RAID on Linux

See http://www.sdmachelp.com/linuxraid.html
http://www.sdmachelp.com/linuxlvm.html

GETTING STATUS!!:

    mdadm --detail /dev/mdX
    cat /proc/mdstat  (if you don't know device names.  See below for details.)


[ For Linux RAID + LVM, do a straight RAID1 ext2 partition for /boot, and
  a straight LVM partition for swap ]

http://www.ibiblio.org/pub/Linux/docs/HOWTO/other-formats/html_single/Software-RAID-HOWTO.html

Can safely RAID /boot, /.
With Solaris and Linux software RAID, you can only "Mirror" /, /usr, /opt.
Don't RAID swap.
RAID partition types should be 0xFD

LVM OVER RAID
SWAP.  Don't ever RAID.  Generally, allocate a straight PV so that you
 can re-allocate this space in future.
/, /usr.  Examples use these as straight RAID, but why???  Defeats main
 purpose of LVM.  Try LVMing over mds.
    

mdadm: Best tool to administer kernel md devices.
N.b.! Never repartition drivers when RAID is running.
According to HOWTO, occurrences of raidstop/raidstart in init scripts should
 be disabled.

AUTODETECTION
    Must create devices with "persistent-superblock".
    Partition types must be 0xFD.  ("fd" in fdisk).
    (STOP raid before changing partition types).
ROOT FS
    Choose any of the / mirror underlying partitions:
        device (hd0) /dev/hdc; root (hd0,0); setup (hd0).
    Generally easiest to just let your Linux installer install to the RAID
     device rather than migrating.

cat /proc/mdstat         # [role], [dev#/?].  role < n: working; >= n: spare.
                         # Failed devices show F.
********************************
I believe that after "mdadm create..."-ing, you must create a config file.
(Otherwise you'll have to probe all the devices as shown in the mdadm man 
page).
Once you create /etc/mdadm.conf according to config file, you start up
all the array definitions in it with "mdadm --assemble --scan".
********************************

Linear mode
    mdadm --create --verbose /dev/md0 --level=linear --raid-devices=2 /dev/sdb6 /dev/sdc5

Mirror
    mdadm --create --verbose /dev/md20 --level=mirror --raid-devices=2 /dev/sda1 /dev/sda2
Unlike Sun Volume Manager, this is destructive.
    

RAID 5
    mdadm -S /dev/md0  # Stop it
    mdadm -R /dev/md0  # Restart it
    (Optional args to mke2fs are important for making ext2 filesystems for
    use under RAID 5).

Failure emulation
    mdadm --manage --set-faulty /dev/md1 /dev/sdc2

RECOVER  (only possible with RAID level > 0)
    mdadm /dev/md1 -r /dev/sdc2  # Can only remove Failed devices
    mdadm /dev/md1 -a /dev/sdc2  # Adds back as active member or spare.

MONITORING  Do not run this interactively!  It's a daemon.
    mdadm --monitor --mail=root@localhost --delay=1800 /dev/md2
    For SuSE, set email addr and MDADM_RAIDDEVICES in /etc/sysconfig/mdadm
    (latter a space-delimited string like "/dev/md0 /dev/md1";
     can get list from /proc/md*); then
     enable the init script "mdadmd".  For some reason, 
     "yas2 runlevel" always hangs when I start the service from there.

Simulate a failure
    mdadm --manage -f /dev/md1 /dev/sdc2
    mdadm --detail /dev/mdX  # and wait for the disk to show failure
    mdadm /dev/md1 -r /dev/sdc2  #  Remove the failed disk
    # ... wait for reconstruction, if any.
    mdadm /dev/md1 -a /dev/sdc2  # If hot-spared, it will become a hot spare.