Mon, 16 Nov 2020 10:29:17 +0000 Tony Mountifield tony@softins.co.uk:
In article 20201115123245.db62b8248e1f248afe02844a@lukaszposadowski.pl, Lukasz Posadowski mail@lukaszposadowski.pl wrote:
Hello everyone.
I'm trying to install CentOS 8 with root and swap partitions on software raid. The plan is:
- create md0 raid level 1 with 2 hard drives: /dev/sda and /dev/sdb,
using Linux Rscue CD,
- install CentOS 8 with Virtual Box on my laptop,
- rsync CentOS 8 root partition on /dev/md0p1,
- chroot in CentOS 8 root partition,
- configure /etc/mdadm.conf, grub.cfg, initramfs, install
bootloader on both sda and sdb drives.
I think I can do first four of the above, but my CentOS installation acts strange after rebooting the server. It recognizes the raid, but boots randomly with root on /dev/sda1 (and recognizes raid with /dev/sdb disk), or with root on /dev/sdb1 (and recognizes raid with /dev/sda disk). When booting from Linux Rescue CD, the raid with two disk is recognized.
I thought it was much more usual to partition both disks to give sda1,2,3 and sdb1,2,3, and then create /dev/md0 from sda1/sdb1, /dev/md1 from sda2/sdb3, and so on.
That's the way I have always done it, and have never had any problems. Never seen an attempt to partition an md device before. In that case, how would the kernel and initrd be found in order to assemble the RAID?
It totaly works either way - raid on partitions and raid on disks. I'm use to mirroring whole disks.