Hi,
I'm attempting to make a RAID 5 on an Apple Network server 700. This
machine has three SCSI busses (one external and two internal). The two
internal are run by a NCR53C825a chip for each bus. There is a hot
swappable backplane that these two busses are split across and I have 7 4
Gb Seagate Barracuda drives. 1 has my boot file system on it and is
regular. The other 6 are spread across the two busses, three per bus. I
am trying to make two RAID 5 arrays (/dev/md0 - 2 disks from bus 0 and 1
disk from bus 1, and /dev/md1 - 1 disk from bus 0 and 2 disks from bus 1).
I can generate /dev/md0 fine. However, I have had continual problems
attempting to generate /dev/md1. For example:
After /dev/md0 has finished constructing and I attempt to mkraid /dev/md1,
I get:
[root@garnet myer]# mkraid --configfile /etc/raidtab /dev/md1
handling MD device /dev/md1
analyzing super-block
/dev/md1: Invalid argument
/var/log/messages shows:
May 12 17:55:53 garnet kernel: bind<sde1,1>
May 12 17:55:53 garnet kernel: nonpersistent superblock ...
May 12 17:55:53 garnet kernel: bind<sdd1,2>
May 12 17:55:53 garnet kernel: nonpersistent superblock ...
May 12 17:55:53 garnet kernel: bind<sdf1,3>
May 12 17:55:53 garnet kernel: nonpersistent superblock ...
May 12 17:55:54 garnet kernel: md: invalid raid superblock magic on sdf1
May 12 17:55:54 garnet kernel: md: bug in file md.c, line 1552
May 12 17:55:54 garnet kernel:
May 12 17:55:54 garnet kernel: **********************************
May 12 17:55:54 garnet kernel: * <COMPLETE RAID STATE PRINTOUT> *
May 12 17:55:54 garnet kernel: **********************************
May 12 17:55:54 garnet kernel: md1: <sdf1><sdd1><sde1> array superblock:
May 12 17:55:54 garnet kernel: SB: (V:2038065518.90.0)
ID:<a0b44513.978103c5.4442051d.4688ed2c> CT:3739f8e9
May 12 17:55:54 garnet kernel: L5 S04193060 ND:3 RD:3 md1 LO:2
CS:4096
May 12 17:55:54 garnet kernel: UT:00000000 ST:0 AD:3 WD:3 FD:0 SD:0
CSUM:00000000 E:00000000
May 12 17:55:54 garnet kernel: D 0: DISK<N:0,sde1(8,65),R:0,S:6>
May 12 17:55:54 garnet kernel: D 1: DISK<N:1,sdd1(8,49),R:1,S:6>
May 12 17:55:54 garnet kernel: D 2: DISK<N:2,sdf1(8,81),R:2,S:6>
(much stuff deleted)
If I stop /dev/md0, I am able to begin the build of /dev/md1 but it
doesn't seem right that I should only be able to have one RAID active at a
time. Now part of this might be because I have been struggling with
figuring out what disks I want to go where and have made many arrays with
different disk combinations so I suppose there could be some old
superblocks hanging around that mkraid doesn't like. Is this normal
behaviour?
I also have had a problem where the last disk I place on /dev/md1 is
marked as bad. It seems to always be the last disk, not any particular
disk because I can change the order in which I assign the disks to the
array and its always disk 2 (out of 0,1,2) that is marked as bad. I'll
post some more about this if I can find the relevant logs AND if it
happens again, as I am currently resyncing /dev/md1 right now.
Other than these peculiarities (and they could be because I'm working with
a hacked kernel since the ANS hasn't been merged into the main kernel
release yet), the new RAID code has been doing just fine on my other Intel
box. Keep up the good work!
Thanks for any help or suggestions.
Kevin
/etc/raidtab follows:
raiddev /dev/md0
raid-level 5
nr-raid-disks 3
chunk-size 4
persistent-superblock 0
parity-algorithm left-symmetric
device /dev/sdb1
raid-disk 0
device /dev/sdc1
raid-disk 1
device /dev/sdg1
raid-disk 2
raiddev /dev/md1
raid-level 5
nr-raid-disks 3
chunk-size 4
persistent-superblock 0
parity-algorithm left-symmetric
device /dev/sde1
raid-disk 0
device /dev/sdd1
raid-disk 1
device /dev/sdf1
raid-disk 2
--
~ Kevin M. Myer
. . Network/System Administrator
/V\ ELANCO School District
// \
/( )\
^`~'^
