raid1_rh.sh: Make md RAID creation fail proof (fixes #40)

[Obihoernchen]

• Make sure previous md RAIDs are deleted c84d0327c4c5d18bb7e1a4cba1c2cca482a0e90b
• Disable md RAID resync during installation a074747deaac847fe424b7a6f63af551d3f068e3

See #40 for more information.

[Obihoernchen]

Hey @neo954 please have a look at my latest commits (especially 0e415d6).

1. Search for disk1 (ex: sda) and disk2 (ex. sdb) in /proc/mdstat to find all md RAIDs on these disks
2. Stop all md RAIDs found in 1.
3. Finally, the superblock zeroing is done for all partitions on disk1 and disk2 only

Works like this (I've just added an echo for each mdadm command for the output):

[rh75-compute-install.partition] get a disk: sdb
[rh75-compute-install.partition] get a disk: sda
[rh75-compute-install.partition] disk sdb has wwn:0x5000c500aadde735
[rh75-compute-install.partition] disk sda has wwn:0x5000c500a8f628fc
[rh75-compute-install.partition] get disk: (sda sdb) with sort_type:wwn
[rh75-compute-install.partition] the final disk order:
[rh75-compute-install.partition]       sda | /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ /dev/disk/by-id/wwn-0x5000c500a8f628fc /dev/disk/by-path/pci-0004:03:00.0-ata-1.0
[rh75-compute-install.partition]       sdb | /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY /dev/disk/by-id/wwn-0x5000c500aadde735 /dev/disk/by-path/pci-0004:03:00.0-ata-2.0
[rh75-compute-install.partition] the output file is: /tmp/xcat_sorted_disks
[rh75-compute-install.partition] disabling md RAID resync during installation
[rh75-compute-install.partition] stopping md device: /dev/md/0
mdadm --stop /dev/md/0
[rh75-compute-install.partition] stopping md device: /dev/md/1
mdadm --stop /dev/md/1
[rh75-compute-install.partition] stopping md device: /dev/md/2
mdadm --stop /dev/md/2
[rh75-compute-install.partition] stopping md device: /dev/md/3
mdadm --stop /dev/md/3
[rh75-compute-install.partition] stopping md device: /dev/md/4
mdadm --stop /dev/md/4
[rh75-compute-install.partition] zeroing superblocks of /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part*
mdadm --zero-superblock /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part1 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part2 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part3 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part4 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part5 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part6 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W4708JNJ-part7
[rh75-compute-install.partition] zeroing superblocks of /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part*
mdadm --zero-superblock /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part1 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part2 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part3 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part4 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part5 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part6 /dev/disk/by-id/ata-ST1000NX0313_00LY266_00LY265IBM_W470V7KY-part7

[immarvin]

hi @neo954 , thanks for your comment, this partition script is just an example/template for admin to create a raid-1 setup on 2 disks, which is far from a common solution for raid setup, maybe we can put some effort to enhance it in the future, but this PR is adequate to for the issue to fix.

After discuss with @neo954 , we would like to merge this, thanks for your contribution @Obihoernchen

[Obihoernchen]

Yes this script is basically only useful for stateful compute nodes or basic service nodes. Raid 5, Raid 10 etc. need some more attention ;)

[Obihoernchen]

/proc/mdstat is not fully populated during script execution :/ Therefore, this still fails if there is an incomplete md setup existing on the disk.

Unfortunately, checking /proc/mdstat does not work.