Configure RAID before deploying the OS

Overview

xCAT provides an user interface linuximage.partitionfile to specify the customized partition script for diskful provision, and provides some default partition scripts.

Deploy Diskful Nodes with RAID1 Setup on RedHat

xCAT provides a partition script raid1_rh.sh which configures RAID1 across 2 disks on RHEL 7.x operating systems.

In most scenarios, the sample partitioning script is sufficient to create a basic RAID1 across two disks and is provided as a sample to build upon.

  1. Obtain the partition script:

    mkdir -p /install/custom/partition/
    wget https://raw.githubusercontent.com/xcat2/xcat-extensions/master/partition/raid1_rh.sh \
         -O /install/custom/partition/raid1_rh.sh
    
  2. Associate the partition script to the osimage:

    chdef -t osimage -o rhels7.3-ppc64le-install-compute \
          partitionfile="s:/install/custom/partition/raid1_rh.sh"
    
  3. Provision the node:

    rinstall cn1 osimage=rhels7.3-ppc64le-install-compute
    

After the diskful nodes are up and running, you can check the RAID1 settings with the following process:

mount command shows the /dev/mdx devices are mounted to various file systems, the /dev/mdx indicates that the RAID is being used on this node.

# mount
...
/dev/md1 on / type xfs (rw,relatime,attr2,inode64,noquota)
/dev/md0 on /boot type xfs (rw,relatime,attr2,inode64,noquota)
/dev/md2 on /var type xfs (rw,relatime,attr2,inode64,noquota)

The file /proc/mdstat includes the RAID devices status on the system, here is an example of /proc/mdstat in the non-multipath environment:

# cat /proc/mdstat
Personalities : [raid1]
md2 : active raid1 sdk2[0] sdj2[1]
      1047552 blocks super 1.2 [2/2] [UU]
        resync=DELAYED
      bitmap: 1/1 pages [64KB], 65536KB chunk

md3 : active raid1 sdk3[0] sdj3[1]
      1047552 blocks super 1.2 [2/2] [UU]
        resync=DELAYED

md0 : active raid1 sdk5[0] sdj5[1]
      524224 blocks super 1.0 [2/2] [UU]
      bitmap: 0/1 pages [0KB], 65536KB chunk

md1 : active raid1 sdk6[0] sdj6[1]
      973998080 blocks super 1.2 [2/2] [UU]
      [==>..................]  resync = 12.8% (125356224/973998080) finish=138.1min speed=102389K/sec
      bitmap: 1/1 pages [64KB], 65536KB chunk

unused devices: <none>

On the system with multipath configuration, the /proc/mdstat looks like:

# cat /proc/mdstat
Personalities : [raid1]
md2 : active raid1 dm-11[0] dm-6[1]
      291703676 blocks super 1.1 [2/2] [UU]
      bitmap: 1/1 pages [64KB], 65536KB chunk

md1 : active raid1 dm-8[0] dm-3[1]
      1048568 blocks super 1.1 [2/2] [UU]

md0 : active raid1 dm-9[0] dm-4[1]
      204788 blocks super 1.0 [2/2] [UU]

unused devices: <none>

The command mdadm can query the detailed configuration for the RAID partitions:

mdadm --detail /dev/md2

Deploy Diskful Nodes with RAID1 Setup on SLES

xCAT provides one sample autoyast template files with the RAID1 settings /opt/xcat/share/xcat/install/sles/service.raid1.sles11.tmpl. You can customize the template file and put it under /install/custom/install/<platform>/ if the default one does not match your requirements.

Here is the RAID1 partitioning section in service.raid1.sles11.tmpl:

<partitioning config:type="list">
   <drive>
     <device>/dev/sda</device>
     <partitions config:type="list">
       <partition>
         <format config:type="boolean">false</format>
         <partition_id config:type="integer">65</partition_id>
         <partition_nr config:type="integer">1</partition_nr>
         <partition_type>primary</partition_type>
         <size>24M</size>
       </partition>
       <partition>
         <format config:type="boolean">false</format>
         <partition_id config:type="integer">253</partition_id>
         <partition_nr config:type="integer">2</partition_nr>
         <raid_name>/dev/md0</raid_name>
         <raid_type>raid</raid_type>
         <size>2G</size>
       </partition>
       <partition>
         <format config:type="boolean">false</format>
         <partition_id config:type="integer">253</partition_id>
         <partition_nr config:type="integer">3</partition_nr>
         <raid_name>/dev/md1</raid_name>
         <raid_type>raid</raid_type>
         <size>max</size>
       </partition>
     </partitions>
     <use>all</use>
   </drive>
   <drive>
     <device>/dev/sdb</device>
     <partitions config:type="list">
       <partition>
         <format config:type="boolean">false</format>
         <partition_id config:type="integer">131</partition_id>
         <partition_nr config:type="integer">1</partition_nr>
         <partition_type>primary</partition_type>
         <size>24M</size>
       </partition>
       <partition>
         <format config:type="boolean">false</format>
         <partition_id config:type="integer">253</partition_id>
         <partition_nr config:type="integer">2</partition_nr>
         <raid_name>/dev/md0</raid_name>
         <raid_type>raid</raid_type>
         <size>2G</size>
       </partition>
       <partition>
         <format config:type="boolean">false</format>
         <partition_id config:type="integer">253</partition_id>
         <partition_nr config:type="integer">3</partition_nr>
         <raid_name>/dev/md1</raid_name>
         <raid_type>raid</raid_type>
         <size>max</size>
       </partition>
     </partitions>
     <use>all</use>
   </drive>
  <drive>
    <device>/dev/md</device>
    <partitions config:type="list">
      <partition>
        <filesystem config:type="symbol">reiser</filesystem>
        <format config:type="boolean">true</format>
        <mount>swap</mount>
        <partition_id config:type="integer">131</partition_id>
        <partition_nr config:type="integer">0</partition_nr>
        <raid_options>
          <chunk_size>4</chunk_size>
          <parity_algorithm>left-asymmetric</parity_algorithm>
          <raid_type>raid1</raid_type>
        </raid_options>
      </partition>
      <partition>
        <filesystem config:type="symbol">reiser</filesystem>
        <format config:type="boolean">true</format>
        <mount>/</mount>
        <partition_id config:type="integer">131</partition_id>
        <partition_nr config:type="integer">1</partition_nr>
        <raid_options>
          <chunk_size>4</chunk_size>
          <parity_algorithm>left-asymmetric</parity_algorithm>
          <raid_type>raid1</raid_type>
        </raid_options>
      </partition>
    </partitions>
    <use>all</use>
  </drive>
</partitioning>

The samples above created one 24MB PReP partition on each disk, one 2GB mirrored swap partition and one mirrored / partition uses all the disk space. If you want to use different partitioning scheme in your cluster, modify this RAID1 section in the autoyast template file accordingly.

Since the PReP partition can not be mirrored between the two disks, some additional postinstall commands should be run to make the second disk bootable, here the commands needed to make the second disk bootable:

# Set the second disk to be bootable for RAID1 setup
parted -s /dev/sdb mkfs 1 fat32
parted /dev/sdb set 1 type 6
parted /dev/sdb set 1 boot on
dd if=/dev/sda1 of=/dev/sdb1
bootlist -m normal sda sdb

The procedure listed above has been added to the file /opt/xcat/share/xcat/install/scripts/post.sles11.raid1 to make it be automated. The autoyast template file service.raid1.sles11.tmpl will include the content of post.sles11.raid1, so no manual steps are needed here.

After the diskful nodes are up and running, you can check the RAID1 settings with the following commands:

Mount command shows the /dev/mdx devices are mounted to various file systems, the /dev/mdx indicates that the RAID is being used on this node.

server:~ # mount
/dev/md1 on / type reiserfs (rw)
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
debugfs on /sys/kernel/debug type debugfs (rw)
devtmpfs on /dev type devtmpfs (rw,mode=0755)
tmpfs on /dev/shm type tmpfs (rw,mode=1777)
devpts on /dev/pts type devpts (rw,mode=0620,gid=5)

The file /proc/mdstat includes the RAID devices status on the system, here is an example of /proc/mdstat:

server:~ # cat /proc/mdstat
Personalities : [raid1] [raid0] [raid10] [raid6] [raid5] [raid4]
md0 : active (auto-read-only) raid1 sda2[0] sdb2[1]
      2104500 blocks super 1.0 [2/2] [UU]
      bitmap: 0/1 pages [0KB], 128KB chunk

md1 : active raid1 sda3[0] sdb3[1]
      18828108 blocks super 1.0 [2/2] [UU]
      bitmap: 0/9 pages [0KB], 64KB chunk

unused devices: <none>

The command mdadm can query the detailed configuration for the RAID partitions:

mdadm --detail /dev/md1

Disk Replacement Procedure

If any one disk fails in the RAID1 array, do not panic. Follow the procedure listed below to replace the failed disk.

Faulty disks should appear marked with an (F) if you look at /proc/mdstat:

# cat /proc/mdstat
Personalities : [raid1]
md2 : active raid1 dm-11[0](F) dm-6[1]
      291703676 blocks super 1.1 [2/1] [_U]
      bitmap: 1/1 pages [64KB], 65536KB chunk

md1 : active raid1 dm-8[0](F) dm-3[1]
      1048568 blocks super 1.1 [2/1] [_U]

md0 : active raid1 dm-9[0](F) dm-4[1]
      204788 blocks super 1.0 [2/1] [_U]

unused devices: <none>

We can see that the first disk is broken because all the RAID partitions on this disk are marked as (F).

Remove the failed disk from RAID array

mdadm is the command that can be used to query and manage the RAID arrays on Linux. To remove the failed disk from RAID array, use the command:

mdadm --manage /dev/mdx --remove /dev/xxx

Where the /dev/mdx are the RAID partitions listed in /proc/mdstat file, such as md0, md1 and md2; the /dev/xxx are the backend devices like dm-11, dm-8 and dm-9 in the multipath configuration and sda5, sda3 and sda2 in the non-multipath configuration.

Here is the example of removing failed disk from the RAID1 array in the non-multipath configuration:

mdadm --manage /dev/md0 --remove /dev/sda3
mdadm --manage /dev/md1 --remove /dev/sda2
mdadm --manage /dev/md2 --remove /dev/sda5

Here is the example of removing failed disk from the RAID1 array in the multipath configuration:

mdadm --manage /dev/md0 --remove /dev/dm-9
mdadm --manage /dev/md1 --remove /dev/dm-8
mdadm --manage /dev/md2 --remove /dev/dm-11

After the failed disk is removed from the RAID1 array, the partitions on the failed disk will be removed from /proc/mdstat and the mdadm --detail output also.

# cat /proc/mdstat
Personalities : [raid1]
md2 : active raid1 dm-6[1]
      291703676 blocks super 1.1 [2/1] [_U]
      bitmap: 1/1 pages [64KB], 65536KB chunk

md1 : active raid1 dm-3[1]
      1048568 blocks super 1.1 [2/1] [_U]

md0 : active raid1 dm-4[1]
      204788 blocks super 1.0 [2/1] [_U]

unused devices: <none>

# mdadm --detail /dev/md0
/dev/md0:
        Version : 1.0
  Creation Time : Tue Jul 19 02:39:03 2011
     Raid Level : raid1
     Array Size : 204788 (200.02 MiB 209.70 MB)
  Used Dev Size : 204788 (200.02 MiB 209.70 MB)
   Raid Devices : 2
  Total Devices : 1
    Persistence : Superblock is persistent

    Update Time : Wed Jul 20 02:00:04 2011
          State : clean, degraded
 Active Devices : 1
Working Devices : 1
 Failed Devices : 0
  Spare Devices : 0

           Name : c250f17c01ap01:0  (local to host c250f17c01ap01)
           UUID : eba4d8ad:8f08f231:3c60e20f:1f929144
         Events : 26

    Number   Major   Minor   RaidDevice State
       0       0        0        0      removed
       1     253        4        1      active sync   /dev/dm-4

Replace the disk

Depends on the hot swap capability, you may simply unplug the disk and replace with a new one if the hot swap is supported; otherwise, you will need to power off the machine and replace the disk and the power on the machine. Create partitions on the new disk

The first thing we must do now is to create the exact same partitioning as on the new disk. We can do this with one simple command:

sfdisk -d /dev/<good_disk> | sfdisk /dev/<new_disk>

For the non-mulipath configuration, here is an example:

sfdisk -d /dev/sdb | sfdisk /dev/sda

For the multipath configuration, here is an example:

sfdisk -d /dev/dm-1 | sfdisk /dev/dm-0

If you got error message “sfdisk: I don’t like these partitions - nothing changed.”, you can add --force option to the sfdisk command:

sfdisk -d /dev/sdb | sfdisk /dev/sda --force

You can run:

fdisk -l

To check if both hard drives have the same partitioning now.

Add the new disk into the RAID1 array

After the partitions are created on the new disk, you can use command:

mdadm --manage /dev/mdx --add /dev/xxx

To add the new disk to the RAID1 array. Where the /dev/mdx are the RAID partitions like md0, md1 and md2; the /dev/xxx are the backend devices like dm-11, dm-8 and dm-9 in the multipath configuration and sda5, sda3 and sda2 in the non-multipath configuration.

Here is an example for the non-multipath configuration:

mdadm --manage /dev/md0 --add /dev/sda3
mdadm --manage /dev/md1 --add /dev/sda2
mdadm --manage /dev/md2 --add /dev/sda5

Here is an example for the multipath configuration:

mdadm --manage /dev/md0 --add /dev/dm-9
mdadm --manage /dev/md1 --add /dev/dm-8
mdadm --manage /dev/md2 --add /dev/dm-11

All done! You can have a cup of coffee to watch the fully automatic reconstruction running…

While the RAID1 array is reconstructing, you will see some progress information in /proc/mdstat:

# cat /proc/mdstat
Personalities : [raid1]
md2 : active raid1 dm-11[0] dm-6[1]
      291703676 blocks super 1.1 [2/1] [_U]
      [>....................]  recovery =  0.7% (2103744/291703676) finish=86.2min speed=55960K/sec
      bitmap: 1/1 pages [64KB], 65536KB chunk

md1 : active raid1 dm-8[0] dm-3[1]
      1048568 blocks super 1.1 [2/1] [_U]
      [=============>.......]  recovery = 65.1% (683904/1048568) finish=0.1min speed=48850K/sec

md0 : active raid1 dm-9[0] dm-4[1]
      204788 blocks super 1.0 [2/1] [_U]
      [===================>.]  recovery = 96.5% (198016/204788) finish=0.0min speed=14144K/sec

unused devices: <none>

After the reconstruction is done, the /proc/mdstat becomes like:

# cat /proc/mdstat
Personalities : [raid1]
md2 : active raid1 dm-11[0] dm-6[1]
      291703676 blocks super 1.1 [2/2] [UU]
      bitmap: 1/1 pages [64KB], 65536KB chunk

md1 : active raid1 dm-8[0] dm-3[1]
      1048568 blocks super 1.1 [2/2] [UU]

md0 : active raid1 dm-9[0] dm-4[1]
      204788 blocks super 1.0 [2/2] [UU]

unused devices: <none>

Make the new disk bootable

If the new disk does not have a PReP partition or the PReP partition has some problem, it will not be bootable, here is an example on how to make the new disk bootable, you may need to substitute the device name with your own values.

  • [RHEL]:

    mkofboot .b /dev/sda
    bootlist -m normal sda sdb
    
  • [SLES]:

    parted -s /dev/sda mkfs 1 fat32
    parted /dev/sda set 1 type 6
    parted /dev/sda set 1 boot on
    dd if=/dev/sdb1 of=/dev/sda1
    bootlist -m normal sda sdb