[solved] mdadm "not an md array" using 3tb disks.

[chenko]

When I try and create a raid array, I get: "mdadm: device /dev/sdf1 exists but is not an md array."

This is on a machine that I previously had a bunch of 750gb drives working, so all is okay on that side.

Difference now is im trying to replace them with 3tb drives. The machine supports it, I can partition, format and mount the disks induvidually.


Normally I would go into fdisk and setup a partition with type fd for raid auto detect. But this time since they are over 3tb I am using parted, mklabel to gpt, and setting the "raid" flag. The partition is aligned (as in, I don't get a warning if I use parted -a optimal).

But I get the above error when using --create, I also get this error if I try and create an array using the devices themselves without any partitions so im not sure if this is a partion/parted problem.

I've tried metadata=1.2, although I think its default anyway for mdadm version I have (v3.1.4 - 31st August 2010), also tried setting --chunk=4096 etc (and some other variants).


Here is parted -l for one of the drives, the others are similar.

[py-ro]

raid Flag ist not used, set type FD00 (or 00FD?)

[chenko]

I'm not really sure which command is which in parted - any that refers to flags doesn't like FD00 (or 00FD).

Just merged gdisk to use instead, set the fd00 flag, if I look at the disk in parted it shows "raid" under flags just like I did myself using "set 1 raid on".

Used gdisk, set fd00 on the drives and I still get the same error.

[py-ro]

Thats simpe you forgot to mention the Device you want to create, write /dev/md0 before the component devices.

Bye
Py

[chenko]

Okay, that was a bit silly. Although when I first started doing this (before I realised to change to gpt) I was putting that info in and getting another issue which I have since sorted.

Thanks