Done on Centos 7.3 very important as clearly based on older guides it was a lot easier and more simpler! Hint do not use grub2-install!
One huge caveat if you are an oldschool user or sysadmin who has avoided UEFI booting
The normal way will not work here if your Centos was using UEFI. Newer systems use it by default.
The easiest way to check is to do an fdisk -l if your system disk and check for an EFI partition.
If using EFI you need a EFI partition usually about 200M.
Instead of the typical partition 1 being normal boot this will now be EFI.
To match you will now have of course 4 mdadm arrays instead of 3 because of EFI.
Centos boot issue
#1 you MUST have an EFI partiton and you MUST be booted as UEFI!
So to make it more complicated with UEFI here is how a standard or at least Centos 7 typically works.
/boot partion mounted as /boot
/efi partition mounted as /boot/efi
b.) EFI partition can be mdadm RAID but must be level 1 and --metadata=1.0 and vfat filesystem
grub2-install --target=x86_64-efi /dev/sda
grub2-mkconfig -o /boot/efi/EFI/redhat/grub.cfg
If you don't tell grub2-mkconfig to update the EFI grub.cfg you will be greeted to a blank grub boot screen with no menu entries. No clue why grub2-mkconfig does not seem to update the correct EFI grub.cfg on its own.
#after booting getting another issue
dracut-initqueue: Warning: dracut-initqueue timeout -starting timeout scripts
Started dracut initqueue hook
Reached target Remote File Systems (Pre).
Reached target Remote File Systems.
A start job is running for dev-disk-by\x2uuid-8f411f86\x2d....device (7min 58s / no limit)
md0 = /boot ext4
*(note md0 this must be --level 1 and --metadata=0.90 or it will not boot/work)
md1 = swap
md2 = / ext4
all mounted in /mnt/
#install grub2-tools so we can install it
yum -y install grub2-tools grub2-efi-modules
lack of efi modules gives the following error (if you didn't install the above): grub2-install: error: /usr/lib/grub/x86_64-efi/modinfo.sh doesn't exist. Please specify --target or --directory.
Setting up swapspace version 1, size = 31440892 KiB
no label, UUID=a30ab8dd-f462-4dab-940d-6658bcae4395
get the UUIDs of each array:
/dev/md0: UUID="5b51dbfb-f361-48c3-bfdf-5da171e4577c" TYPE="ext4"
/dev/md1: UUID="a30ab8dd-f462-4dab-940d-6658bcae4395" TYPE="swap"
/dev/md2: UUID="8f411f86-3816-409b-b291-ebe1a7a170ef" TYPE="ext4"
mount /dev/md2 /target
rsync -Pha --exclude="/boot/*" --exclude="/sys/*" --exclude=/proc/* --exclude=/mnt/* / /target
mount /dev/md0 /target/boot
#if you have an EFI partition /dev/md3 or whatever you've used it for
mount /dev/md3 /target/boot/efi
#rsync the rest
rsync -Pha /boot/ /target/boot/
#if you have an EFI partition
rsync -Pha /boot/efi /target/boot/efi
Change /mnt/md2 to your mounted root /
for bind in dev sys proc; do
mount -o bind /$bind /$mount/$bind
#run the swapon command on your swap partition othewrise it soudns weird but mdadm WILL not activate the array even if it's in mdadm.conf.
mdadm --detail --scan > /etc/mdadm.conf Update initramfs #you must specify the exact initramfs image and exact kernel name (matching it) or it will not work/be unbootable - much more finicky than Centos 6 dracut /boot/initramfs-3.10.0-514.21.2.el7.x86_64.img 3.10.0-514.21.2.el7.x86_64 --force #if you get a message like this you specified the wrong kernel details in the second parameter Kernel version 3.10.0-514.21.2.el7 has no module directory /lib/modules/3.10.0-514.21.2.el7 #in the above case I forgot the .x86_64 at the end
dracut --mdadmconf --fstab --add="mdraid" --filesystems "vfat xfs ext4 ext3 tmpfs devpts sysfs proc" --add-drivers="raid1 raid10" --force #this dracut step is critical because by default it will not have mdadm raid support so it will keep looking for devcies that don't get created #also MAKE sure to use the latest or current initramfs (may not be obvious ifyou are booting from a recovery disc). dracut --mdadmconf --fstab --add="mdraid" --filesystems "xfs ext4 ext3 tmpfs devpts sysfs proc" --add-drivers="raid1 raid10" /boot/initramfs-3.10.0-514.21.2.el7.x86_64.img --force
#to avoid problems remove any old grub.cfg
echo y| rm /boot/efi/EFI/centos/grub.cfg
echo y| rm /boot/grub2/grub.cfg
#something went wrong as it doesn't boot
The above happens if you are running EFI and didn't specify output to your EFI partition for grub.cfg
If you have EFI
grub2-mkconfig -o /boot/efi/EFI/centos/grub.cfg
#not sure why but the above still results in a blank grub screen with no menu entries. Maybe grub2 needs to be reinstalled as a package or something else first. I suspect it doesn't contain the right modules to read the boot or efi partition.
#NO NO NO - DO NOT USE GRUB2-INSTALL
This Fedora warning explains it all:
grub2-install shouldn't be used on EFI systems. The grub2-efi package installs a prebaked grubx64.efi on the EFI System partition, which looks for grub.cfg on the ESP in /EFI/fedora/ whereas the grub2-install command creates a custom grubx64.efi, deletes the original installed one, and looks for grub.cfg in /boot/grub2/.
One other solution is if you really want grub2-install is to make /boot/grub2/grub.cfg a symlnk to /boot/efi/EFI/centos/grub.cfg ln -s /boot/efi/EFI/centos/grub.cfg /boot/grub2/grub.cfg this doesn't work I am certain grub cannot make use of that symlink
#if you get an error of "grub2-install: error /boot/efi doesn't look like an EFI partition"
It means you haven't mounted your EFI partition OR you haven't set the partition as EFI (in gdisk this is type ef00). You'll have to stop the mdadm array for EFI and then do a mkfs.vfat on it and try again (partition must be VFAT and no other fs such as ext4/xfs etc.. it won't work unless it's VFAT).