I just discovered that two of my disks have exactly the same UUIDs.
A couple of years back, I bought these drives to install RAID on them,
but gave up on that. Now, I've decided to do "manual" RAID, but I'm
wondering if the fact that the two drives have the same UUID is causing
whoever it is who sets up /dev/disk (I'm still trying to find that
culprit) is croaking on two different devices with the same UUID.
Where is the UUID determined? I'd presumed that it was derived from
some characteristics of the drive, determined by the device controller,
but now I'm wondering if my initial RAID configuration set some
drive-internal variable to be identical?
And, how does one /*reset*/ it?
On 2020-03-19 07:08, n952162 wrote:
On 2020-03-19 00:12, Ian Zimmerman wrote:
On 2020-03-18 22:57, n952162 wrote:
Well, some new recognitions ...
It turns out that those /dev/disk subdirectories don't necessarily have
all the disk devices represented:
1. by-id/
2. by-partuuid/
3. by-path/
4. by-uuid/
There is also by-label, which you can reference from fstab like
LABEL=foobar /home ext4 defaults ...
If predictability and readability is the goal, I think using labels is
the best option, because you have complete control over them, unlike the
device IDs. For example:
LABEL=my-machine-home-part /home ext4 defaults ...
This doesn't solve your underlying timing problem, of course. Just apropos.
I had by-label/ in the back of my mind, thinking that was always
available as a fall-back, but then forgot about it because it's /not
there/! That list above I got with ls(1).
I wonder if whatever process it is that builds /dev/disk (/udev/?) is
getting aborted, because I can't imagine a reason that
/dev/disk/by-uuid/* -> /dev/sda* shouldn't be there.
