Date: Fri, 29 Sep 2006 16:21:15 -0700 From: John-Mark Gurney <gurney_j@resnet.uoregon.edu> To: Robert Watson <rwatson@FreeBSD.org> Cc: ru@FreeBSD.org, Pawel Jakub Dawidek <pjd@FreeBSD.org>, current@FreeBSD.org Subject: Re: lockf in installworld -- not a good idea Message-ID: <20060929232115.GT80527@funkthat.com> In-Reply-To: <20060929172321.R74256@fledge.watson.org> References: <20060929141709.E70454@fledge.watson.org> <20060929155526.GA9194@garage.freebsd.pl> <20060929172321.R74256@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Robert Watson wrote this message on Fri, Sep 29, 2006 at 17:24 +0100:
>
> On Fri, 29 Sep 2006, Pawel Jakub Dawidek wrote:
>
> >On Fri, Sep 29, 2006 at 02:20:06PM +0100, Robert Watson wrote:
> >>I've noticed an increasing intolerance in our tools for system install
> >>and maintenance to locking not being implemented over the past few years.
> >>I no longer get working
> >>cron on boxes with neither rpc.lockd nor local locking enabled, for
> >>example. [...]
> >
> >If you are refering to my change in which cron(8) started to use
> >pidfile(3), then I'm sorry, but you're wrong.
> >
> >cron(8) from the very beginning was exiting when it had problems with
> >creating a pidfile, please check function acquire_daemonlock() in:
>
> Cron may have been a poor example, I picked it from the top of a list of a
> half dozen failed services, many of which used not to fail. Sorry about
> that.
>
> >The way I prefer is to ignore errors other than EEXIST - you can check
> >EXAMPLES section in the pidfile(3) manual page for more info. I just
> >didn't wanted to change cron(8)'s original behaviour.
> >
> >I do agree, that this shouldn't be treated as critical error and I can
> >change it if you like.
>
> I think if we get back EOPNOTSUPP from lock attempts in libpidfile, we
> should ignore it and hope for the best. In an ideal world, we wouldn't do
> this, but in an ideal world we also wouldn't need to. :-)
Why not:
-L Do not forward fcntl(2) locks over the wire. All locks will be
local and not seen by the server and likewise not seen by other
NFS clients. This removes the need to run the rpcbind(8) service
and the rpc.statd(8) and rpc.lockd(8) servers on the client.
Note that this option will only be honored when performing the
initial mount, it will be silently ignored if used while updating
the mount options.
Seems like most of these cases -L makes the most sense... I don't see
a reason you'd share var's between machines, and doubt that you'd be
installing world to the same tree from two different machines...
--
John-Mark Gurney Voice: +1 415 225 5579
"All that I will do, has been done, All that I have, has not."
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060929232115.GT80527>