I'm running Ubuntu 16.04, and I'm having a problem with a new RAID 10 array. The array consists of four 5 TB drives, all the same exact model. Two of the drives were used by me previously, and two are new. I made sure that they synced cleanly and wiped out old data and partitions. The array forms properly afterwards, with a capacity of 9.1 TB. However, after I reboot, the array degrades to a RAID 0 with a capacity of 13 TB. I've verified that I've got the correct UUID in /etc/mdadm/mdadm.conf, but the array always boots up one disk short. It's always the same device as well: /dev/sdd1 I'm able to force the array to start with this command: sudo mdadm --create /dev/md0 -v --assume-clean --level=raid10 --raid-devices=4 /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1 The array acts normally and can be mounted after I make a filesystem with fdisk. It just degrades to RAID 0 again after I reboot. What am I doing wrong here? It seems like one disk is failing to be found at boot, and thus is missing from the array. Do the disks require something which I'm missing for this one disk? Should each disk in the array be specified in my fstab file, or should they be excluded from it? Any insight would be greatly appreciated.
1 Answer
Sounds like one of the drives is slow to start up and be ready. That may either be because it's simply slow to spin up, or because something is wrong with it.
Check the drive's SMART log (gnome-disk-tool can access that for you), and if the drive is healthy, increase the grub boot timer a bit to give the drive time to spin up before the OS boots.
-
Ok, I thought that it could be something like that. Do you know if there's a way to specifiy that GRUB has to wait until all disks are ready, or do I just make it wait like 60 seconds?– some_guyCommented Apr 10, 2017 at 14:07
-
You just set an arbitrary
GRUB_TIMEOUT=that's long enough. 30 seconds should be plenty. 60 would be a bit overkill.– Mio RinCommented Apr 10, 2017 at 14:13