Re: NFS - lock failover

["J. Bruce Fields" <bfields@fieldses.org>]

On Mon, Dec 12, 2011 at 09:30:27PM +0100, Frank van Maarseveen wrote:
> On Mon, Dec 12, 2011 at 02:56:13PM -0500, J. Bruce Fields wrote:
> > On Fri, Dec 09, 2011 at 11:36:46AM +0100, Frank van Maarseveen wrote:
> > > On Thu, Dec 01, 2011 at 06:24:16PM +0100, Frank van Maarseveen wrote:
> > > > On Thu, Dec 01, 2011 at 07:07:22PM +0200, Pavel A wrote:
> > > > > Hi everyone!
> > > > > 
> > > > > I was trying not to create new topics, but it seems that posting to an
> > > > > old one doesn't bring it up. Here is the original topic I'm referring
> > > > > to: http://comments.gmane.org/gmane.linux.nfs/13108
> > > > > 
> > > > > I'm building an A/A cluster using NFS v3 and local file systems, and
> > > > > looking for
> > > > > efficient ways for failover (for now I have to restart nfs-kernel-server on
> > > > > Takeover node to be able to initiate grace period), so the discussed solutions
> > > > > are very interesting to me.
> > > > > 
> > > > > Now (4 years after) in current nfs-utils packages (v. 1.2.2-4 and later) I can
> > > > > see that the ability to release locks was really implemented and is
> > > > > working well
> > > > > (I mean interfaces /proc/fs/nfsd/unlock_ip and
> > > > > /proc/fs/nfsd/unlock_filesystem),
> > > > > but how about reacquiring locks on the node, share migrates to? - I've been
> > > > > going through various mailing lists and found a lot of discussions on the topic
> > > > > (also dated mainly 2007), but don't seem to find any rpc-based mechanism or
> > > > > interface like /proc/fs/nfsd/nlm_set_grace to do that, was it ever made?
> > > > 
> > > > I've posted a patch some time ago implementing
> > > > /proc/fs/nfsd/relock_filesystem:
> > > > 
> > > > 	http://permalink.gmane.org/gmane.linux.nfs/42360
> > > > 	http://permalink.gmane.org/gmane.linux.nfs/42361
> > > > 	http://permalink.gmane.org/gmane.linux.nfs/42362
> > > 
> > > Can this patch be scheduled for inclusion in mainline (3.3)?
> > 
> > Could you resend and I'll take another look?
> 
> I'll check it again but except for one fuzz it should still apply on
> 3.2-rc. Patch offsets and an incidental fuzz is sometimes informative
> about possible conflict areas. It has been verified on 3.0 and 3.1

Thanks.  It would be helpful to have them resent in any case.

> > But this should really be fixed to handle v4 locks as well.
> 
> Why do you think it doesn't?

The fact that unlock_* doesn't.  Looking at it.... I suppose that
doesn't matter as much on the grace-period starting side.  How much use
does that have on its own?

--b.