[OpenAFS] I'm my own worst enemy sometimes...

[Derrick J Brashear]

On Sat, 28 Jun 2003, Steven N. Hirsch wrote:


> Thanks for giving me a nudge in the right direction.  I wasn't seeing the
> forrest for the trees.  Now that I have thought about things, the
> semantics I was expecting were unreasonable.  Naturally, anything added or
> changed directly inside a replicated volume must occur on the
> /afs/.my_cell path.  Once I added an /afs/.my_cell/u volume (without
> replication) I could operate on it through /afs/my_cell/u/blah, blah.
>
> So, to make sure I understand, one should never replicate anything below
> /afs/my_cell ?

That's not necessarily true.

We replicate the volumes that contain user home directory mount points,
for instance:

/afs/andrew.cmu.edu/usr16 is cmu.usr16; shadow is a mountpoint for
user.shadow in that volume. Changes are sporadic and it means that only
people whose user volumes are offline lose, not all the people with mount
points there if one server with that volume goes down.

And, heck, that ignores the idea of having system binaries distributed in
afs, which will certainly be below /afs/my_cell

I think the rule is "don't replicate data considered generally read/write"

-D