why won't this RAID5 device start?

why won't this RAID5 device start?

[David T-G]

Hi, all --

I recently migrated our disk farm to a new box with a new OS build
(openSuSE from KNOPPIX).  Aside from the usual challenges of setting
up the world again, I have a 3-device RAID5 volume that won't start.
The other metadevice is fine, though; I think we can say that the md
system is running.  Soooooo ...  Where do I start?

  diskfarm:~ # cat /proc/mdstat 
  Personalities : [raid6] [raid5] [raid4] 
  md0 : active raid5 sdc1[3] sdd1[4] sdb1[0]
        11720265216 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/3] [U_UU]
        
  md127 : inactive sdl2[0](S) sdj2[3](S) sdf2[1](S)
        2196934199 blocks super 1.2
         
  unused devices: <none>

[No, I don't know why md127 was first a moment ago!]  I tried reassembling
the device, but mdadm doesn't like it.

  diskfarm:~ # mdadm --stop /dev/md127
  mdadm: stopped /dev/md127
  diskfarm:~ # mdadm --assemble --scan
  mdadm: /dev/md/750Graid5md assembled from 1 drive - not enough to start the array.
  mdadm: Found some drive for an array that is already active: /dev/md/0
  mdadm: giving up.
  mdadm: No arrays found in config file or automatically

But ... what's with just 1 drive?

  diskfarm:~ # for D in /dev/sd[fjl] ; do parted $D print ; done
  Model: ATA WDC WD7500BPKX-7 (scsi)
  Disk /dev/sdf: 750GB
  Sector size (logical/physical): 512B/4096B
  Partition Table: gpt
  Disk Flags: 

  Number  Start   End    Size   File system  Name              Flags
   2      1049kB  750GB  750GB  ntfs         Linux RAID        raid
   3      750GB   750GB  134MB  ext3         Linux filesystem

  Model: ATA WDC WD7500BPKX-7 (scsi)
  Disk /dev/sdj: 750GB
  Sector size (logical/physical): 512B/4096B
  Partition Table: gpt
  Disk Flags: 

  Number  Start   End    Size   File system  Name              Flags
   2      1049kB  750GB  750GB  ntfs         Linux RAID        raid
   3      750GB   750GB  134MB  xfs          Linux filesystem

  Model: ATA Hitachi HDE72101 (scsi)
  Disk /dev/sdl: 1000GB
  Sector size (logical/physical): 512B/512B
  Partition Table: msdos
  Disk Flags: 

  Number  Start   End     Size    Type     File system  Flags
   1      1049kB  4227MB  4226MB  primary  ntfs         diag, type=27
   2      4227MB  754GB   750GB   primary  reiserfs     raid, type=fd
   3      754GB   754GB   134MB   primary  reiserfs     type=83
   4      754GB   1000GB  246GB   primary  reiserfs     type=83

Slice 2 on each is the RAID partition, slice 3 on each is a little 
filesystem for bare-bones info, and slice 4 on sdl is a normal basic
filesystem for scratch content.

  diskfarm:~ # mdadm --examine /dev/sd[fjl]2 | egrep '/dev|Name|Role|State|Checksum|Events|UUID'
  /dev/sdf2:
       Array UUID : 88575f01:592167fd:bd9f9ba1:a61fafc4
             Name : diskfarm:750Graid5md  (local to host diskfarm)
            State : clean
      Device UUID : e916fc67:b8b7fc59:51440134:fa431d02
         Checksum : 43f9e7a4 - correct
           Events : 720
     Device Role : Active device 1
     Array State : AAA ('A' == active, '.' == missing, 'R' == replacing)
  /dev/sdj2:
       Array UUID : 88575f01:592167fd:bd9f9ba1:a61fafc4
             Name : diskfarm:750Graid5md  (local to host diskfarm)
            State : clean
      Device UUID : 0b847f84:83e80a3d:a0dc11e7:60bffc9f
         Checksum : 9522782b - correct
           Events : 177792
     Device Role : Active device 2
     Array State : A.A ('A' == active, '.' == missing, 'R' == replacing)
  /dev/sdl2:
       Array UUID : 88575f01:592167fd:bd9f9ba1:a61fafc4
             Name : diskfarm:750Graid5md  (local to host diskfarm)
            State : clean
      Device UUID : cc53440e:cb9180e4:be4c38d4:88a676eb
         Checksum : fef95256 - correct
           Events : 177794
     Device Role : Active device 0
     Array State : A.. ('A' == active, '.' == missing, 'R' == replacing)

Slice f2 looks great, but slices j2 & l2 seem to be missing -- even though
they are present.  Worse, the Events counter on sdf2 is frighteningly
small.  Where did it go?!?  So maybe I consider sdf2 failed and reassemble
from the other two [only] and then put f2 back in?

Definitely time to stop, take a deep breath, and ask for help :-)


TIA & HAND

:-D
-- 
David T-G
See http://justpickone.org/davidtg/email/
See http://justpickone.org/davidtg/tofu.txt

Re: why won't this RAID5 device start?

[Wols Lists]

On 28/03/21 03:14, David T-G wrote:
> Hi, all --
> 
> I recently migrated our disk farm to a new box with a new OS build
> (openSuSE from KNOPPIX).  Aside from the usual challenges of setting
> up the world again, I have a 3-device RAID5 volume that won't start.
> The other metadevice is fine, though; I think we can say that the md
> system is running.  Soooooo ...  Where do I start?

https://raid.wiki.kernel.org/index.php/Linux_Raid#When_Things_Go_Wrogn

Sorry I don't have time to look deeper right now. If someone else
doesn't chime in first, I'll take a look when I get back from work.

Cheers,
Wol

Re: why won't this RAID5 device start?

[Wols Lists]

On 28/03/21 03:14, David T-G wrote:
> Hi, all --
> 
> I recently migrated our disk farm to a new box with a new OS build
> (openSuSE from KNOPPIX).  Aside from the usual challenges of setting
> up the world again, I have a 3-device RAID5 volume that won't start.
> The other metadevice is fine, though; I think we can say that the md
> system is running.  Soooooo ...  Where do I start?
> 
>   diskfarm:~ # cat /proc/mdstat 
>   Personalities : [raid6] [raid5] [raid4] 
>   md0 : active raid5 sdc1[3] sdd1[4] sdb1[0]
>         11720265216 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/3] [U_UU]

This looks wrong - is this supposed to be a four-drive array? U_UU
implies a missing drive ... as does [4/3]
>         
>   md127 : inactive sdl2[0](S) sdj2[3](S) sdf2[1](S)
>         2196934199 blocks super 1.2
>          
>   unused devices: <none>
> 
> [No, I don't know why md127 was first a moment ago!]  I tried reassembling
> the device, but mdadm doesn't like it.
> 
>   diskfarm:~ # mdadm --stop /dev/md127
>   mdadm: stopped /dev/md127
>   diskfarm:~ # mdadm --assemble --scan
>   mdadm: /dev/md/750Graid5md assembled from 1 drive - not enough to start the array.
>   mdadm: Found some drive for an array that is already active: /dev/md/0
>   mdadm: giving up.
>   mdadm: No arrays found in config file or automatically
> 
> But ... what's with just 1 drive?
> 
>   diskfarm:~ # for D in /dev/sd[fjl] ; do parted $D print ; done
>   Model: ATA WDC WD7500BPKX-7 (scsi)
>   Disk /dev/sdf: 750GB
>   Sector size (logical/physical): 512B/4096B
>   Partition Table: gpt
>   Disk Flags: 
> 
>   Number  Start   End    Size   File system  Name              Flags
>    2      1049kB  750GB  750GB  ntfs         Linux RAID        raid
>    3      750GB   750GB  134MB  ext3         Linux filesystem
> 
>   Model: ATA WDC WD7500BPKX-7 (scsi)
>   Disk /dev/sdj: 750GB
>   Sector size (logical/physical): 512B/4096B
>   Partition Table: gpt
>   Disk Flags: 
> 
>   Number  Start   End    Size   File system  Name              Flags
>    2      1049kB  750GB  750GB  ntfs         Linux RAID        raid
>    3      750GB   750GB  134MB  xfs          Linux filesystem
> 
>   Model: ATA Hitachi HDE72101 (scsi)
>   Disk /dev/sdl: 1000GB
>   Sector size (logical/physical): 512B/512B
>   Partition Table: msdos
>   Disk Flags: 
> 
>   Number  Start   End     Size    Type     File system  Flags
>    1      1049kB  4227MB  4226MB  primary  ntfs         diag, type=27
>    2      4227MB  754GB   750GB   primary  reiserfs     raid, type=fd
>    3      754GB   754GB   134MB   primary  reiserfs     type=83
>    4      754GB   1000GB  246GB   primary  reiserfs     type=83
> 
> Slice 2 on each is the RAID partition, slice 3 on each is a little 
> filesystem for bare-bones info, and slice 4 on sdl is a normal basic
> filesystem for scratch content.
> 
>   diskfarm:~ # mdadm --examine /dev/sd[fjl]2 | egrep '/dev|Name|Role|State|Checksum|Events|UUID'
>   /dev/sdf2:
>        Array UUID : 88575f01:592167fd:bd9f9ba1:a61fafc4
>              Name : diskfarm:750Graid5md  (local to host diskfarm)
>             State : clean
>       Device UUID : e916fc67:b8b7fc59:51440134:fa431d02
>          Checksum : 43f9e7a4 - correct
>            Events : 720
>      Device Role : Active device 1
>      Array State : AAA ('A' == active, '.' == missing, 'R' == replacing)
>   /dev/sdj2:
>        Array UUID : 88575f01:592167fd:bd9f9ba1:a61fafc4
>              Name : diskfarm:750Graid5md  (local to host diskfarm)
>             State : clean
>       Device UUID : 0b847f84:83e80a3d:a0dc11e7:60bffc9f
>          Checksum : 9522782b - correct
>            Events : 177792
>      Device Role : Active device 2
>      Array State : A.A ('A' == active, '.' == missing, 'R' == replacing)
>   /dev/sdl2:
>        Array UUID : 88575f01:592167fd:bd9f9ba1:a61fafc4
>              Name : diskfarm:750Graid5md  (local to host diskfarm)
>             State : clean
>       Device UUID : cc53440e:cb9180e4:be4c38d4:88a676eb
>          Checksum : fef95256 - correct
>            Events : 177794
>      Device Role : Active device 0
>      Array State : A.. ('A' == active, '.' == missing, 'R' == replacing)
> 
> Slice f2 looks great, but slices j2 & l2 seem to be missing -- even though
> they are present.  Worse, the Events counter on sdf2 is frighteningly
> small.  Where did it go?!?  So maybe I consider sdf2 failed and reassemble
> from the other two [only] and then put f2 back in?

Yes I'm afraid so. I'd guess you've been running a failed raid-5 for
ages, and because something hiccuped when you shut it down, the two good
drives drifted apart, and now they won't start ...
> 
> Definitely time to stop, take a deep breath, and ask for help :-)
> 
Read up about overlays on the web site, use an overlay to force-assemble
the two good drives, run fsck etc to check everything's good (you might
lose a bit of data, hopefully very little), then if everything looks
okay do it for real. Ie force-assemble the two good drives, then re-add
the third. I'd just do a quick smartctl health check on all three drives
first, just to make sure nothing is obviously wrong with them - a
problem could kill your array completely! Then add the third drive back
in (whether you use --add or --re-add probably won't make much difference).

Oh - and fix whatever is wrong with md0, too, before that dies on you!

Cheers,
Wol

Re: why won't this RAID5 device start?

[David T-G]

Wol, et al --

...and then Wols Lists said...
% 
% On 28/03/21 03:14, David T-G wrote:
% > 
...
% > system is running.  Soooooo ...  Where do I start?
% > 
% >   diskfarm:~ # cat /proc/mdstat 
% >   Personalities : [raid6] [raid5] [raid4] 
% >   md0 : active raid5 sdc1[3] sdd1[4] sdb1[0]
% >         11720265216 blocks super 1.2 level 5, 512k chunk, algorithm 2 [4/3] [U_UU]
% 
% This looks wrong - is this supposed to be a four-drive array? U_UU
% implies a missing drive ... as does [4/3]

Yep :-(  I sent a separate note about that one.


% >         
% >   md127 : inactive sdl2[0](S) sdj2[3](S) sdf2[1](S)
% >         2196934199 blocks super 1.2
% >          
...
% >   diskfarm:~ # mdadm --examine /dev/sd[fjl]2 | egrep '/dev|Name|Role|State|Checksum|Events|UUID'
% >   /dev/sdf2:
...
% >            Events : 720
% >      Device Role : Active device 1
% >      Array State : AAA ('A' == active, '.' == missing, 'R' == replacing)
% >   /dev/sdj2:
...
% >            Events : 177792
% >      Device Role : Active device 2
% >      Array State : A.A ('A' == active, '.' == missing, 'R' == replacing)
% >   /dev/sdl2:
...
% >            Events : 177794
% >      Device Role : Active device 0
% >      Array State : A.. ('A' == active, '.' == missing, 'R' == replacing)
% > 
% > Slice f2 looks great, but slices j2 & l2 seem to be missing -- even though
% > they are present.  Worse, the Events counter on sdf2 is frighteningly
% > small.  Where did it go?!?  So maybe I consider sdf2 failed and reassemble
% > from the other two [only] and then put f2 back in?
% 
% Yes I'm afraid so. I'd guess you've been running a failed raid-5 for
% ages, and because something hiccuped when you shut it down, the two good
% drives drifted apart, and now they won't start ...

Arrrgh :-(  I was afraid of that ...


% > 
% > Definitely time to stop, take a deep breath, and ask for help :-)
% > 
% Read up about overlays on the web site, use an overlay to force-assemble

Ouch.  I tried that with a hiccup a few years ago and never really got it
down.  This is gonna take some reading again ...


% the two good drives, run fsck etc to check everything's good (you might
% lose a bit of data, hopefully very little), then if everything looks
% okay do it for real. Ie force-assemble the two good drives, then re-add

It looks like I want

  mdadm --assemble --force /dev/md127 /dev/sd[jl]2

for that based on the Go Wrong page and Go Wrogn collection.  Does that
sound right?


% the third. I'd just do a quick smartctl health check on all three drives
% first, just to make sure nothing is obviously wrong with them - a

That looks good, although perhaps quite simple:

  diskfarm:~ # for D in /dev/sd[fjl] ; do echo "### $D" ; smartctl --health $D ; done
  ### /dev/sdf
  smartctl 7.0 2019-05-21 r4917 [x86_64-linux-5.3.18-lp152.63-default] (SUSE RPM)
  Copyright (C) 2002-18, Bruce Allen, Christian Franke, www.smartmontools.org

  === START OF READ SMART DATA SECTION ===
  SMART overall-health self-assessment test result: PASSED

  ### /dev/sdj
  smartctl 7.0 2019-05-21 r4917 [x86_64-linux-5.3.18-lp152.63-default] (SUSE RPM)
  Copyright (C) 2002-18, Bruce Allen, Christian Franke, www.smartmontools.org

  === START OF READ SMART DATA SECTION ===
  SMART overall-health self-assessment test result: PASSED

  ### /dev/sdl
  smartctl 7.0 2019-05-21 r4917 [x86_64-linux-5.3.18-lp152.63-default] (SUSE RPM)
  Copyright (C) 2002-18, Bruce Allen, Christian Franke, www.smartmontools.org

  === START OF READ SMART DATA SECTION ===
  SMART overall-health self-assessment test result: PASSED

But --info is longer and still doesn't give THAT much info, so maybe this
is fine :-)


% problem could kill your array completely! Then add the third drive back
% in (whether you use --add or --re-add probably won't make much difference).

OK.


% 
% Oh - and fix whatever is wrong with md0, too, before that dies on you!

Yeah, about that ...  Haven't seen any replies there yet :-/


% 
% Cheers,
% Wol


Thanks again & HANN

:-D
-- 
David T-G
See http://justpickone.org/davidtg/email/
See http://justpickone.org/davidtg/tofu.txt