[OpenAFS] "No Such Device" error on newly created and mounted volume

Jeffrey Altman jaltman@secure-endpoints.com
Wed, 29 Jul 2009 19:47:19 -0400

J. Maynard Gelinas wrote:
> We're in the process of transitioning from OpenAFS 1.2.11 to 1.4.7. Both

rxdebug -ver says your 1.2 servers are 1.2.10.  Not that this is
particularly relevant.

1.4.7 is quite old.  1.4.11 is current.  There are certainly OpenAFS
packages available for Debian.

> fs3:/afs/.lns.mit.edu/user# vos listvol afs3 vicepj
> Total number of volumes on server afs3 partition /vicepj: 66
> test                              536876681 RW          2 K On-line
> [...]
> afs3:/afs/.lns.mit.edu/public# fs mkmount test test
> afs3:/afs/.lns.mit.edu/public# ls test
> ls: cannot access test: No such device
> afs3:/afs/.lns.mit.edu/public#

Sure because:

[\\afs\.lns.mit.edu\public]vos exa test -cell lns.mit.edu
VLDB: no such entry

there is no volume named "test" in your vldb.  Therefore, the client
cannot find it.

> afsdbserv1:/var/lib/openafs/db# udebug afs2 7002
> Host's addresses are: ***.***.***.134

There is no benefit to hiding the addresses of you VLDB servers since
they are being published in your DNS AFSDB records.

> Is my problem the difference between openafs 1.2.11 vs. 1.4.7 or do I
> have a deeper problem going on here? There are no clock skew issues
> between any of the servers.

The vldb is out of sync with the servers.  I would start by upgrading
your 1.4.7 servers to 1.4.11 and then synchronize the contents of the
vldb with the list of volumes on each of your file servers.

I would avoid running with a mix of 1.2.x and 1.4.x database servers.
Mixing 1.4.x database servers with 1.2.x file servers should be fine.

Jeffrey Altman