I have 4 ST2000DL003-9VT166 (2Tbyte) disks in a RAID 5 array. Because
of the size I built them as a GPT partitioned disk. They were originally
built on a CentOS 5.x machine but more recently plugged into a CentOS
6.2 machine where they were detected just fine
e.g.
% parted /dev/sdj print
Model: ATA ST2000DL003-9VT1 (scsi)
Disk /dev/sdj: 2000GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Number Start End Size File system Name Flags
3 2097kB 2000GB 2000GB Linux RAID raid
The MBR is empty except for the standard "protection" partition
% fdisk -l /dev/sdj
WARNING: GPT (GUID Partition Table) detected on '/dev/sdj'! The util fdisk doesn't support GPT. Use GNU Parted.
Disk /dev/sdj: 2000.4 GB, 2000398934016 bytes
256 heads, 63 sectors/track, 242251 cylinders
Units = cylinders of 16128 * 512 = 8257536 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000
Device Boot Start End Blocks Id System
/dev/sdj1 1 242252 1953514583+ ee GPT
% dd if=/dev/sdj count=1 2> /dev/null | hexdump
0000000 0000 0000 0000 0000 0000 0000 0000 0000
*
00001c0 0002 ffee ffff 0001 0000 88af e8e0 0000
00001d0 0000 0000 0000 0000 0000 0000 0000 0000
*
00001f0 0000 0000 0000 0000 0000 0000 0000 aa55
0000200
So far, so normal. This works fine under 2.6.32-220.23.1.el6.x86_64
Personalities : [raid1] [raid10] [raid6] [raid5] [raid4]
md127 : active raid5 sdj3[2] sdi2[1] sdk4[3] sdh1[0]
5860537344 blocks level 5, 64k chunk, algorithm 2 [4/4] [UUUU]
However, I just patched to CentOS 6.3 and on reboot this array failed
to be built. The 2.6.32-279 kernel complained that /dev/sdj was too
similar to /dev/sdj3. But I reboot to -220.23.1 then it works.
And, indeed, if I run "mdadm --examine /dev/sdj" then it _does_ look
like mdadm thinks the raw device is part of the array! But with a
different name...
% mdadm --examine /dev/sdj
/dev/sdj:
Magic : a92b4efc
Version : 0.90.00
UUID : 79b0ccbb:8f11154b:7df88b83:4a0b6975
Creation Time : Thu Sep 8 22:14:01 2011
Raid Level : raid5
Used Dev Size : 1953512448 (1863.01 GiB 2000.40 GB)
Array Size : 5860537344 (5589.04 GiB 6001.19 GB)
Raid Devices : 4
Total Devices : 4
Preferred Minor : 5
Update Time : Tue Jul 10 09:10:38 2012
State : clean
Active Devices : 4
Working Devices : 4
Failed Devices : 0
Spare Devices : 0
Checksum : 8cc92232 - correct
Events : 36410
Layout : left-symmetric
Chunk Size : 64K
Number Major Minor RaidDevice State
this 2 8 147 2 active sync /dev/sdj3
0 0 8 113 0 active sync /dev/sdh1
1 1 8 130 1 active sync /dev/sdi2
2 2 8 147 2 active sync /dev/sdj3
3 3 8 164 3 active sync /dev/sdk4
This is the same output as if I'd run it against /dev/sdj3
What's odd is that I also have 4*1Tbyte disks in another raid array but
these are partitioned with standard MBR partitions, and these are not
reporting this problem.
% dd if=/dev/sdd count=1 2> /dev/null | hexdump
0000000 0000 0000 0000 0000 0000 0000 0000 0000
*
00001b0 0000 0000 0000 0000 0000 0000 0000 0100
00001c0 0001 fefd ffff 003f 0000 5982 7470 0000
00001d0 0000 0000 0000 0000 0000 0000 0000 0000
*
00001f0 0000 0000 0000 0000 0000 0000 0000 aa55
0000200
% mdadm --examine /dev/sdd
/dev/sdd:
MBR Magic : aa55
Partition[0] : 1953520002 sectors at 63 (type fd)
So, finally, the question: Is there anything I could do inside the MBR
of these GPT disks to stop mdadm from thinking the whole disk is part
of the array? Is there anything I can do that'll let the new kernel
assemble the array?
Thanks!
--
rgds
Stephen