[OpenAFS] Invalid cross-device link

[Stefan Strandberg]

It was determined out of list that the particular error I'm seeing in
this case is because I was adding a RO volume on the same server but
different partition as the RW volume.  While I know it's terrible
practice, it did work in previous versions and I was using it for
testing purposes.  Apparently, this is no longer allowed, and gives the
errors that I'm seeing.

So, that mystery is solved - however, why I got stuck in salvage loops
with 1.4.10 is not, and as I don't have logs and am wary to bring
production machines back to 1.4.10, it'll remain a mystery for the
forseeable future I imagine.

Thanks for the help!

-stefan

On Thu, Aug 20, 2009 at 09:19:01AM -0500, Stefan Strandberg wrote:
> Hi,
>=20
> 1.4.11 isn't really doable until it's at least in lenny-backports as we
> don't want to roll our own versions of this.
>=20
> As for a stale replica existing, I may be misunderstanding.  If you're
> saying that it's a replica of that volume, I don't see how that's the
> case.
>=20
> Here's the creation and subsequent release attempt for a brand new
> volume:
>=20
> stefan@cog ~ $ vos create beth a foo.bar
> Volume 536885610 created on partition /vicepa of beth
> stefan@cog ~ $ vos addsite beth b foo.bar
> Added replication site beth /vicepb for volume foo.bar
> stefan@cog ~ $ vos rel -v foo.bar
>=20
> foo.bar=20
>     RWrite: 536885610=20
>     number of sites -> 2
>        server beth.cae.wisc.edu partition /vicepa RW Site=20
>        server beth.cae.wisc.edu partition /vicepb RO Site  -- Not relea=
sed
> This is a complete release of volume 536885610
> Cloning RW volume 536885610 to temporary RO... done
> Getting status of RW volume 536885610... done
> Ending cloning transaction on RW volume 536885610... done
> Starting transaction on cloned volume 536885611... done
> Creating new volume 536885611 on replication site beth.cae.wisc.edu: Fa=
iled to create the ro volume: : Input/output error
> The volume 536885610 could not be released to the following 1 sites:
>                           beth.cae.wisc.edu /vicepb
> VOLSER: release could not be completed
> Error in vos release command.
> VOLSER: release could not be completed
>=20
> And here's the VolserLog output:
>=20
> Thu Aug 20 09:15:14 2009 1 Volser: CreateVolume: volume 536885610 (foo.=
bar) created
> Thu Aug 20 09:15:24 2009 1 Volser: Clone: Cloning volume 536885610 to n=
ew volume 536885611
> Thu Aug 20 09:15:24 2009 VAttachVolume: Failed to open /vicepb/V0536885=
611.vol (errno 2)
> Thu Aug 20 09:15:24 2009 1 Volser: CreateVolume: Unable to create the v=
olume; aborted, error code 18
> Thu Aug 20 09:15:24 2009 : Invalid cross-device link
>=20
> Turning up debugging doesn't show any extra anything really.
>=20
> Thanks again,
>=20
> -stefan
>=20
> On Thu, Aug 20, 2009 at 09:42:46AM -0400, Derrick Brashear wrote:
> > On Thu, Aug 20, 2009 at 9:39 AM, Jeffrey
> > Altman<jaltman@secure-endpoints.com> wrote:
> > > Stefan Strandberg wrote:
> > >> Anyone have any ideas? =A0I would really like to get everything on=
 1.4.10
> > >> for the performance increases.
> > >
> > > The current version of OpenAFS is 1.4.11 which addresses:
> > >
> > > - Fix race in background sync code which could cause volumes to go
> > > =A0offline. (124359)
> > >
> > > This is not the issue you are describing. =A0However, please test w=
ith
> > > 1.4.11 and see if the problem is still present. =A0 If so, send log=
s and
> > > report to openafs-bugs@openafs.org.
> >=20
> >=20
> > it will still be present. the real problem is you have a stale copy o=
f
> > the replica elsewhere on the disk. there should be exactly one copy o=
f
> > 536885604, and it should be on the same partition as 536885602, both
> > according to the vldb and in vos listvol output. arrange to make that
> > true, and your issue will go away.
> > _______________________________________________
> > OpenAFS-info mailing list
> > OpenAFS-info@openafs.org
> > https://lists.openafs.org/mailman/listinfo/openafs-info
> >=20
>=20
> --=20
> Stefan Strandberg
> UNIX group
> Computer Aided Engineering - UW Madison
> stefan@cae.wisc.edu
>=20
>=20
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info
>=20

--=20
Stefan Strandberg
UNIX group
Computer Aided Engineering - UW Madison
stefan@cae.wisc.edu