I'm facing a weird situation on one of my server on the provisioning process.
Here is the machine HW Model: HPE Proliant DL365 Gen 10 Plus with 6x 1.9TB SATA 6GBps SSD disks.
What's going on: When I'm provisioning this machine using CentOS Stream 8 kickstart, the installer consistently install the OS on the first disk of the server as SDA.
When I'm provisioning this machine using CentOS Stream 9 kickstart however, the installer isn't able to install the OS on the first disk of the server as SDA, it always choose another disk name.
Kickstart paste: https://paste.opendev.org/show/bjeFoS3svA3dILjMEP8n/
Additional misc informations:
Any server of this type that continue to install correctly on CentOS Stream 9 is having all its disks using the same product ID (HPG3/HPG1 let say).
The disk that I'm targeting as SDA on this specific server is the only one having a product ID/reference different from the others and it looks like it is slower to activate.
On this server, when I do trick it and manage to install the OS on this specific disk, at the next reboot (correctly working), it then do not use SDA as a name for the disk.
On this server, when I do trick it and manage to install the OS forcing the installer to use the SDA naming instead of forcing it to use a path, then during the installation the selected disk is one of the other 6th (expected) and SDA naming is assigned to this disk, then when the installer reboot, udev assign this disk another naming sdf/etc and this can change at each reboot or not.
So all in all, I think I've gave you all information I've got.
If anyone have a clue, on how to force the installer and afterward the system (probably udev) to correctly wait for all disks to be up and available before assigning them a name and root pivot on it, I'm all in!
Thanks!
