On Wed, Oct 29, 2008 at 10:35:39AM +0100, Hans de Goede wrote:
Richard Hughes wrote:
> mean we are still loading sendmail, isdn, nfslock, and all other legacy
> stuff in ten years time.
>
Talking about the nfs (and related rpc daemons) I recently had a good
idea for those, some will hate it, but those are the people who so no to
anything looking like progress, so I say lets just ignore them.
This idea is taken from some work I've done recently to stop the iscsi
related daemons from starting unnecessary be default, something which has
become relevant as the libvirt stuff drags in iscsi-initiator-utils on
very installation that has libvirt, of which there are quite a few.
So the idea is to patch all the initscripts which currently only get
started by default for NFS, to as the first thing in their start method
grep for nfs in /etc/fstab and if that is not found exit quietly without
doing anything further.
Then to make manual mounting of nfs exports not in fstab work properly,
the suggested fix is to patch mount to check if the necessary daemons are
running and if not start them. Since this will probably turn out to be
distro specific I think it would be best to have a checknfsdaemons shell
script in some standard location provided by the distro and let mount
call that before each nfs mount. (Oh the horror an additional entire fork
and exec on each nfs mount, hello people this is going over the network,
forget about the fork and exec cost, thank you).
You are forgetting autofs in your scheme, my systems never have nfs in
/etc/fstab but if nfs doesn't start the machines are useless. Why not
let people do /sbin/chkconfig nfs on|off instead of trying to guess what
they want.
Kostas Georgiou