[Gluster-users] 3.2.5 client connectivity issues...

JJ jjones at cirrhus9.com
Sat Mar 10 05:12:38 UTC 2012


Hello gluster-users:

I have installed and configured a 2 brick replication setup and it 
appears to be a successful install. Writes on 1 brick are replicated on 
the other.

My current dilemma is my client will NOT stay connected.

I don't know what information to provide, so I will just give you the
install steps I have taken THIS TIME.

Recipe for this Ubuntu 12.04/32 host in Vbox:
Yes, I realize it's "beta". If using Ubuntu 11.x would be better, I can
re-install this Vbox host using that.

cd /usr/src
apt-get install -y bison flex g++-4.4 make
wget 
http://download.gluster.com/pub/gluster/glusterfs/LATEST/glusterfs-3.2.5.tar.gz
tar zxf glusterfs-3.2.5.tar.gz && cd glusterfs-3.2.5
./configure && make && make install && ldconfig
vi /usr/local/etc/glusterfs/glusterfs.vol
volume cloud_c9
   type protocol/client
   option transport-type tcp
   option remote-host xx.xxx.xx.xxx
   option remote-subvolume brick
end-volume
mount -t glusterfs xxx.xx.xxx.xxx:cloud_c9 /mnt/cloud_c9 (or
mount -t glusterfs xxx.xx.xxx.xxx:cloud_c9 /mnt/cloud_c9 -o 
log-level=WARNING,log-file=/root/gclient.out)

mount shows it but when/if I df -h, it's gone and /root/gclient.out 
shows http://susepaste.org/97668544 (expires in 6 days).

No firewall rules in the way.
I can ping each and every host from each and every other host and the 
client. I added my IP using "gluster volume set cloud_c9 auth.allow 
xxx.xxx.xxx.xxx" successfully
fuse init (API version 7.17) is loaded.

My Questions are:
What should a "./configure --prefix=..." be to get the logs normalized
(into /var/log/glusterfs/) and the glusterfs to /etc/glusterfs/?'

./configure --prefix=/something --sysconfdir=/etc/something...?

On the bricks, it goes to /var/log/glusterfs...and /etc/glusterfs
but that's a CentOS host. IF the paths are just a Debian-derivative 
"thing" then I'll live with that.

I have to get the BOSS connected next week, or else, this project might 
get sidelined, or delayed, or worse!? (Not really worse)

We are a Cloud Hosting Provider of the AppLogic Grid Computing Software
and I'd really like to be able to provide some storage using glusterfs 
and I don't have the luxury of letting the clients "figure out" their 
connectivity issues.

It's not a marketable feature if the client(s) can't connect.

IT REALLY SHOULD NOT BE THIS DIFFICULT to get a client connected.
And some frequent convos on #gluster indicate that I am not the only 
person having difficulty with this task.

Are bounties allowed to be posted via this list? Because I am seriously
about fed up trying to MAKE this work.

Thank you all for your time.

-- 

JJ / Habitual on #gluster



More information about the Gluster-users mailing list