Gluster users - Mar 2013 - Can't get gluster to mount on boot

This is something I've been trying to get to work for the last year or so,
but have still not found a working solution.

I have two servers in a simple replication config of a single volume running
gluster 3.3.0 (I'm avoiding 3.3.1 because of its known NFS issues) on Ubuntu
Lucid. As well as hosting a gluster server, each mounts the gluster volume via
NFS over tcp on localhost. There is also a third server that mounts the NFS
volume but is not running any gluster code (client nor server), and that
displays exactly the same symptoms, so I don't think running gluster server
or accessing via localhost is causing the problem.

Firstly, a manual mount works fine, so I know that the NFS settings, paths,
permissions and the gluster volume itself are ok.

If I try to mount via fstab (even with _netdev), it hangs the machine on boot
and it must be put into single-user mode in order to restore it, so I'm not
trying that approach for now.

This list suggested using autofs, so I've been trying that, with no success.
I've created a /etc/auto.nfs file containing this:

/var/lib/sitedata/aegir
-fstype=nfs,vers=3,hard,noexec,nosuid,nodev,rsize=32768,wsize=32768,intr,noatime,mountproto=tcp
127.0.0.1:/shared

and referred to it from /etc/auto.master, and commented out the
'+auto.master' line, as suggested on docs I found (it doesn't work
work either way). autofs does do *something* here, but it appears to completely
ignore NFS and instead creates and rebinds /shared locally at
/var/lib/sitedata/aegir. I've tried with a minimal options list (just
-fstype=nfs,mountproto=tcp) but it doesn't help.

gluster peer status and gluster volume status both say all is present and
correct with respect to the underlying gluster volume. The only anomaly I am
seeing is this happening once per second in syslog:

Mar 12 17:09:49 web2 init: glusterd main process ended, respawning
Mar 12 17:09:50 web2 init: glusterd main process (20365) terminated with status
255

I've googled that and not found anything at all.

How does everyone else do automounts?

Marcus
-- 
Marcus Bointon
Synchromedia Limited: Creators of http://www.smartmessages.net/
UK info at hand CRM solutions
marcus at synchromedia.co.uk | http://www.synchromedia.co.uk/

On 12 Mar 2013, at 19:14, Marcus Bointon <marcus at synchromedia.co.uk>
wrote:
> This list suggested using autofs, so I've been trying that, with no
success. I've created a /etc/auto.nfs file containing this:
> 
> /var/lib/sitedata/aegir
-fstype=nfs,vers=3,hard,noexec,nosuid,nodev,rsize=32768,wsize=32768,intr,noatime,mountproto=tcp
127.0.0.1:/shared
> 
> and referred to it from /etc/auto.master, and commented out the
'+auto.master' line, as suggested on docs I found (it doesn't work
work either way). autofs does do *something* here, but it appears to completely
ignore NFS and instead creates and rebinds /shared locally at
/var/lib/sitedata/aegir. I've tried with a minimal options list (just
-fstype=nfs,mountproto=tcp) but it doesn't help.
I've set up autofs on a separate server that has no gluster components
installed and mounts the volume via nfs (exactly as I want my gluster nodes to
do), and I'm happy to report it works perfectly. The only difference between
the configs is the ip that the NFS mount points at (remote vs 127.0.0.1). My
suspicion is that the NFS client is seeing the 127.0.0.1 address and thinking it
can use a local bind instead of actually going over NFS, and is thus failing.
How can I force the NFS client go via the network? Or do you think the problem
is elsewhere?

Marcus
-- 
Marcus Bointon
Synchromedia Limited: Creators of http://www.smartmessages.net/
UK info at hand CRM solutions
marcus at synchromedia.co.uk | http://www.synchromedia.co.uk/