Bug#622394: systemd: nfs-common and rpcbind unit files to fix systemd NFS issues properly

[Niels Thykier <niels@thykier.net>]

Hi,

Can anyone provide me with an up to date status on this bug?  It seems
to be a mismatch of no less than 3 different bugs if I understand
Simon's comment in message #41 correctly (see [1] for the full text).

In particular, this bug is correctly marked as a Jessie blocker due to
the current bug title, which suggests that nfs-common has a circular
dependency causing NFS to break under systemd.  However, reading the bug
I am understanding that this loop is now assumed to be fixed?

Simon McVittie wrote (cite [1]):
> Bug A, reported by Alban Browaeys, present in 1:1.2.3-2: "systemd[1]:
> Found ordering cycle on basic.target/start" and an unbootable system. I
> believe newer systemd and/or nfs-common have probably fixed this, so
> #622394 should probably have been closed, and the other symptoms should
> probably have had distinct bug numbers.
> 

If this is indeed closed, I would very much like this bug to be closed.
 If the other bugs (see below) are still present, let us clone this
bug/file separate bugs for those!  NB: If you clone this bug, please
remember to remove the "jessie-is-blocker" tag.

If these bugs below lack information or/and are unreproducible, please
also tag them as such.

> Bug M1, reported by Matthew Grant, present in 1:1.2.8-6: "NFS mounts
> from /etc/fstab do not work". (What symptoms? What error messages?) This
> one might be related to NFS entries in /etc/fstab that don't have _netdev.
> 
> 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".
> 
> Bug C1, sort-of-reported by Chris Butler, present in 1:1.2.8-9: "the
> boot order problem on NFS clients that mount remote filesystems on
> boot". This might be the same thing as M1. (Do you have _netdev in the
> options of the relevant mounts?)
> 
> Bug C2, reported by Chris Butler, present in 1:1.2.8-9 with Matthew's
> systemd units added: 'tries to start "statd" before "rpcbind"'. I
> believe this is an omission in one of Matthew's systemd units, and I am
> about to test a version of nfs-common that should fix that bit.


On the topic of "_netdev": Did you conclude that systemd correctly
auto-discovered NFS as a _netdev *without* explicitly stating it in
/etc/fstab?  There is a mail between Simon and Ben that suggests this to
be the case.  If *not* the case, we /can/ add a warning in the release
notes problems for that.
  We already have a clause for updating fstab with nodefault and nofail
on disks are not critical.  Having a paragraph for adding _netdev is not
going to be a huge problem at this point. :)

Thanks,
~Niels

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=622394#41