Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 12 Apr 2012 17:46:57 +0200
From:      Fabian Keil <>
To:        "Christopher J. Ruwe" <>
Subject:   Re: zpool creation on geli failed with FreeBSD-9.0
Message-ID:  <>
In-Reply-To: <>
References:  <> <> <>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

"Christopher J. Ruwe" <> wrote:

> On Tue, 10 Apr 2012 08:13:36 +1000
> Da Rock <> wrote:
> > On 04/10/12 02:07, Christopher J. Ruwe wrote:
> > > I was trying to install FreeBSD 9.0 using a geli encrypted disk and
> > > ZFS on my ThinkPad R500 this weekend. I failed.
> > >
> > > Having sucessfully initialized the geli part and having attached the
> > > provider, my attempt to create a zpool on the geli section thus
> > >
> > > $>  zpool create ntank /dev/ada0p2.eli
> > >
> > > failed with the message
> > >
> > > Cannot create 'ntank': invalid argument for this pool operation.
> > >
> > > I could not convince the system to create the zpool on the geli
> > > part, so I gave up and created the zpool on the unencrypted
> > > partition instead to have a working machine for the week. I would,
> > > however, like to have my data on an encrypted partition though. Has
> > > anyone witnessed and resolved this issue or does anyone have other
> > > ideas?

I have never seen this error message and its meaning isn't
obvious to me either (which is why I didn't comment earlier,
before Da Rock "provoked" me).

You could try truss'ing the operating to see what is actually
going on.

You could also try to see if it makes a difference if you
put a label on top of the eli geom and try to create the
pool on that.

> > Someone using ZFS will be able to verify this, but from my
> > understanding ZFS runs on the hardware and you can *possibly* put
> > geli on top of ZFS.

While you can indeed put geli on a ZFS volume, putting the
ZFS pool on geli is supposed to work as well and does work
for me:

fk@r500 ~ $zpool status tank
  pool: tank
 state: ONLINE
  scan: scrub repaired 0 in 3h1m with 0 errors on Sun Mar 25 20:47:28 2012

        NAME           STATE     READ WRITE CKSUM
        tank           ONLINE       0     0     0
          ada0s1d.eli  ONLINE       0     0     0

errors: No known data errors

The pool was created with a command rather similar to the one
used by Christopher:

fk@r500 ~ $sudo zpool history | head -n 2
History for 'tank':
2009-08-09.02:53:52 zpool create tank ad4s1d.eli

While I'm using this pool on a 10.0-CURRENT system, the pool worked
with previous FreeBSD versions in the past. I currently have no
9.0 system to test with.

Here's a more recently created pool where the eli is on a label:

fk@r500 ~ $zpool status lexmark
  pool: lexmark
 state: ONLINE
  scan: scrub repaired 0 in 0h2m with 0 errors on Sun Mar 11 22:10:49 2012

        NAME                 STATE     READ WRITE CKSUM
        lexmark              ONLINE       0     0     0
          label/lexmark.eli  ONLINE       0     0     0

errors: No known data errors
fk@r500 ~ $sudo zpool history lexmark | head -n 2
History for 'lexmark':
2012-02-04.13:41:05 zpool create lexmark /dev/label/lexmark.eli

This is the configuration I usually use for external
devices, as it makes automatically importing the pool

While this configuration is known to trigger bugs under
certain conditions, I don't know for sure that using less
geoms would actually help:

> > You can put geom on ZFS but not the other way around.

Except for test pools, I always use geoms as vdevs.
This has been working (for me) since I started using
ZFS shortly after Pawel imported it.

> I had a zpool on a geli on FreeBSD-8.0 to 8.2 and
> 8-stable. I opted for a complete reinstall of 9.0 instead of upgrading
> due to issues with xmonad, which I did not understand then. I am
> "furious" at my decision now that I have seen that a update would
> possibly have been much easier than to reinstall, had I at that time
> understood what was causing problems with xmonad.

Do you know if you used the same zpool command and the same
vdev in the past?


Content-Type: application/pgp-signature; name=signature.asc
Content-Disposition: attachment; filename=signature.asc

Version: GnuPG v2.0.19 (FreeBSD)



Want to link to this message? Use this URL: <>