raidstop failes - device busy

[Simone]

Hi list,

have a question for the experts. I set up a raid1, 2 disks, each has 3
partitions, /boot /root and the swap partition mirrored (/dev/md0
/dev/md1 /dev/md1). The partitions are hda1 hda2 hda3 - hdc1 hdc2 hdc3.
I started the raid no problems it boots from /dev/md0, and goes well.
Then I needed 2 more eide controllers for adding 2 more disks and I
moved everything to another motherboard that has integrated a promise
raid controller (still with only two disks, waiting the 2 new ones). It
started fine with no need of configuring anything more, recognized the
new motherboard processor ram and so on. The thing is that now it sees
the hda missing, but still I can mount the partitions and work on the
files. In the messages it looks at the hda partitions but says hdc is
fresher and boots from there. So I tried raidstop but it didn't work, it
says the device is busy. Tried mkraid -R but still device busy. Tried to
put in raidtab hda like spare device hoping it would reconstruct the
array on it (when shutting down i got the message - no spare device to
reconstruct array, going on degraded mode) but still no luck. I think
the promise changed the order of the eide ports, but still don't know
enough to say if this could be the problem.
At the moment i've no problems reinstalling linux and setup all in the
new motherboard environment, but i'm trying to understand this for the
future, on a production server in case of disk failure.

The thing I'd like someone to clear is this: is it normal that it says
device busy since it is the /root directory (and the same for the swap
and boot partition that are in use and i cannot unmount), and if so, how
can I replace a disk in case of failure? I'm going to have 2 disks
mirrored with os and 2 mirrored just for data (which i can unmount when
i need to replace and remake raid). 

Almost sure i haven't been clear enough but i've done my best.  
Thanks in advance to anyone that'll spend time reading this,
have a nice day!!

Simone