2007/416 NFSv4 Mirror-mounts

[Bill Sommerfeld]

On Wed, 2007-07-18 at 00:58 +0100, Calum Mackay wrote:
> Bill Sommerfeld wrote:
> > Is there a server-side change associated with this project or is it
> > entirely a client-side feature?
> 
> entirely client-side; apols if that wasn't clear.

> I also failed to note that mirror-mounts should work with any vendor's 
> NFSv4 server implementation.

thanks for the clarification.

> >> Recursive manual hierarchical unmounting
> >>
> >> If an NFS filesystem is manually unmounted, then any mirror-mounted
> >> filesystems contained within it will also be unmounted, if idle. If
> >> there is an active mirror-mounted filesystem within, the manual unmount
> >> will fail, as though that original filesystem were busy. A forced
> >> unmount will, however, be propagated through to all enclosed
> >> mirror-mounted filesystems.
> > 
> > this seems odd to me -- is there some reason the unforced unmount can't
> > always recurse into the enclosed mirror mountpoints?
> 
> but if they're active - busy - then we shouldn't allow them to be 
> unmounted, surely? Just as if the enclosing mount were busy, we would 
> fail that unmount.
> 
> or am I missing your point, Bill?

No, I missed yours -- I managed to completely miss the second half of
the first sentence of the paragraph I quoted and skipped ahead to the
second sentence. 

now that i actually read it correctly, the behavior appears to be what
I'm expecting.  

						- Bill