Skip to content

Latest commit

 

History

History
160 lines (130 loc) · 6.41 KB

mdadm.md

File metadata and controls

160 lines (130 loc) · 6.41 KB
Raw

MDADM@dubnet (2017)

A guide to madam

Re-name
Shrink
Disk removal & cleanup

Raid 1 (mirror)

https://www.digitalocean.com/community/tutorials/how-to-create-raid-arrays-with-mdadm-on-ubuntu-16-04
how-to , except use parted (not fdisk) to gpt-format disk(s) or gdisk.
cheat-sheet
email-alert

Re-use (large volume) existing disk:

create (degraded) raid1, to transfer existing data :: ref_01

  1. create an empty (broken) mirror with the new disk
  2. copy the existing data onto it
  3. VERIFY USEABLE DATA
  4. create mountpoint
  5. edit fstab to use new array
  6. VERIFY USEABLE DATA
  7. format old drive
  8. add old drive to array

Tools du -sh file_path
du command estimates file_path space usage
The options -sh are (from man du):

-s, --summarize  
display only a total for each argument

-h, --human-readable  
print sizes in human readable format (e.g., 1K 234M 2G)

Prepare disk:

  1. Install disk &...
    1. Identify the device names of your new hard drives
      sudo fdisk -l or lsblk
    2. Determine whether there is any existing raid configuration
      mdadm -E /dev/sd[x,y]
  2. Create partition.
    1. The partition table needs to be GPT because desired volume > 2TB; use parted, not fdisk.
      parted /dev/sd(x)
    2. At the (parted) prompt, create the partition table:
      mklabel gpt
    3. Check the free space on the drive by typing
      print free
    4. Create the partition
      mkpart primary 1M 3001GB This starts the partition at 1M offset giving a 4096 byte alignment. This may or may not be necessary, but won't hurt if its not.
    5. p #displays partition setup
    6. q #quit

Create Raid array:

  1. initialize array /dev/md(#) using /dev/sd(x#) and a missing disk
    sudo mdadm --create --verbose /dev/md0 --level=raid1 --raid-devices=2 /dev/sd(x#) missing
  2. Create a file system on the array:
    sudo mkfs.ext4 /dev/md0
  3. Create a mountpoint for the array, & mount volumes
    sudo mkdir -p /storage/share mount /dev/md0 /mnt/zero
  4. & copy old stuff to new /dev/md0 :: rsync_01 :: rsync_if_again rsync -avhW --progress --no-compress /src/ /dst/

USE the (degraded) array

  1. Mount it
    sudo mount /mnt/md0
  2. Next, save the raid configuration manually to ‘mdadm.conf‘ file using the below command.
    mdadm --detail --scan --verbose >> /etc/mdadm.conf
  3. All good? Add entry to /etc/fstab: [dev/ID]
    /dev/md0 /storage/share auto defaults 0 0
  4. All good? Add entry to /etc/fstab: [UUID]
    1. find device
      cat /proc/mdstat
    2. find uuid of device
      sudo blkid /dev/md0
    3. add entry to fstab
      UUID=4a2b3c6d-0ada-4228-8043-7a2f40a13d4a /storage/share auto defaults 0 0

data verified?

perform 1a through 2f on the old (3tb wd black)

  1. let’s add the other drive...
    mdadm /dev/md0 --add /dev/sd(y#)
  2. Save the raid configuration manually to /etc/mdadm.conf
    mdadm --detail --scan --verbose >> /etc/mdadm.conf
  3. Now you can start using your array. Bear in mind, however, that before it is fully operational it will need to complete its initial sync.
  4. Track/Watch sync progress:
    watch -n1 sudo mdadm --detail /dev/md0
    cat /proc/mdstat

removal of disk from mdadm array (& disk wipe):

  • fail disk
    sudo mdadm /dev/md0 --fail --remove /dev/sdb1
  • stop array
    sudo mdadm --stop md0
  • wipe the filesystem
    wipefs -a /dev/sd(x#)

shrink intact array :: shrink degraded array ::

shrink intact array:

  1. Shrink (Resize) filesystem to just larger than the space used by the content resize2fs /dev/md# [size]
  2. since intact, spends time rebuilding watch -n1 cat /proc/mdstat
  3. Reshape RAID volume to just larger than than the (shrunken) filesystem
    mdadm --grow /dev/md# --size [size+]
  4. Modify partition(s) as needed
  5. Grow the RAID volume to fill the re-worked/new partition
    mdadm --grow /dev/md# -z max
  6. Resize filesystem to occupy entire RAID volume restore filesystem to full raid vol?
    resize2fs /dev/md#

shrink filesystem: resize2fs to-size
shrink (reshape) raid vol: mdadm --grow
shrink partition by delete/recreation: gdisk

(Re)name an array

Helps to communicate array purpose/function/intent. Links the array to /dev/md/newname, and applies newname to its listing in yast and elsewhere.

  1. unmount, then stop the array
    umount /dev/md#
    mdadm --stop /dev/md# or mdadm --stop --scan

  2. Define & use "newname"
    mdadm --assemble /dev/md/newname --name=newname --update=name /dev/sd[xy]#
    edit /etc/mdadm.conf; change oldname to newname

  3. Make persistent across reboots
    dracut -f

System on RAID-0

This is somewhat faster, but forfeits grub-rollbacks (thumbs-down).

System (root) on a raid-0 volume seems to have exposed a limitation of the Dell optiplex 3010 UEFI firmware.
Corresponding efi and boot partitions are expected to be on the same physical disk.

If the root is on soft-raid0, then /boot may need to be on a separate partition, because of firmware limitations similar to that experienced above.
EFI and /boot cannot share a partition because EFI needs to have a 'FAT' filesystem, and '/boot' needs a posix compliant filesystem (which 'FAT' is NOT).

Config used:

partition sdX sdY raid-0
1 260M [FAT] /boot/efi 260M -
3 500M [XFS] /boot 500M -
4 20G -> 20G -> [Btrfs] /
5 remainder -> remainder -> [XFS] /home
2 2G -> 2G -> [swap]

Add & grow an array