We are using a SuperMicro computer that has 24 removable hot-swappable disks. During use, disks will independently be swapped in and out.

For this to work, we need to know which physical slot a disk gets inserted into. There are many many disks, so tracking a disk id of the disk itself is not really feasible. And the disks are a mix of ext4 and ntfs.

We have done this before with SuperMicro computers that had, say, 4 such drive bays. We have udev rules that track activity on adds to the block subsystem

The detection and all works fine. The problem we have is that detecting which physical slot the disk has been inserted into is not consistent. We use the /dev/disk/by-path entries for this. As we have done in the past.

This hardware information must, I guess, be assigned by the SuperMicro BIOS (or the modern equivalent) each time a disk is inserted. It really looks like the hardware info changes. There are 24 hardware values that seem to be used (this is consistent), but the physical slot to which they refer seems somewhat random.

Does it make sense to expect the hardware information in /dev/disk/by-path to consistently refer to the same physical SCSI interface? I know that at the end of the day it all comes down to what the BIOS tells when a disk is added. But I want to make sure that I am not doing something silly in the Linux part of this. Like I should be looking elsewhere, or running some obscure command to get this consistent hardware address.

-- Roger Oberholtzer