[OpenAFS] Understanding questions backup volume

Jeffrey Hutzelman jhutz@cmu.edu
Fri, 10 Feb 2006 12:39:08 -0500


On Friday, February 10, 2006 01:46:31 PM +0200 Juha J=E4ykk=E4 =
<juolja@utu.fi>=20
wrote:

> Is there something conceptually wrong or very stupid here? Our use of the
> RO clones would basically be the same as that of constantly mounted =
backup
> volumes, the difference being that they'd be on a different site to keep
> them up if the corresponding RW volume fails.

I don't think there's anything that won't work.  But it's unusual enough=20
that you need to be careful that you, and whoever maintains this after you, =

fully understand how things are set up.  It _will_ confuse anyone who's not =

familiar with what you've done.



> Currently, we have all RO volumes mounted without leading dots and RW =
ones
> with leading dots. Sergio Gelato suggested that this may not be THE way =
to
> go, though. What's your opinion?

That is a widely-used convention, and it's appropriate for things like=20
system software, where accesses to the RW volumes are rare.  For things=20
like user directories, it probably makes sense to use RW mount points for=20
the "normal" path, and mount the RO volume somewhere else, like you would=20
for a backup volume.  Obviously mounting the RO volumes inside the RW ones=20
is not terribly helpful, but you could create a parallel tree of RO mount=20
points:

/afs/your.cell/user/jhutz -> %jhutz
/afs/your.cell/user-readonly/jhutz -> #jhutz.readonly

-- Jeff