On Tue, Jul 11, 2006 at 01:45:14PM -0500, Nicolas Williams wrote:
> Is the directory a ZFS filesystem root? If yes...
No.
> If the latter, then if the /net/<host> mounts had been done before
> cretion of the server-side filesystem in question, which means that the
> client won't see the new filesystem until after the /net/<host> mounts
> are auto-unmounted.
None of the filesystems in question are being created, destroyed, shared, or
unshared. None of the machines in question are being rebooted or reconfigured.
The only slightly strange thing here is that I have a shared filesystem
/net/host/foo with a nested filesystem which is also shared /net/host/foo/bar.
> > Is this a bug in Solaris? If not, is there some way to address the problem
> > without rebooting?
>
> It's an irritating lack of a feature.
>
> If using NFSv3 the missing feature is, I think, periodic probing of the
> server's MOUNTs and updating of the client mounts accordingly. this
> will never happen for v3.
>
> If using NFSv4 the missing feature is client-side crossing of
> server-side mount points ("mirror mounts"). Currently when using NFSv4
> the -hosts automount map still uses the NFSv3 MOUNT protocol to figure
> out what the server exports.
It's NFSv4. I think it's a little generous to describe this as a "lack of
a feature" -- sporadically my data disappears.
Adam
--
Adam Leventhal, Solaris Kernel Development http://blogs.sun.com/ahl
