[systemd-devel] [PATCH] util: always consider glusterfs as a network filesystem

[Lennart Poettering]

On Mon, 24.03.14 17:48, Lennart Poettering (lennart at poettering.net) wrote:

> On Mon, 24.03.14 17:38, Lukáš Nykrýn (lnykryn at redhat.com) wrote:
> 
> > 
> > Po 24. březen 2014, 17:12:30 CET, Lennart Poettering napsal:
> > >On Mon, 24.03.14 15:55, Lukas Nykryn (lnykryn at redhat.com) wrote:
> > >
> > >Are you sure this does what you think it does? I mean, AFAIR glusterfs
> > >is a fuse FS, and hence will nevershow up like this in fstab nor
> > >/proc/self/mountinfo... Or am I missing something?
> > >
> > Reason for this was
> > https://bugzilla.redhat.com/show_bug.cgi?id=974626 , it is probably
> > not accessible so in shortcut: In rhel 6 someone is using
> > glusterfs,_netdev in fstab and they want me to patch rc.sysinit to
> > recognize glusterfs as a network fs by default (so no need for
> > _netdev). I don't want to introduce such change in rhel6 and that
> > have a regression in rhel7 with systemd.
> 
> Well, but how exactly do those lines look in fstab? Are you sure they
> have "glusterfs" in the fstype field? And not "fuse.glusterfs" or so?

Hmm, now, reading the actual BZ, it says:

    "However, my understanding is that _netdev should be dedicated to file
    system relying on network devices (iSCSI for e.g.), not "network file
    systems" as such."

The mount(8) man page however says:

    _netdev
              The filesystem resides on a device that requires network
              access (used to  prevent  the  system
              from attempting to mount these filesystems until the
              network has been enabled on the system).

To me it really appears as if "_netdev" is the right thing to use here,
and we really should advertise its use for cases like this, and close
the bug as WONTFIX or NOTABUG...

I mean, the fact that we imply _netdev behaviour with nfs, and so on is
mostly to be compatibile with old fstabs. But given that glusterfs is
already documented explicitly to require _netdev, like for example here:

http://gluster.org/community/documentation/index.php/Gluster_3.1:_Automatically_Mounting_Volumes

And given that we really shouldn't attempt to add every possible network
fs in the world to our list I think we shouldn't add this at all here.

Lennart

-- 
Lennart Poettering, Red Hat