[Gluster-users] Geo-replication fails to even TRY to start doesnt create /tmp/gsyncd-* dir

[Tony Maro]

Setting up Geo-replication with an existing 3 TB of data is turning out to
be a huge pain.

It was working for a bit but would go faulty by the time it hit 1TB synced.
 Multiple attempts resulted in the same thing.

Now, I don't know what's changed, but it never actually tries to log into
the remote server anymore.  Checking "last" logs on the destination shows
that it never actually attempts to make the SSH connection.  The
geo-replication command is as such:

gluster volume geo-replication docstore1
root at backup-ds2.gluster:/data/docstore1
start

>From the log:

[2013-07-26 10:26:04.317667] I [gsyncd:354:main_i] <top>: syncing:
gluster://localhost:docstore1 -> ssh://root@backup-ds2.gluster
:/data/docstore1
[2013-07-26 10:26:08.258853] I [syncdutils(monitor):142:finalize] <top>:
exiting.
[2013-07-26 10:26:08.259452] E [syncdutils:173:log_raise_exception] <top>:
connection to peer is broken
*[2013-07-26 10:26:08.260386] E [resource:191:errlog] Popen: command "ssh
-oControlMaster=auto -S /tmp/gsyncd-aux-ssh-WlTfNb/gsycnd-ssh-%r@%h:%p
root at backup-ds2.gluster /usr/lib/glusterfs/glusterfs/gsyncd --session-owner
24f8c92d-723e-4513-9593-40ef4b7e766a -N --listen --timeout 120
file:///data/docstore1" returned with 143*

When I attempt to run the SSH command from the logs directly in the
console, ssh replies with:

muxserver_listen bind(): No such file or directory

And, there's no gsyncd temp directory where specified.  If I manually
create that directory and re-run the same command it works.  The problem of
course is that the tmp directory is randomly named and starting Gluster
geo-rep again will result in a new directory it tries to use.

Running Gluster 3.3.1-ubuntu1~precise9

Any ideas why this would be happening?  I did find that my Ubuntu packages
were trying to access gsyncd in the wrong path so I corrected things.  I've
also got auto-ssh login using root so I changed my ssh command (and my
global ssh config) to make sure the options would work.  Here's the
important geo-rep configs:

ssh_command: ssh
remote_gsyncd: /usr/lib/glusterfs/glusterfs/gsyncd
gluster_command_dir: /usr/sbin/
gluster_params: xlator-option=*-dht.assert-no-child-down=true

Thanks,
Tony
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130726/159755e8/attachment.html>