Device /dev/md0 already has a non-bcache superblock, remove it using wipefs and wipefs -a

[Ante Karamatić]

When configuring a machine in MAAS to use bcache for /var, curtin fails because it gets an unexpected exit from make-bcache. I suspect curtin doesn't handle dirty devices? While similar to bug 1750519, these are not the same.

I've attached curtin user-data.

[Ryan Harper (raharper)]

On Tue, Feb 20, 2018 at 8:37 AM, Ante Karamatić <
<email address hidden>> wrote:

> Public bug reported:
>
> When configuring a machine in MAAS to use bcache for /var, curtin fails
> because it gets an unexpected exit from make-bcache. I suspect curtin
> doesn't handle dirty devices? While similar to bug 1750519, these are
> not the same.
>

It's not the same, but it's nothing to do with curtin itself. The error
message
comes from bcache itself.

make-bcache.c:204

 if (!blkid_do_probe(pr)) {
        /* XXX wipefs doesn't know how to remove partition tables */
        fprintf(stderr, "Device %s already has a non-bcache superblock, "
                "remove it using wipefs and wipefs -a\n", dev);
        exit(EXIT_FAILURE);
    }

That looks to be a unsophisticated check that the backing device may
already be
claimed by some other filesystem or partition (which it is, via md).

I'll look to see if this configuration is supported upstream.

>
> I've attached curtin user-data.
>
> ** Affects: curtin
> Importance: Undecided
> Status: New
>
>
> ** Tags: cpe-onsite
>
> ** Attachment added: "curtin.txt"
> https://bugs.launchpad.net/bugs/1750580/+attachment/
> 5058939/+files/curtin.txt
>
> --
> You received this bug notification because you are subscribed to curtin.
> Matching subscriptions: curtin-bugs-all
> https://bugs.launchpad.net/bugs/1750580
>
> Title:
> Device /dev/md0 already has a non-bcache superblock, remove it using
> wipefs and wipefs -a
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/curtin/+bug/1750580/+subscriptions
>

[Ryan Harper (raharper)]

I've marked this as duplicate as it's a manifestation of the same issue in curtin's clear-holders code which didn't wipe the raid device before breaking it apart.