I would recommend you try the compile from latest source code:

https://git.kernel.org/pub/scm/utils/mdadm/mdadm.git/

There are some changes in the udev rules, which may help here.

VMware should not be a problem.

Thanks,
Song

On Wed, Nov 15, 2017 at 11:31 PM, Tobx [off-list ref] wrote:

Larkin Lowrey wrote:

quoted

The "spares=1" is for the journal device. The "mdadm --examine --scan" command produced that for me as well but since it seemed odd I deleted that part. Since there isn't actually a spare it seems incorrect. I've never tried to boot with "spares=1" so I don't know if removing it will make any difference.

Yes, I tried that, seemed odd to me too, I also tried just:
ARRAY /dev/md/test UUID=d59de5cc:02f560ed:ebcb9400:491955d7

Rudy Zijlstra wrote:

quoted

Did you update the initrd after creating the array?

Yes.

quoted

Tobx wrote:

quoted

quoted

I tried this with Debian Stretch (4.9.0-4), Debian Stretch (4.13.0-0) and Ubuntu 17.10 (4.13.0-16).

quoted

Tobx wrote:

quoted

I compiled mdadm from source and tried again with:

# mdadm --version
mdadm - v4.0 - 2017-01-09

Song Liu wrote:

quoted

Your mdadm is a little old. Could you please try rebuild it from
latest source code?

Do you have something newer?

It always works without the journal option (except I set something very wrong of course) and also with the option, but then it never assembles automatically after reboot.

I am testing everything in VMware mostly on a clean Debian installation. Might VMware be the problem? I have no other test system in place.

Cheers,
Tobi