I have change a disk of the raid 5 (sde) to growing the raid. 
sylvain> But a second disk (sdc) have too many bad sector during the 
sylvain> re-assemble, and shutdown the re-assemble. "mdadm --assemble 
sylvain> --force /dev/md2 /dev/sd[bcde]1" I have try to correct bad 
sylvain> sectors with badblocks, but it's finished by no more spare 
sylvain> sectors and the disk still have some bad sector. badblocks -b 
sylvain> 512 -o badblocks-sdc.txt -v -n /dev/sdc 1140170000 1140169336 
sylvain> 1140169400 1140169401 1140169402 1140169403 1140169404 
sylvain> 1140169405 1140169406 1140169407 1140169416 1140169417 
sylvain> 1140169418 1140169419 1140169420 1140169421 1140169422 
sylvain> 1140169423 
>>> 
sylvain> For information the mdadm examine return : cat mdadm-exam.txt 
sylvain> /dev/sdb: MBR Magic : aa55 Partition[0] : 1953523120 sectors 
sylvain> at 2048 (type fd) /dev/sdc: MBR Magic : aa55 Partition[0] : 
sylvain> 1953520002 sectors at 63 (type fd) /dev/sdd: MBR Magic : aa55 
sylvain> Partition[0] : 1953520002 sectors at 63 (type fd) /dev/sde: 
sylvain> MBR Magic : aa55 Partition[0] : 4294965247 sectors at 2048 
sylvain> (type fd) I have 2 way to solve the issue. The first, is to 
sylvain> have special command to pass bad sector during re-assemble as 
sylvain> "mdadm --assemble --force /dev/md2 /dev/sd[bcde]1" The second 
sylvain> is change the disk sde with the old good one, but some datas 
sylvain> have been changed on the raid since i have remove it. But 
sylvain> these datas are not important. It's only logs and history 
sylvain> activity. What can i do to recover a maximum datas without 
sylvain> too many risk? Thank's in advance Best Regards 
sylvain> ---------------------------------- Sylvain Depuille (in 
sylvain> trouble) sylvain.depuille@laposte.net -- To unsubscribe from 
sylvain> this list: send the line "unsubscribe linux-raid" in the body 
sylvain> of a message to majordomo@vger.kernel.org More majordomo info 
sylvain> at http://vger.kernel.org/majordomo-info.html 
>>> 

--
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: Re: Re : Re: Big trouble during reassemble a Raid5

[John Stoffel]

sylvain> It's my fault too, because the mailing list doesn't accept
sylvain> HTML format.  And my provider doesn't offer plain text option
sylvain> with the mobile site :-(

It's a total pain at times.  I'm still reading email with plain ascii
most of the time, I don't care for HTML email generally since it just
gets in the way.  

sylvain> Back to the issue :

sylvain> I have stop the raid.
sylvain> mdadm --stop /dev/md2
sylvain> mdadm: stopped /dev/md2

sylvain> cat /proc/mdstat
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> unused devices: <none>

sylvain> And when i try to re-assemble, the command is aborting :
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1
sylvain> mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array.

So if I look at the event counts, there's a mismatch between
/dev/sd[bde]1 and /dev/sdc1.  So let's try to assemble without /dev/sdc1:

  mdadm -vv --assemble --force /dev/md2 /dev/sdb1 missing /dev/sdd1 /dev/sde1

and see what happens then.  With three out of the four devices, the
RAID5 array should start up, and then you can add in /dev/sdc1 to
rebuild the parity.

John

Re: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hi John,

I do the summary of disk and states :
/dev/sdb : 1TB : OK
/dev/sdc : 1TB : ??
/dev/sdd : 2TB : ddrescued
/dev/sde : 3TB : Not synchronized

Something is wrong in this list.
It seems that i have remove a wrong disk.

I should investigate with my friends tomorrow evening for us, afternoon for you.
I will on the road tomorrow.
And it is may be free for you?

I wish you a good eve for this evening, and see you next year ;-).

Best Regards
Sylvain Depuille
 
----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "Sylvain Depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org> 
Envoyé: Mercredi 31 Décembre 2014 18:27:03 
Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


Sylvain> But the missing disk is the rescued disk! 

I understand that, but we're looking to get the RAID5 up and working 
again, but the event count looks wrong on that disk, so it shouldn't 
be used to re-assemble the array. 

There is going to be some data loss here, but let's minimize it and 
get the array going again. 

Sylvain> And the 3TB should be rebuild, because i have change It just 
Sylvain> before the issue :-(sorry iPhone in French) 

I think it's ok, since the other three disks have matching event 
counts. It's the /dev/sdc1 that's behind the rest, so it needs to be 
excluded at first. 

Try it out and let me know. 


Sylvain> I haven't mailings list in copy because html :-( 
Sylvain> If necessary i Will revend the mail from House! 

Sylvain> Best Regards 
Sylvain> Sylvain Depuillr 

Sylvain> Envoyé depuis un telephone portable 

>> Le 31 déc. 2014 à 17:08, John Stoffel <john@stoffel.org> a écrit : 
>> 
>> 
sylvain> It's my fault too, because the mailing list doesn't accept 
sylvain> HTML format. And my provider doesn't offer plain text option 
sylvain> with the mobile site :-( 
>> 
>> It's a total pain at times. I'm still reading email with plain ascii 
>> most of the time, I don't care for HTML email generally since it just 
>> gets in the way. 
>> 
sylvain> Back to the issue : 
>> 
sylvain> I have stop the raid. 
sylvain> mdadm --stop /dev/md2 
sylvain> mdadm: stopped /dev/md2 
>> 
sylvain> cat /proc/mdstat 
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> unused devices: <none> 
>> 
sylvain> And when i try to re-assemble, the command is aborting : 
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 
sylvain> mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array. 
>> 
>> So if I look at the event counts, there's a mismatch between 
>> /dev/sd[bde]1 and /dev/sdc1. So let's try to assemble without /dev/sdc1: 
>> 
>> mdadm -vv --assemble --force /dev/md2 /dev/sdb1 missing /dev/sdd1 /dev/sde1 
>> 
>> and see what happens then. With three out of the four devices, the 
>> RAID5 array should start up, and then you can add in /dev/sdc1 to 
>> rebuild the parity. 
>> 
>> John 

--
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: Re: Re : Re: Big trouble during reassemble a Raid5

[John Stoffel]

>>>>> "sylvain" == sylvain depuille <sylvain.depuille@laposte.net> writes:

sylvain> Hi John,
sylvain> I do the summary of disk and states :
sylvain> /dev/sdb : 1TB : OK
sylvain> /dev/sdc : 1TB : ??
sylvain> /dev/sdd : 2TB : ddrescued
sylvain> /dev/sde : 3TB : Not synchronized

sylvain> Something is wrong in this list.  It seems that i have remove
sylvain> a wrong disk.

It might be that is the problem.  One thing to do is to run 'smartctl'
on all the disks and get the UUIDs/serial numbers so you can confirm
which disk is which.  

The 'lsscsi' is also useful to help make sure you know which drives
are mapped to which drive letter, since they can move around.  But if
you do get a chance, try what I suggested, which is to assemble with
just three of the drives.  Maybe don't even hook up /dev/sdc1 at all
to the system.  

sylvain> I should investigate with my friends tomorrow evening for us,
sylvain> afternoon for you.  I will on the road tomorrow.  And it is
sylvain> may be free for you?

I won't be around tomorrow afternoon, January 1st, I've got a birthday
party to attend.  I'll try to be online at some point, but it's not
likely.  

sylvain> I wish you a good eve for this evening, and see you next year ;-).

Happy New Year to you and your friends and family as well!  Enjoy!  I
will talk with you in the new year for sure.  

 
sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "Sylvain Depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org> 
sylvain> Envoyé: Mercredi 31 Décembre 2014 18:27:03 
sylvain> Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


Sylvain> But the missing disk is the rescued disk! 

sylvain> I understand that, but we're looking to get the RAID5 up and working 
sylvain> again, but the event count looks wrong on that disk, so it shouldn't 
sylvain> be used to re-assemble the array. 

sylvain> There is going to be some data loss here, but let's minimize it and 
sylvain> get the array going again. 

Sylvain> And the 3TB should be rebuild, because i have change It just 
Sylvain> before the issue :-(sorry iPhone in French) 

sylvain> I think it's ok, since the other three disks have matching event 
sylvain> counts. It's the /dev/sdc1 that's behind the rest, so it needs to be 
sylvain> excluded at first. 

sylvain> Try it out and let me know. 


Sylvain> I haven't mailings list in copy because html :-( 
Sylvain> If necessary i Will revend the mail from House! 

Sylvain> Best Regards 
Sylvain> Sylvain Depuillr 

Sylvain> Envoyé depuis un telephone portable 

>>> Le 31 déc. 2014 à 17:08, John Stoffel <john@stoffel.org> a écrit : 
>>> 
>>> 
sylvain> It's my fault too, because the mailing list doesn't accept 
sylvain> HTML format. And my provider doesn't offer plain text option 
sylvain> with the mobile site :-( 
>>> 
>>> It's a total pain at times. I'm still reading email with plain ascii 
>>> most of the time, I don't care for HTML email generally since it just 
>>> gets in the way. 
>>> 
sylvain> Back to the issue : 
>>> 
sylvain> I have stop the raid. 
sylvain> mdadm --stop /dev/md2 
sylvain> mdadm: stopped /dev/md2 
>>> 
sylvain> cat /proc/mdstat 
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> unused devices: <none> 
>>> 
sylvain> And when i try to re-assemble, the command is aborting : 
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 
sylvain> mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array. 
>>> 
>>> So if I look at the event counts, there's a mismatch between 
>>> /dev/sd[bde]1 and /dev/sdc1. So let's try to assemble without /dev/sdc1: 
>>> 
>>> mdadm -vv --assemble --force /dev/md2 /dev/sdb1 missing /dev/sdd1 /dev/sde1 
>>> 
>>> and see what happens then. With three out of the four devices, the 
>>> RAID5 array should start up, and then you can add in /dev/sdc1 to 
>>> rebuild the parity. 
>>> 
>>> John 
--
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: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello,
Happy New Year and received all my best wishes for 2015.
i'm back to home.
Now, the failed raid is at 700km (~400 miles) :-( I should work only by phone and ssh to repair the Raid.
We have changed the right disk and the raid is rebuilding. I wille write you when it's finished.
Many many thank's for your help and your patience.
I have a question to migrate Raid 5 to Raid 6 :
This site http://www.ewams.net/?date=2013/05/02&view=Converting_RAID5_to_RAID6_in_mdadm describe the right cookbook?
I don't want re-do this bad Story.

Thank's in advance.
Best Regards.
----------------------------------
Sylvain Depuille (Happy at this time).
sylvain.depuille@laposte.net

----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
Envoyé: Mercredi 31 Décembre 2014 20:11:32 
Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 

>>>>> "sylvain" == sylvain depuille <sylvain.depuille@laposte.net> writes: 

sylvain> Hi John, 
sylvain> I do the summary of disk and states : 
sylvain> /dev/sdb : 1TB : OK 
sylvain> /dev/sdc : 1TB : ?? 
sylvain> /dev/sdd : 2TB : ddrescued 
sylvain> /dev/sde : 3TB : Not synchronized 

sylvain> Something is wrong in this list. It seems that i have remove 
sylvain> a wrong disk. 

It might be that is the problem. One thing to do is to run 'smartctl' 
on all the disks and get the UUIDs/serial numbers so you can confirm 
which disk is which. 

The 'lsscsi' is also useful to help make sure you know which drives 
are mapped to which drive letter, since they can move around. But if 
you do get a chance, try what I suggested, which is to assemble with 
just three of the drives. Maybe don't even hook up /dev/sdc1 at all 
to the system. 

sylvain> I should investigate with my friends tomorrow evening for us, 
sylvain> afternoon for you. I will on the road tomorrow. And it is 
sylvain> may be free for you? 

I won't be around tomorrow afternoon, January 1st, I've got a birthday 
party to attend. I'll try to be online at some point, but it's not 
likely. 

sylvain> I wish you a good eve for this evening, and see you next year ;-). 

Happy New Year to you and your friends and family as well! Enjoy! I 
will talk with you in the new year for sure. 


sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "Sylvain Depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org> 
sylvain> Envoyé: Mercredi 31 Décembre 2014 18:27:03 
sylvain> Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


Sylvain> But the missing disk is the rescued disk! 

sylvain> I understand that, but we're looking to get the RAID5 up and working 
sylvain> again, but the event count looks wrong on that disk, so it shouldn't 
sylvain> be used to re-assemble the array. 

sylvain> There is going to be some data loss here, but let's minimize it and 
sylvain> get the array going again. 

Sylvain> And the 3TB should be rebuild, because i have change It just 
Sylvain> before the issue :-(sorry iPhone in French) 

sylvain> I think it's ok, since the other three disks have matching event 
sylvain> counts. It's the /dev/sdc1 that's behind the rest, so it needs to be 
sylvain> excluded at first. 

sylvain> Try it out and let me know. 


Sylvain> I haven't mailings list in copy because html :-( 
Sylvain> If necessary i Will revend the mail from House! 

Sylvain> Best Regards 
Sylvain> Sylvain Depuillr 

Sylvain> Envoyé depuis un telephone portable 

>>> Le 31 déc. 2014 à 17:08, John Stoffel <john@stoffel.org> a écrit : 
>>> 
>>> 
sylvain> It's my fault too, because the mailing list doesn't accept 
sylvain> HTML format. And my provider doesn't offer plain text option 
sylvain> with the mobile site :-( 
>>> 
>>> It's a total pain at times. I'm still reading email with plain ascii 
>>> most of the time, I don't care for HTML email generally since it just 
>>> gets in the way. 
>>> 
sylvain> Back to the issue : 
>>> 
sylvain> I have stop the raid. 
sylvain> mdadm --stop /dev/md2 
sylvain> mdadm: stopped /dev/md2 
>>> 
sylvain> cat /proc/mdstat 
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> unused devices: <none> 
>>> 
sylvain> And when i try to re-assemble, the command is aborting : 
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 
sylvain> mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array. 
>>> 
>>> So if I look at the event counts, there's a mismatch between 
>>> /dev/sd[bde]1 and /dev/sdc1. So let's try to assemble without /dev/sdc1: 
>>> 
>>> mdadm -vv --assemble --force /dev/md2 /dev/sdb1 missing /dev/sdd1 /dev/sde1 
>>> 
>>> and see what happens then. With three out of the four devices, the 
>>> RAID5 array should start up, and then you can add in /dev/sdc1 to 
>>> rebuild the parity. 
>>> 
>>> John 

--
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: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello,

First good news of the year : the dirty Raid is online.

cat /proc/mdstat
Personalities : [raid1] [raid6] [raid5] [raid4] 
md2 : active raid5 sdc1[3] sdd1[2] sde1[4] sdb1[5]
      2930276352 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/4] [UUUU]
      bitmap: 0/8 pages [0KB], 65536KB chunk


Many thank's for your help.

Now, i have unmounted all filesystems and do e2fsck on all.

With ext4, it's running very quickly, and no errors found. Is it reliable?
But it's may be not your part? ;-)

Thank's in advance
Best Regards

Sylvain Depuille (Happy).

----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
Envoyé: Mercredi 31 Décembre 2014 20:11:32 
Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 

>>>>> "sylvain" == sylvain depuille <sylvain.depuille@laposte.net> writes: 

sylvain> Hi John, 
sylvain> I do the summary of disk and states : 
sylvain> /dev/sdb : 1TB : OK 
sylvain> /dev/sdc : 1TB : ?? 
sylvain> /dev/sdd : 2TB : ddrescued 
sylvain> /dev/sde : 3TB : Not synchronized 

sylvain> Something is wrong in this list. It seems that i have remove 
sylvain> a wrong disk. 

It might be that is the problem. One thing to do is to run 'smartctl' 
on all the disks and get the UUIDs/serial numbers so you can confirm 
which disk is which. 

The 'lsscsi' is also useful to help make sure you know which drives 
are mapped to which drive letter, since they can move around. But if 
you do get a chance, try what I suggested, which is to assemble with 
just three of the drives. Maybe don't even hook up /dev/sdc1 at all 
to the system. 

sylvain> I should investigate with my friends tomorrow evening for us, 
sylvain> afternoon for you. I will on the road tomorrow. And it is 
sylvain> may be free for you? 

I won't be around tomorrow afternoon, January 1st, I've got a birthday 
party to attend. I'll try to be online at some point, but it's not 
likely. 

sylvain> I wish you a good eve for this evening, and see you next year ;-). 

Happy New Year to you and your friends and family as well! Enjoy! I 
will talk with you in the new year for sure. 


sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "Sylvain Depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org> 
sylvain> Envoyé: Mercredi 31 Décembre 2014 18:27:03 
sylvain> Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


Sylvain> But the missing disk is the rescued disk! 

sylvain> I understand that, but we're looking to get the RAID5 up and working 
sylvain> again, but the event count looks wrong on that disk, so it shouldn't 
sylvain> be used to re-assemble the array. 

sylvain> There is going to be some data loss here, but let's minimize it and 
sylvain> get the array going again. 

Sylvain> And the 3TB should be rebuild, because i have change It just 
Sylvain> before the issue :-(sorry iPhone in French) 

sylvain> I think it's ok, since the other three disks have matching event 
sylvain> counts. It's the /dev/sdc1 that's behind the rest, so it needs to be 
sylvain> excluded at first. 

sylvain> Try it out and let me know. 


Sylvain> I haven't mailings list in copy because html :-( 
Sylvain> If necessary i Will revend the mail from House! 

Sylvain> Best Regards 
Sylvain> Sylvain Depuillr 

Sylvain> Envoyé depuis un telephone portable 

>>> Le 31 déc. 2014 à 17:08, John Stoffel <john@stoffel.org> a écrit : 
>>> 
>>> 
sylvain> It's my fault too, because the mailing list doesn't accept 
sylvain> HTML format. And my provider doesn't offer plain text option 
sylvain> with the mobile site :-( 
>>> 
>>> It's a total pain at times. I'm still reading email with plain ascii 
>>> most of the time, I don't care for HTML email generally since it just 
>>> gets in the way. 
>>> 
sylvain> Back to the issue : 
>>> 
sylvain> I have stop the raid. 
sylvain> mdadm --stop /dev/md2 
sylvain> mdadm: stopped /dev/md2 
>>> 
sylvain> cat /proc/mdstat 
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> unused devices: <none> 
>>> 
sylvain> And when i try to re-assemble, the command is aborting : 
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 
sylvain> mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array. 
>>> 
>>> So if I look at the event counts, there's a mismatch between 
>>> /dev/sd[bde]1 and /dev/sdc1. So let's try to assemble without /dev/sdc1: 
>>> 
>>> mdadm -vv --assemble --force /dev/md2 /dev/sdb1 missing /dev/sdd1 /dev/sde1 
>>> 
>>> and see what happens then. With three out of the four devices, the 
>>> RAID5 array should start up, and then you can add in /dev/sdc1 to 
>>> rebuild the parity. 
>>> 
>>> John 

--
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: Re: Re : Re: Big trouble during reassemble a Raid5

[John Stoffel]

sylvain> First good news of the year : the dirty Raid is online.

Awesome news!  A great start to the New Year.  And a happy new year to
you all.  

sylvain> cat /proc/mdstat
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> md2 : active raid5 sdc1[3] sdd1[2] sde1[4] sdb1[5]
sylvain> 2930276352 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/4] [UUUU]
sylvain> bitmap: 0/8 pages [0KB], 65536KB chunk


sylvain> Many thank's for your help.

sylvain> Now, i have unmounted all filesystems and do e2fsck on all.

sylvain> With ext4, it's running very quickly, and no errors found. Is it reliable?
sylvain> But it's may be not your part? ;-)

You can force ext4 to do a more thorough check if you like with:

   fsck.ext4 -f /dev/....

and it will force a check, even if everything looks to be ok.  

sylvain> Thank's in advance
sylvain> Best Regards

sylvain> Sylvain Depuille (Happy).

sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "sylvain depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
sylvain> Envoyé: Mercredi 31 Décembre 2014 20:11:32 
sylvain> Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 

>>>>> "sylvain" == sylvain depuille <sylvain.depuille@laposte.net> writes: 

sylvain> Hi John, 
sylvain> I do the summary of disk and states : 
sylvain> /dev/sdb : 1TB : OK 
sylvain> /dev/sdc : 1TB : ?? 
sylvain> /dev/sdd : 2TB : ddrescued 
sylvain> /dev/sde : 3TB : Not synchronized 

sylvain> Something is wrong in this list. It seems that i have remove 
sylvain> a wrong disk. 

sylvain> It might be that is the problem. One thing to do is to run 'smartctl' 
sylvain> on all the disks and get the UUIDs/serial numbers so you can confirm 
sylvain> which disk is which. 

sylvain> The 'lsscsi' is also useful to help make sure you know which drives 
sylvain> are mapped to which drive letter, since they can move around. But if 
sylvain> you do get a chance, try what I suggested, which is to assemble with 
sylvain> just three of the drives. Maybe don't even hook up /dev/sdc1 at all 
sylvain> to the system. 

sylvain> I should investigate with my friends tomorrow evening for us, 
sylvain> afternoon for you. I will on the road tomorrow. And it is 
sylvain> may be free for you? 

sylvain> I won't be around tomorrow afternoon, January 1st, I've got a birthday 
sylvain> party to attend. I'll try to be online at some point, but it's not 
sylvain> likely. 

sylvain> I wish you a good eve for this evening, and see you next year ;-). 

sylvain> Happy New Year to you and your friends and family as well! Enjoy! I 
sylvain> will talk with you in the new year for sure. 


sylvain> ----- Mail original ----- 
sylvain> De: "John Stoffel" <john@stoffel.org> 
sylvain> À: "Sylvain Depuille" <sylvain.depuille@laposte.net> 
sylvain> Cc: "John Stoffel" <john@stoffel.org> 
sylvain> Envoyé: Mercredi 31 Décembre 2014 18:27:03 
sylvain> Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


Sylvain> But the missing disk is the rescued disk! 

sylvain> I understand that, but we're looking to get the RAID5 up and working 
sylvain> again, but the event count looks wrong on that disk, so it shouldn't 
sylvain> be used to re-assemble the array. 

sylvain> There is going to be some data loss here, but let's minimize it and 
sylvain> get the array going again. 

Sylvain> And the 3TB should be rebuild, because i have change It just 
Sylvain> before the issue :-(sorry iPhone in French) 

sylvain> I think it's ok, since the other three disks have matching event 
sylvain> counts. It's the /dev/sdc1 that's behind the rest, so it needs to be 
sylvain> excluded at first. 

sylvain> Try it out and let me know. 


Sylvain> I haven't mailings list in copy because html :-( 
Sylvain> If necessary i Will revend the mail from House! 

Sylvain> Best Regards 
Sylvain> Sylvain Depuillr 

Sylvain> Envoyé depuis un telephone portable 

>>>> Le 31 déc. 2014 à 17:08, John Stoffel <john@stoffel.org> a écrit : 
>>>> 
>>>> 
sylvain> It's my fault too, because the mailing list doesn't accept 
sylvain> HTML format. And my provider doesn't offer plain text option 
sylvain> with the mobile site :-( 
>>>> 
>>>> It's a total pain at times. I'm still reading email with plain ascii 
>>>> most of the time, I don't care for HTML email generally since it just 
>>>> gets in the way. 
>>>> 
sylvain> Back to the issue : 
>>>> 
sylvain> I have stop the raid. 
sylvain> mdadm --stop /dev/md2 
sylvain> mdadm: stopped /dev/md2 
>>>> 
sylvain> cat /proc/mdstat 
sylvain> Personalities : [raid1] [raid6] [raid5] [raid4] 
sylvain> unused devices: <none> 
>>>> 
sylvain> And when i try to re-assemble, the command is aborting : 
sylvain> mdadm --assemble --force /dev/md2 /dev/sd[bcde]1 
sylvain> mdadm: /dev/md2 assembled from 2 drives and 1 spare - not enough to start the array. 
>>>> 
>>>> So if I look at the event counts, there's a mismatch between 
>>>> /dev/sd[bde]1 and /dev/sdc1. So let's try to assemble without /dev/sdc1: 
>>>> 
>>>> mdadm -vv --assemble --force /dev/md2 /dev/sdb1 missing /dev/sdd1 /dev/sde1 
>>>> 
>>>> and see what happens then. With three out of the four devices, the 
>>>> RAID5 array should start up, and then you can add in /dev/sdc1 to 
>>>> rebuild the parity. 
>>>> 
>>>> John 
--
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: Re: Re : Re: Big trouble during reassemble a Raid5

[John Stoffel]

sylvain> Happy New Year and received all my best wishes for 2015.  i'm
sylvain> back to home.

Me too, but heading to bed soon, been a busy day here.  

sylvain> Now, the failed raid is at 700km (~400 miles) :-( I should
sylvain> work only by phone and ssh to repair the Raid.

It sucks having to work at a distance like this.  

sylvain> We have changed the right disk and the raid is rebuilding. I
sylvain> wille write you when it's finished.  Many many thank's for
sylvain> your help and your patience.

My real suggestion to you is to now collect all the information you
can on the system, including drive serial numbers, and how they map to
each device and member of the RAID array.  Sometimes when a system
reboots, or you add new disks, things get re-arranged and you don't
notice at first.  And that leads you down the wrong path quite
quickly.

sylvain> I have a question to migrate Raid 5 to Raid 6 :

sylvain> This site
sylvain> http://www.ewams.net/?date=2013/05/02&view=Converting_RAID5_to_RAID6_in_mdadm
sylvain> describe the right cookbook?  I don't want re-do this bad
sylvain> Story.

It looks right to me.  But I would also suggest that you look through
the archives of this list, and make sure you have a newer version of
the mdadm tool before you try to do this.

I would also suggest that you add two new disks, one to be used as a
hot spare, and the other used to grow the array into a RAID6.  

Then go through and make sure that all your drive timeouts are set
properly on system boot, so that if a disk hangs when trying to
re-read a bad block, it won't cause it to be kicked out of the array
too quickly.  Again, check the recent archives (last three months of
2014) of the list for examples and discussions of this issue.

Good luck!
John

Re: Re : Re: Big trouble during reassemble a Raid5

[Can Jeuleers]

On 02/01/15 05:07, John Stoffel wrote:
> My real suggestion to you is to now collect all the information you
> can on the system, including drive serial numbers, and how they map to
> each device and member of the RAID array.  Sometimes when a system
> reboots, or you add new disks, things get re-arranged and you don't
> notice at first.  And that leads you down the wrong path quite
> quickly.

Sylvain, a nice script for documenting device serial numbers,
partitions, RAID memberships etc can be found here:

https://github.com/pturmel/lsdrv/raw/master/lsdrv

Written by Phil Turmel; a frequent contributor to this list.

Jan

Re: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello,

the raid is completly operationnal.

I take linux up to date with emerge Gentoo stable each day or each week.

I know the attribute pv_timeout on SAN with HP-UX, but not with Linux.
I haven't find the discussion over this point with Google.
Could you help me to find the thread on that?

PS: could i write what's your explain me on my wiki http://wiki.depuille.net/index.php/Accueil (it's a French Unix tricks/memorendum Wiki)?

Thank's in advance.
Best Regards
Sylvain Depuille

----- Mail original ----- 
De: "John Stoffel" <john@stoffel.org> 
À: "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: "John Stoffel" <john@stoffel.org>, linux-raid@vger.kernel.org 
Envoyé: Vendredi 2 Janvier 2015 05:07:32 
Objet: Re: Re: Re : Re: Big trouble during reassemble a Raid5 


sylvain> Happy New Year and received all my best wishes for 2015. i'm 
sylvain> back to home. 

Me too, but heading to bed soon, been a busy day here. 

sylvain> Now, the failed raid is at 700km (~400 miles) :-( I should 
sylvain> work only by phone and ssh to repair the Raid. 

It sucks having to work at a distance like this. 

sylvain> We have changed the right disk and the raid is rebuilding. I 
sylvain> wille write you when it's finished. Many many thank's for 
sylvain> your help and your patience. 

My real suggestion to you is to now collect all the information you 
can on the system, including drive serial numbers, and how they map to 
each device and member of the RAID array. Sometimes when a system 
reboots, or you add new disks, things get re-arranged and you don't 
notice at first. And that leads you down the wrong path quite 
quickly. 

sylvain> I have a question to migrate Raid 5 to Raid 6 : 

sylvain> This site 
sylvain> http://www.ewams.net/?date=2013/05/02&view=Converting_RAID5_to_RAID6_in_mdadm 
sylvain> describe the right cookbook? I don't want re-do this bad 
sylvain> Story. 

It looks right to me. But I would also suggest that you look through 
the archives of this list, and make sure you have a newer version of 
the mdadm tool before you try to do this. 

I would also suggest that you add two new disks, one to be used as a 
hot spare, and the other used to grow the array into a RAID6. 

Then go through and make sure that all your drive timeouts are set 
properly on system boot, so that if a disk hangs when trying to 
re-read a bad block, it won't cause it to be kicked out of the array 
too quickly. Again, check the recent archives (last three months of 
2014) of the list for examples and discussions of this issue. 

Good luck! 
John 


--
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: Re : Re: Big trouble during reassemble a Raid5

[sylvain.depuille]

Hello, 

thank's for the tricks.

I have modify the #! as #!/usr/bin/python2 because i have python 2.7 and 3.3 as default on my PCs.

I keep this script on my PCs, and the output ;-)

Thank's a lot.
Best Regards

----- Mail original ----- 
De: "Can Jeuleers" <can.jeuleers@gmail.com> 
À: "John Stoffel" <john@stoffel.org>, "sylvain depuille" <sylvain.depuille@laposte.net> 
Cc: linux-raid@vger.kernel.org 
Envoyé: Vendredi 2 Janvier 2015 08:59:39 
Objet: Re: Re : Re: Big trouble during reassemble a Raid5 

On 02/01/15 05:07, John Stoffel wrote: 
> My real suggestion to you is to now collect all the information you 
> can on the system, including drive serial numbers, and how they map to 
> each device and member of the RAID array. Sometimes when a system 
> reboots, or you add new disks, things get re-arranged and you don't 
> notice at first. And that leads you down the wrong path quite 
> quickly. 

Sylvain, a nice script for documenting device serial numbers, 
partitions, RAID memberships etc can be found here: 

https://github.com/pturmel/lsdrv/raw/master/lsdrv 

Written by Phil Turmel; a frequent contributor to this list. 

Jan 

--
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