Bug#622394: systemd: nfs-common and rpcbind unit files to fix systemd NFS issues properly
On Sat, 08 Nov 2014 at 01:16:27 +0000, Ben Hutchings wrote:
> On Sat, 2014-11-08 at 00:37 +0000, Simon McVittie wrote:
> > The symptom that *I* get on some boots is these messages in the NFS
> > client's Journal/syslog (my test-case for NFS is mounting one machine's
> > /srv onto the other machine's /srv):
> >
> > mount[309]: mount.nfs: Network is unreachable
> > systemd[1]: srv.mount mount process exited, code=exited status=32
> > systemd[1]: Failed to mount /srv.
> > systemd[1]: Dependency failed for Remote File Systems.
>
> This line seems to indicate that /srv was a dependency for 'Remote File
> Systems', i.e. it has correctly been recognised as remote.
That's a good point.
> But network-online.target doesn't seem to have any integration with
> ifupdown, i.e. there is no ifupdown-wait-online.service.
As is that. Michael Biebl has proposed an ifupdown patch in
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=766943#129
which makes /etc/init.d/networking wait for "udevadm settle",
then bring up allow-hotplug interfaces in addition to auto interfaces.
As far as I can see, that should make networking.service wait for
network interfaces (or at least, the network interfaces that would
have been brought up at this point by Debian 7); if network.target
gained After=networking.service, then that ought to be enough to
fix what I described as bug M1?
(Elsewhere on that bug he also provided a proof-of-concept
ifupdown-wait-online.service which would remove the need for network.target
to have an explicit After=networking.service, AIUI.)
> > > Bug M2, reported by Matthew Grant, present in 1:1.2.8-6: "NFS exports
> > > also fail due to rpcbind not starting before nfs-common and nfs-
> > > kernel-server".
> >
> > I have not been able to reproduce this, with or without native systemd
> > units.
> >
> > One possible reason why this could happen sometimes is that
> > /etc/init.d/rpcbind has "Provides: rpcbind", which seems ... redundant.
> > I would expect it to have "Provides: $portmap" in order to be ordered
> > before nfs-common, which has "Required-Start: $portmap".
>
> So would I. But this should result in breakage under sysvinit too.
I opened https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=768548.
Regards,
S
Reply to: