[OpenAFS] Re: AIX v3fshelper

Hartmut Reuter reuter@rzg.mpg.de
Fri, 03 May 2002 15:43:03 +0200


> So far, with every AIX maintenance update, we just copied the old
> Transarc AFS 3.6 fsck utility back over the AIX updated version and
> it worked. There are two problems though.
> 
> 1) IBM may change the filesystem code enough to break the Transarc
> fsck.
> 2) Transarc's fsck may not work with OpenAFS (now or in the future).
> 
> Does it mean OpenAFS on AIX is not or will not be a viable option?
> Any clarification would be appreciated.

With OpenAFS you may use the NAMEI-Interface also along with AIX or any
other UNIX. We are running MR-AFS servers on AIX, Solaris and Linux and
are using the NAMEI-Interface on all our servers. It may be a little bit
slower because of the additional directory lookups, but it has a lot of
advantages:

1) You don't need vfsck or v3fshelper.

2) All files are visible for the administrator and could be dumped or
backed up.

3) You can have any kind of filesystem for AIX also filesystems with
large file support or GPFS

4) last not least: the salvage of a single volume is much much faster
because all files belonging to a volume group are under a single
directory.

There are two things you have to keep in mind:

1) You cannot upgrade a running server to the new software because there
is no way to convert a classical AFS psrtition to an NAMEI-partition.
That means you have to move all volumes to the new server.

2) I am not sure someone has used the NAMEI-interface for other
platforms than Linux and Windows in OpenAFS. Before you move your
volumes to such a server do enaugh tests to make sure everything works
correctly. (MR-AFS has slightly different namei_ops.c and header-files
so I cannot guarantee for normal OpenAFS servers).

Hartmut
-----------------------------------------------------------------
Hartmut Reuter                           e-mail reuter@rzg.mpg.de
					   phone +49-89-3299-1328
RZG (Rechenzentrum Garching)               fax   +49-89-3299-1301 
Computing Center of the Max-Planck-Gesellschaft (MPG) and the
Institut fuer Plasmaphysik (IPP)
-----------------------------------------------------------------