[OpenAFS] [ OpenAFS - Cache FileSystem ]

Marc Dionne marc.c.dionne@gmail.com
Wed, 2 Oct 2013 18:18:05 -0400

On Wed, Oct 2, 2013 at 3:48 PM, nicolas prochazka
<prochazka.nicolas@gmail.com> wrote:
> Hello,
> For your intention,
> In Faq, we can read :
> The OpenAFS cache manager will detect an unsupported filesystem and
> refuse to start.
> The following file systems have been reported to work for the AFS client cache:
> ext2
> ext3
> hfs (HP-UX)
> xfs (at least on IRIX 6.5)
> ufs (Solaris, ?Tru64Unix)

That information is somewhat outdated; for more recent OpenAFS
releases and Linux kernels there are no runtime checks, and any
filesystem that is exportable by NFS has a good chance of being usable
for holding the client cache.  So the list of possible filesystems is
substantially larger than the above.

> But if I configure cache on zfs on linux ( zfsonlinux.org) ,
> i got kernel panic :
> [114328.841466] Starting AFS cache scan...
> [114349.618208] openafs: Inconsistent file handles within cache

I haven't looked at zfs code closely, but that message suggests that
- zfs does provide the necessary exportfs API calls, so should be
usable in theory
- maybe your cache location crosses a mount point between zfs and some
other filesystem
- or, zfs produces file handles of variable length.  The current
client code assumes that handles have a constant size, so if this is
the case, some changes will be required before zfs can be used to hold
the AFS disk cache.