[mdlug] "Device busy" errors and RAID setup issues

[David McMillan]

Jeff Hanson wrote:
> On Tue, Nov 17, 2009 at 7:54 PM, David McMillan <skyefire at skyefire.org> wrote:
>>        Busy, busy, busy.  But I finally got time to sit down and do all of
>> this.  So, I did:
>>        dd if=zero of=/dev/sdx for all five of the drives I want to RAID.
>>        Confirmed that each disk no longer had a partition table using fdisk.
>>        Used fdisk to partition each drive as Linux Raid Autodetect.
>>        Ran the following command:  sudo mdadm --create --verbose /dev/md0
>> --level=10 --raid-devices=4 /dev/sdb1 /dev/sdc1 /dev/sdd1 /dev/sde1
>> --spare-devices=1 /dev/sdf1
>>        And what I got was:
>> mdadm: layout defaults to n1
>> mdadm: chunk size defaults to 64K
>> mdadm: Cannot open /dev/sde1: Device or resource busy
>> mdadm: /dev/sdf1 appears to be part of a raid array:
>>     level=raid10 devices=4 ctime=Sat Oct 10 02:50:35 2009
>> mdadm: create aborted
>>
> 
> mdadm --zero-superblock  should be enough to clear the old RAID data.
> 
> Try setting up the array with the minimal number of members (like
> RAID0) with missing members then add spare drives to complete it.  I
> think that's what I did with my server but I was using RAID1 and only
> had one member to add.

	Well, I did that.  Successfully created /dev/md0.  /proc/mdstat showed 
/dev/md0 active and containing all of the hard drives except the one I'd 
left as "missing" in the initial setup.  I even put a label on it and 
created a filesystem on it.
	Then I rebooted the server.  And now /dev/md0 is gone.  /proc/mdstat 
just gives me this:
sudo mdadm /dev/md0 --add /dev/sde1
mdadm: error opening /dev/md0: No such file or directory
david at Archive:~$ cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] 
[raid4] [raid10]
md_d0 : inactive sdf1[3](S)
       976759936 blocks

	/dev/sdf1 was the last drive I added to /dev/md0.  /dev/sde1 was the 
one I couldn't add due to the "device busy" error.  Before the reboot, 
/dev/sdb, c, d, and f all showed in /proc/mdstat.  What the heck?  Why 
would a RAID array that was fully built and appeared to be fine before 
the reboot just vanish after?  And why does sdf1, and only sdf1, still 
show up in /proc/mdstat?

	While we're on the subject, I've got one other question:  when adding a 
drive to a RAID array with mdadm, how do you tell mdadm to treat the 
drive as a member of the main array, or as a hot spare?  So far, all the 
HOWTOs I've found on adding drives just use --add for both actions, 
without any apparent differentiation between regular and spare drives.