<quote who="Yuri"/>
> a) How do you know when one of the harddrive fails? Do i have to look in the
> messages all the time for it or will it ( or can i make it) beep on bootup or
> smth like that to let me know that it failed?
cat /proc/mdstat ... From mine (software RAID-1):
Personalities : [raid1]
read_ahead 1024 sectors
md0 : active raid1 ide/host0/bus1/target0/lun0/part2[0]
ide/host0/bus0/target0/lun0/part2[1]
38832576 blocks [2/2] [UU]
unused devices: <none>
Those two U's mean "up". :-) It also gives you a cute little status bar
whilst the drives are syncing. Also, most distros will have a cronjob that
mails you if a hard drive has failed.
> b) Is it CPU-intensive? I have 4 drives right now in P133 with 98MBram all
> using Promise cards and it's running 2.4.18-xfs kernel (rh7.2), and not sure
> if my setup will handle RAID-5.
RAID is pretty simple stuff, so it's usually I/O speed that you should look
into if it's important to you. The gnome.org machines run on software
RAID-5, and we haven't had too many problems... However, if there are lot of
writes going on, it will take a fair bit of cpu / kernel time.
> c) This is silly, but can i add more drives later to that RAID-5? Like another
> 1 or 2 drives and not loose anything, or should i be looking at LVM rather
> than raid5?
Because of the way the data is laid out, you can't add more drives later
with RAID-5 without backing up and creating a new array. However, if you're
just interested in more redundancy, you can create a RAID-5 array of RAID-1
arrays. This sounds silly, but sometimes it's useful.
LVM is worth looking at, but is going to change in 2.6 anyway. :-)
- Jeff
--
chown -R us:us yourbase
--
SLUG - Sydney Linux User's Group - http://slug.org.au/
More Info: http://lists.slug.org.au/listinfo/slug
