Need Help Understanding

[etpoole60]

Let say that my installation of 1708 went well and then the update to 1810 went well also.
I'm stuck on several issues:

When I 'cat /etc/fstab" I see "UUID=099d7278-4cb9-434a-b933-c684434d6365 /boot ext defaults 1 2"
But when I enter 'df -h' I see " /dev/md126 2.0G 270M 1.6G 15% /boot"
When I comment out the UUID line and add the /dev/md126 I get an error - sometimes (requires a boot to rescue go back to the UUID).

When I installed CentOS 6 it allowed me to control the name of the MD devices for both /boot and the LVM raid devices. Why not with CentOS 7?

Why is it when I install CentOS 7 over an existing CentOS 6 file systems (I want to keep the allocations the same) CentOS 7 really doesn't like it??? I found myself deleting and reallocating the same partitions. Especially the /boot raid-1 allocations - it always comes up as 'unknown' .

I guess I'm old school because I like the flow of CentOS 6 as-far-as disks are concerned (i..e. allocate raid device from first disk; allocate raid device from second disk; create raid partition by adding the 2 raid devices; label it as /dev/md0 with a mount point of /boot; etc). CentOS 7 just seems odd they way it's done.

Can someone help me understand what I'm doing?

TIA
Gene

[hunter86_bg]

I don't see any issues with UUID .
Most probably , in initramfs there are some naming issues with MD devices (most probably your MD Is named different) and later when the switch root comes - everything goes to it's place.

Out can try to create an /etc/mdadm.conf and run dracut to rebuild your initramfs.
Once done, check the content of the initramfs to verify the config is inside. Then, give it a try

[etpoole60]

I'm guessing that the 'unknown' issue is due to the difference in the metadata for /boot raid-1 device. That just tells me that there are some 'backwards compatibility' issues.

After install and I enter 'cat /proc/mdstat' the names of the MD devices are /dev/md124, /dev/md125, etc which under CentOS 6 aren't allowed (I can only name them /dev/md0 through /dev/md15 - which seems logical to me - BTW, IBM started the naming of things from 0). So, Why can't I name the MD devices as in CentOS 6? Why take a (under CentOS 6) default name? Should I forget everything (in this area) I learned in the past? Was the kernel jump from 2.6 to 3.10 so big that they eliminated so much?

I googled looking for a CentOS 7 Installation Guide with no luck so I started checking out the RHEL 7 installation guide but they didn;t spend much time on Software RAID. Has any one seen a good guide or tutorial on this? Where I'm lost is on the screem where you define the mount points and stuff, is 2 fields ('label' and 'name') I 'm looking for a explanation on how to use them. OK, I guessed and started using the 'name' field for the LV name (i.e. /home would be lv_home because I use LVM). But, shouldn't that mean I could use the 'label' field to put, for example, /dev/md0 for the /boot raid-1, etc? If this is important, should I just define everything singular and after the install is complete using sfdisk and mdadm to create the RAID devices manually?

Just don't understand.
Gene

[TrevorH]

What's the matter with using UUIDs? They are unique and are the preferred method to mount things.

[etpoole60]

I've no problem using the UUID as that's the way CentOS 7 is running now - but just for the /boot raid-1 device. I just like things consistent. If something goes wrong with the hard drive(s) that the /boot raid device is on, by looking at /dev/md0 with cat /proc/mdstat I know exactly what disk drive is causing me issues. It become more difficult when you have several raid-1 devices.
On CentOS 6 the /boot raid-1 device would always be /dev/md0.

It just seems things have been made different when it wasn't necessary.

Gene

[etpoole60]

OK, So far I think I've solved all of my hard drive issues!
What I did was to reinstall CentOS 6 because I can define everything just like I want it to be.
The hard drive environment is:
2 - 120 GB WD SATA Drives
4 - 4 TB WD SATA Drives
On the 120 GB drives I've defined the following RAID-1 devices:
/dev/md0 - /dev/sda1 | /dev/sdb1 - /boot - 2 GB
/dev/md1 - /dev/sda2 | /dev/sdb2 - /dev/vg_osswap/ 32 GB - lv_sysswap 16 GB - swap
/dev/md2 - /dev/sda3 | /dev/sdb3 - /dev/vg_opsys/ 78 GB - lv_root 8 GB - /name I entered lv_sysswap
- lv_archive 2 GB - /archive
- lv_home 4 GB - /home
- lv_opt 2 GB - /opt
- lv_tmp 5 GB - /tmp
- lv_usr - 12 GB - /usr
- lv_usr_local 2 GB - /usr/local
- lv_var 8 GB - /var

On the 4 - 4 TB dribes I've defined the following RAID-10 device:
/dev/md3 - /dev/sd(c-f)1 - /dev/vg_sysdata 7.28 TB - Many lv_ devices.

The definitions for /dev/md0; /dev/md1; /dev/md2 are built from a custom kickstart DVD that was built from RHEL 4 and constantly modified since then so it works with CentOS 6. However, I do not have custion install DVD for CentOS 7 so I have to do the install manually.

When I did the CentOS 7 1708 install and I got to the 'Custom Partition' page I did the following:
The allocation for /boot shows up as 'unknown' so I entered the mount point as /boot; I selected reformat as ext4; it did show up as raid 1; in the label I entered /dev/md0; in the name I entered boot.
The allocation for swap shows up as CentOS 6.10 so there was no mount point; reformated as swap; in the label I entered /dev/md1; in the name I entered lv_sysswap.
For all of the other defined OS file systems I added the mount points; reformatted as ext4; they showed up as raid-1 but in the label I entered /dev/md2 for all of them; in the name I entered the lv name as shown above.
What was strange was that when I did the raid-10 file systems there was information in the label and name fields already.

So going forward I will continue to use the 1708 DVD after I verify that all of the allocations are basically the same across board.

I think I understand as much as I need to at this point.

Gene

[etpoole60]

How do I mark this post as solved?