NFS relationship logic is broken

[Charles Butler]

I've discovered by deploying through LXC containers the NFS relationship is broken. This is thanks to some sync issues and in reality should be done via lxc.host.mount in the container configuration, no big deal.

This however prompted me to investigate how well it functions in AWS. I deployed owncloud, deployed the nfs charm, and attempted to relate the services.

The hooks executed and exited without any sign of schenanigans, however when I remoted into the machine there was no external NFS mount present, as illustrated:

root@ip-10-169-8-130:/var/www# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/xvda1 7.9G 1.1G 6.5G 15% /
udev 819M 12K 819M 1% /dev
tmpfs 331M 180K 331M 1% /run
none 5.0M 0 5.0M 0% /run/lock
none 827M 92K 827M 1% /run/shm
/dev/xvdb 147G 188M 140G 1% /mnt

From what i saw, the shared-fs-relation-changed hook is supposed to symlink /var/www/owncloud to /var/lib/owncloud as the mounted FS, and push those files via RSYNC.. none of this was executed, nor was there any error output:

diving deeper into the logs however, I got a single line of ouput:
2014-01-30 18:07:11 INFO juju-log shared-fs:2: remote host not ready

so far i've tried remoting into both hosts and running the hooks interactively, but it appears that this message is there to stay. Somethings missing in translation between the hosts.

it appears that on the NFS host, the fail case at the top of the script is blocking it from transmitting the relationship data:

remote_ip=`relation-get private-address`
if [ -z "$remote_ip" ] ; then
    juju-log "wait for related service to start"
    exit 0
fi

therefore the changed hook never executes on owncloud:

if [ -z "${export_path}" ]; then
    juju-log "remote host not ready"
    exit 0
fi