RAID6 simply does not start as /dev/md8

Hello everybody,
I've now been trying for over a week to get my RAID6 working. I have set up the array by partitioning all six disks (partition type:
Linux (83)) and executing "mdadm -C /dev/md8 --level=6 --raid-devices=6 /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1 /dev/sdf1".

When the array had finished syncing, I rebooted. /proc/mdstat contained

	md126 : inactive sde1[4](S) sdd1[3](S)
	      3907025072 blocks super 1.2

	md127 : inactive sdb1[1](S) sda1[0](S) sdc1[2](S) sdf1[5](S)
	      7814050144 blocks super 1.2

So I put the following lines into /etc/mdadm.conf (I had forgot this before):
"DEVICE /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1 /dev/sdf1 ARRAY /dev/md8 devices=/dev/sda1,/dev/sdb1,/dev/sdc1,/dev/sdd1,/dev/sde1,/dev/sdf1" So I rebooted again, now the array was started as follows:

	Personalities : [raid1] [raid6] [raid5] [raid4]
	md8 : inactive sda1[0](S)
	      1953512536 blocks super 1.2

I found out, that the assembly was aborted due to the error, that /dev/sda1 has no superblock. It seems, as if the kernel would start the array while booting the system, so mdadm is unable to assemble it afterwards. If I stop and re-assemble the array, it works fine.

Unfortunately, I have to keep "CONFIG_MD_AUTODETECT" enabled, because / and the whole system runs on a RAID1. But why does the kernel start the array, although the used partition type of the disks in my RAID6 is NOT "linux raid autodetect"?
I simply want to start my RAID6 as /dev/md8.

I use mdadm-3.1.4 and linux-2.6.36-gentoo-r8. The superblocks seem to be correct - "mdadm -E /dev/sd[a,b,c,d,e,f]1 | grep Name" returns

           Name : Atlas:8  (local to host Atlas)
           Name : Atlas:8  (local to host Atlas)
           Name : Atlas:8  (local to host Atlas)
           Name : Atlas:8  (local to host Atlas)
           Name : Atlas:8  (local to host Atlas)
           Name : Atlas:8  (local to host Atlas)

If I get this right, this array should be started as /dev/md8. However, it does not. I have tried several combinations of configurations ("linux raid autodetect" without using mdadm for assembly, "linux" together with mdadm, and so on), but without any success.
Any help would be really apprectiated.

Best regards,
Yann

After you do various changes to your mdadm.conf, don't forget to update your initrd. AFAIK mdadm.conf is copied into the compressed initrd image to be used early at boot, and this copy has to be kept in sync with your version in /etc.

Typically you can update initrd for all installed kernels by running something like:

update-initramfs -k all -u

With respect,
Roman

Subject: Re: RAID6 simply does not start as /dev/md8
From: Roman Mamedov <[email protected]>
To: Yann Ormanns <[email protected]>
Date: 2011-04-08 16:15 (+0600)

Thank you for your quick response. I do not use an initrd - during the boot process the kernel reconnects and assembles md1,md2,md3,md5,md6 and md7 (these are all RAID1) without any problems. Only the assembly of the RAID6 does not work at all.

Best regards,
Yann

Do you have perhaps old raid metadata on the disks? Are you certain you have no raid metadata on the whole disks? you can check with mdadm --Examine /dev/sdc

Cheers

Rudy

Subject: Re: RAID6 simply does not start as /dev/md8
From: Rudy Zijlstra <[email protected]>
To: Yann Ormanns <[email protected]>
Date: 2011-04-08 16:36 (+0200)

To eliminate this possibility, I "zeroed" all six disks before I created the array:
dd if=/dev/zero of=/dev/sda bs=1M
...
dd if=/dev/zero of=/dev/sdf bs=1M

Output of mdadm -E /dev/sda1:

	/dev/sda1:
        	  Magic : a92b4efc
	        Version : 1.2
	    Feature Map : 0x0
	     Array UUID : 6286b922:2da22a6c:f8fba333:fc88f457
        	   Name : Atlas:8  (local to host Atlas)
	  Creation Time : Fri Apr  8 00:58:12 2011
	     Raid Level : raid6
	   Raid Devices : 6

	 Avail Dev Size : 3907025072 (1863.01 GiB 2000.40 GB)
	     Array Size : 15628095488 (7452.06 GiB 8001.58 GB)
	  Used Dev Size : 3907023872 (1863.01 GiB 2000.40 GB)
	    Data Offset : 2048 sectors
	   Super Offset : 8 sectors
        	  State : clean
	    Device UUID : 1a016923:a408d8ee:d8f6625f:bcf55b66

	    Update Time : Fri Apr  8 09:53:13 2011
	       Checksum : 69969b9b - correct
        	 Events : 19

	         Layout : left-symmetric
	     Chunk Size : 512K

	   Device Role : Active device 0
	   Array State : AAAAAA ('A' == active, '.' == missing)

The complete output of mdadm -E /dev/sd[a,b,c,d,e,f]1 can be found here:
http://nopaste.info/658b72791c_nl.html

Best regards,
Yann

Just to humour me, the output without the partition number? so /dev/sda ?

Cheers,

Rudy

Subject: Re: RAID6 simply does not start as /dev/md8
From: Rudy Zijlstra <[email protected]>
To: Yann Ormanns <[email protected]>
Date: 2011-04-08 20:22 (+0200)

Output of "mdadm -E /dev/sd[a,b,c,d,e,f]":
mdadm: No md superblock detected on /dev/sda.
mdadm: No md superblock detected on /dev/sdb.
mdadm: No md superblock detected on /dev/sdc.
mdadm: No md superblock detected on /dev/sdd.
mdadm: No md superblock detected on /dev/sde.
mdadm: No md superblock detected on /dev/sdf.

Best regards,
Yann

Why? From your description, it sounds to me like md8 is built using
the whole disk on all 6 drives. If so, there's really no need to partition the drives.

You should not need the "devices=..." entry, but I would include the name and the UUID. Here are the relevant parts of my mdadm.conf:

# by default, scan all partitions (/proc/partitions) for MD superblocks.
# alternatively, specify devices to scan, using wildcards if desired.
DEVICE partitions

# auto-create devices with Debian standard permissions
CREATE owner=root group=disk mode=0660 auto=yes

# automatically tag new arrays as belonging to the local system
HOMEHOST <system>

ARRAY /dev/md0 level=raid6 metadata=1.2 num-devices=14 UUID=5ff10d73:a096195f:7a646bba:a68986ca name=RAID-Server:0

ARRAY /dev/md1 level=raid1 metadata=0.90 num-devices=2 UUID=f0a63cae:10406a7a:fa72b0ce:3d8e1057

ARRAY /dev/md2 level=raid1 metadata=1.2 num-devices=2 UUID=4b466602:fb81286c:4ad8dc5c:ad0bd065 name=RAID-Server:2

ARRAY /dev/md3 level=raid1 metadata=1.2 num-devices=2 UUID=5bc11cda:e1b4065f:fbf2fca5:8b12e0ba name=RAID-Server:3

I had similar problems with a name conflict, but that doesn't seem
to be the case here.

Mdadm builds any arrays local to the system at boot. Why do you
have grief with the kernel starting the array?

First of all:
I finally solved the problem by creating an initrd, which assembles all arrays, so I could deactivate CONFIG_MD_AUTODETECT in the kernel. I also do not need /etc/mdadm.conf anymore - and it works like a charm.

Thanks to everybody who attended to this problem!
The summary can be found here:
https://forums.gentoo.org/viewtopic-t-868017-highlight-.html

Subject: RE: RAID6 simply does not start as /dev/md8
From: Leslie Rhorer <[email protected]>
To: Yann Ormanns <[email protected]>
Date: 2011-04-11 10:06 (-0500)

That is correct. I simply wanted to eliminate possible mistakes - after 6 or 7 "useless" syncs of the array I tried _everything_, just to get this thing work.

The problem was, that - although I did not use the filesystem type "linux raid autodetect" for my RAID6 - the kernel still seemed to assemble it. Now, without its autodetection, it works fine.

Best regards,
Yann