Hi folks,

Yesterday I created a raid5 array /dev/md0 consisting of 5 harddisks, named sda thru sde on the time of creation.
After that I stored some data into the arry without any difficulties, then shutdown the computer.
Early this morning when starting the computer I got a message that /dev/md0 was not ready to be mounted.
So I checked the raid array and discovered that the enumerator had been messing with the harddisks.
Harddisk sda was now sdc etc. etc.
After I rebooted, the harddisks got the original names again: sda was sda again.
When I mounted the array no problems occurred.
So, it seems that the order in which the harddisks are enumerated influences the availability of the raid array.
Is there a way to avoid this kind of problems with a raid array?

Did you create the /etc/mdadm.conf?

The UUIDs will be used instead of device names, so how they are enumerated is immaterial.

1 members found this post helpful.

I'm curious about this, how would you assign a UUID to a raw disk such as sda? Although mdadm will handle raw disks, the issue of device names changing seems to require that you create 'fd' partitions on those disks just to get a UUID to work with. Another alternative I understand is to create custom udev rules to assign permanent names to the devices.
Is there a better way?

The md superblock will contain a UUID. In addition, each filesystem superblock contains a UUID. The /etc/fstab and mount command also accept UUIDs.

macemoneta, thank you very much. Your advice was very helpful!