My backup process uses rsync to duplicate my backup onto hot-swapped hard drives. The SATA hot-swap unit is wired to SATA1 on the mobo. My script mounts /dev/sdb1 does the rsync, then umounts the disk. The disks used vary in size but they’re all ext3 file systems and it works pretty well most of the time. Every month or two I get a mount failure on the swapped drive. I use fdisk -l and discover that the drive is now seen as /dev/sdc1. I restart the machine and the disk appears as /dev/sdb1 again but it’s read-only. Disgusted, I format the disk and run the backup copy (which works fine). When next disk is replaced the system doesn’t see a disk at all. I restart the machine and everything hums along again until the next incident. CentOS 5.7.

Should I be doing more than umount prior to changing the drives? Having the disk come up /dev/sdc1 challenges my understanding of how these device names get assigned. Under what circumstance does the the second disk on a machine come up as /dev/sdc? It seems complicated to mount by UUID or Label (since my swap person sometimes forgets a disk, or is sick).

Since the UUID uniquely identifies each device (and partition on a drive), you could try writing UDEV rules to mount each UUID at /dev/sdb1, assuming each backup drive has only one partition.

Bear in mind that if you try to attach two or more backup drives as the same time, you will have a conflict, with only the first drive mounted and the others generating error messages.