mjhjmtu

We are thinking about to change all Linux fstab configuration to work with UUID instead the current configuration

Some of the disks are with non RAID and some of the disks are with RAID10

I searched in google and find complain about using UUID for RAID1 :

" Unfortunately you MUST NOT use UUID in /etc/fstab if you use software RAID1. Why? Because the RAID volume itself and the first element of the mirror will appear to have the same filesystem UUID. If the mirror breaks or for any other reason the md device isn't started at boot, the system will mount any random underlying disk instead, clobbering your mirror. Then you'll need a full resync. Bad juju."

So I just want to know if we can use UUID for RAID10 ?

and in which cases ( RAID configuration ) not to use UUID?

second - in few lines what are the benefit to use UUID ?

Answer to your second question: an UUID allows you to uniquely identify a device.

Devices are assigned as /dev/sda, /dev/sdb, etc. depending on the order the system discovers them. While the drive the system boots on is always the first, for the others their name assignment depends on the order of discovery and might change after a reboot.

Also, imagine you have drives /dev/sdc and /dev/sdd, and you physically remove the first drive; after reboot, what was known as /dev/sdd is now called /dev/sdc.

This makes identification of devices ambiguous. UUIDs avoid all ambiguity; as the UUID is stored in the superblock (for a block device), it pertains to the device itself.

Adding to @dr01 answer: about the RAID questions, you can also use UUID in RAID settings.

If using mdadm, the UUID will be stored in the local filesystem, if using hardware RAID, it will be presented as a virtual physical disk, which again will have a UUID.

I think this is only an issue in cases where the data on the RAID device is also identically present in the same position on the underlying devices (or at least some of them). In practice that means the RAID formats where the RAID superblock (metadata) is at the end of the partition.

Since the UUID is part of the filesystem^(*), the system must find a supported filesystem on the device first. Filesystems are identified by reading some set locations on the device, usually at the beginning, and looking for identifying signatures. If the same data is visible in the same position in a raw disk (like /dev/sda) and a raid device (/dev/md0) the same UUID can be found in two devices. Or more, if there are copies of the data on other devices, i.e. the other side of the mirror.
_{(* The UUIDs of GPT partitions are a different matter)}

The Linux software RAID system knows of two main superblock formats, the original (v. 0.90) format puts the superblock at the end, and the current (v. 1) has three subformats for different positions of the superblock. Since the superblock formats 1.1 and 1.2 put the superblock in the beginning, they should be safe to use. 0.9 and 1.0 put the superblock at the end of the device, so they are the ones you can get into issues. /proc/mdstat should show you the superblock format for each device.

The RAID wiki page also contains a warning about placing the superblock at the end of the device.

With RAID 0 or RAID 10 it's hard to read the filesystem through the underlying disks, since the data is striped. But the UUID's might still be detectable, so it's probably best to use the 1.2 format RAID superblock in
any case.

For ext2/3/4 filesystems, tune2fs -l $device might be able to show you the filesystem UUID if it's possible to find (blkid might find the RAID UUID instead.)

Note that I'm writing this based on my understanding of the documentation, I didn't explicitly test with a RAID superblock at the end.