Well, still doing some testing here w/ boot raid and have come up with some
more
observations & misc that I thought to send to the list.
In one instance I was playing with two different machines and found a hairy
problem
when moving arrays around with the same name. Ie:
- Need to add to raid supperblocks some kind of unique identifier. ie
say the user has two arrays each on a separate machine. Each array is
called
'md1' and is set to auto-run (0xfd). One machine dies (hardware
problem).
The user wants to physically move his disks to the second machine.
It appears under the current code that the second machine will use
whatever disk
that is auto-run and has the highest event count. This turns into a
crap-shoot as to
which array will get auto-started on the next boot.
- Mainly due to my own stupidity here, but two things I've gotten bit by:
I have an array that is in degraded mode and want to hot-add a
replacement
partition to the array. Unfortunately Murphy struck and I added the
wrong partition
to the array. This was compounded by the problem that the resync
process
starts automatically.
When hot-adding a disk to an array a flag should be available to the
user (to the
raidhotadd command) to flag the new partition as 'inactive' or 'active'.
If the partition
is marked as 'inactive' the resync process is not started yet, only on
'actively'
marked partitions. This should give fools like me a chance to remove a
partition
before data is overwritten (and to verify in /proc/mdstat).
something like:
raidhotadd /dev/md1 -i /dev/sdb1 (adding /dev/sdb1 as an
inactive partition)
raidhotadd /dev/md2 -a /dev/sdb2 (adding /dev/sdb2 as an active
partition)
raidhotadd /dev/md1 -a (activating whatever partitions that
were already added?)
Lastly (also due to my fooling around here). This ties into the first issue
on moving disk
from different systems/arrays.
Say you have a RAID-1, one disk is dead. You have another raid-1 (same
configuration) on another machine but it's no longer being used. You
want
to move that un-used disk to the degraded RAID-1. There should be an
option to indicate that you want to use the current mounted disk as the
primary
disk in the mirror and wipe out the newly added one irregardless of
event count(s).
Any comments/heckles?
Steve
----------------------------------------------
Steve Costaras Unix Systems Administrator
[EMAIL PROTECTED] Celestial Haven
Ph: (630) 852-9765 FAX: (630) 852-9263
-
If you don't understand, read the RFC's. If
you do, re-write them. HTTP://3502543125/rfcs
----------------------------------------------
