Making Software RAID 0 Arrays in Modern Linux

With a new disk, you have to prepare or create partitions, using the parted command.  
For md RAID arrays here is only a need for a single partition (ie, number 1, as in /dev/sdc1).  
See the following web page on preparation (but not for mdadm, since this page used raidtools instead):
It summarizes as follows:
 Prepare The Partitions With FDISK:
fdisk /dev/sdc
 Use FDISK Help  (command is m)
 Set The ID Type To FD
 Make Sure The Change Occurred (command p yields the next two lines)
   Device Boot      Start         End      Blocks   Id  System
/dev/sdc1               1       48641   390708801   fd  Linux raid autodetect
 Save The Changes (command is w)

To recreate a new array, first unmount the old array.
To disassemble the old array (temporarily):
[root@mps-pc10 ~]# mdadm --stop /dev/md0

To verify that array is temporarily disassembled:
[root@mps-pc10 ~]# cat /proc/mdstat
Personalities : [raid0]
unused devices: <none>

To look at component disks in old array:
[root@mps-pc10 ~]# mdadm --query /dev/md0
/dev/md0: is an md device which is not active
[root@mps-pc10 ~]# mdadm --examine /dev/sdd1
          Magic : a92b4efc
        Version : 00.90.00
           UUID : 023c49a0:72d521d1:223c6636:4abdbfc7
  Creation Time : Mon May  8 17:36:39 2006
     Raid Level : raid0
    Device Size : 390708736 (372.61 GiB 400.09 GB)
   Raid Devices : 4
  Total Devices : 4
Preferred Minor : 0

    Update Time : Mon May  8 17:36:39 2006
          State : dirty
 Active Devices : 4
Working Devices : 4
 Failed Devices : 0
  Spare Devices : 0
       Checksum : 2b4167f3 - correct
         Events : 0.1

     Chunk Size : 64K

      Number   Major   Minor   RaidDevice State
this     3       8       49        3      active sync   /dev/sdd1
   0     0       8        1        0      active sync   /dev/sda1
   1     1       8       17        1      active sync   /dev/sdb1
   2     2       8       33        2      active sync   /dev/sdc1
   3     3       8       49        3      active sync   /dev/sdd1

Making a new smaller md array:

[root@mps-pc10 ~]# mdadm --create /dev/md0 --level=0 --raid-devices=2 /dev/sda1 /dev/sdb1

mdadm: /dev/sda1 appears to contain an ext2fs file system
    size=1562834944K  mtime=Wed Jul 12 20:11:41 2006
mdadm: /dev/sda1 appears to be part of a raid array:
    level=0 devices=4 ctime=Mon May  8 17:36:39 2006
mdadm: /dev/sdb1 appears to be part of a raid array:
    level=0 devices=4 ctime=Mon May  8 17:36:39 2006
Continue creating array? y
mdadm: array /dev/md0 started.

[root@mps-pc10 ~]# cat /proc/mdstat
Personalities : [raid0]
md0 : active raid0 sdb1[1] sda1[0]
      781417472 blocks 64k chunks

unused devices: <none>

To create a filesystem on the new smaller md0 and run the badblocks program.

mke2fs -b 4096 -c  -j -T largefile4 -v /dev/md0

I wanted to have a journal on these filesystems, so the second one (/dev/md1, made from /dev/sd[cd]1) is done with the ext3 option 
(I believe that this is equivalent to adding -j to the above command line).

mkfs.ext3 -b 4096 -c -T largefile4 -v /dev/md1

Finally, all you have to do is to edit /etc/fstab to make things mount automatically.  Fist mount them by hand and test.

The only thing that I did not do is to use the -R option in mke2fs.  I thought that the default was probably fine, but maybe this is a problem.  But the web page said as follows:
 Format The New RAID Set
Your new RAID device now has to be formatted. The next example uses the -j qualifier to ensure that a journaling filesystem is created. Here a block size of 4KB (4096 bytes) is used with each chunk, which is comprised of 8 blocks. It is very important that the chunk-size parameter in the /etc/raidtab file match the value of the block size multiplied by the stride value in the command below. If the values don't match, you will get parity errors:
[root@bigboy tmp]# mke2fs -j -b 4096 -R stride=8 /dev/md0

I think that mdadm takes care of this automatically, by default.

To check for bad blocks on the fielsystem, run
e2fsck -c -c -k -C -y /dev/md1

The filesystem MUST be unmounted before checking for bad blocks.