ejabberd - Comments for "Clustering two servers"

Database removal

bbarnes — Wed, 22 Feb 2006 20:17:45 +0000

Is there more to it than deleting the contents of the spool directory where the files are? If not then I did that and get the same error.

Re: Hostnames on my server

sander — Wed, 22 Feb 2006 18:34:05 +0000

bbarnes wrote:

The good news is that I am now able to cluster, if I start node2 with erl -sname ejabberd -mnesia extra_db_nodes "['ejabberd@node1']" -s mnesia I get the following error:

** ERROR ** (ignoring core) ** FATAL ** Failed to merge schema: Incompatible schema storage types (local). on ejabberd@node2 storage ram_copies, on ejabberd@node1 storage disc_copies

Maybe that is because you started ejabberd first wrong on the node so that the database contains the wrong domain names. Maybe you can try to remove the database of the server with the error? (make a backup of course)

Hostnames on my server

bbarnes — Wed, 22 Feb 2006 14:51:53 +0000

Ok, I get the following results:

Hostname -s
node1

Hostname
node1@example.net

Ejabberd will only start and accept connections if I start it with -sname ejabberd. All others fail saying the system is configured to use fully configured hostnames.

The good news is that I am now able to cluster, if I start node2 with erl -sname ejabberd -mnesia extra_db_nodes "['ejabberd@node1']" -s mnesia I get the following error:

** ERROR ** (ignoring core) ** FATAL ** Failed to merge schema: Incompatible schema storage types (local). on ejabberd@node2 storage ram_copies, on ejabberd@node1 storage disc_copies

But at least they are talking to each other now.

erlang node name with fully qualified hostname

mfoss — Wed, 22 Feb 2006 10:40:35 +0000

This is how it works for me, I don't know if it's the same for everybody:

Short hostname

$ hostname -s
monster

Fully Qualified hostname

$ hostname
monster.example.net

Start Erlang with a short hostname

$ erl -sname ejaby
Erlang (BEAM) emulator version 5.4.12 [source] [hipe] [kernel-poll]

Eshell V5.4.12  (abort with ^G)
(ejaby@monster)1>

Start Erlang with a short hostname (alternate)

$ erl -sname ejaby@monster
Erlang (BEAM) emulator version 5.4.12 [source] [hipe] [kernel-poll]

Eshell V5.4.12  (abort with ^G)
(ejaby@monster)1>

Start Erlang with a Fully Qualified hostname

$ erl -name ejaby
Erlang (BEAM) emulator version 5.4.12 [source] [hipe] [kernel-poll]

Eshell V5.4.12  (abort with ^G)
(ejaby@monster.example.net)1>

Start Erlang with a Fully Qualified hostname (alternate)

$ erl -name ejaby@monster.example.net
Erlang (BEAM) emulator version 5.4.12 [source] [hipe] [kernel-poll]

Eshell V5.4.12  (abort with ^G)
(ejaby@monster.example.net)1>

Re: Clustering two servers

bbarnes — Tue, 21 Feb 2006 19:49:10 +0000

If I attempt to run the following command:
erl -name ejabberd@node2 -mnesia extra_db_nodes "['ejabberd@node1']" -s mnesia

I get the following error:

** System running to use fully qualified hostnames **
** Hostname node1 is illegal **

And there is no connection attempt shown in the log of node1.

If I run the following command from node1:

erl -name test@node1 -remsh ejabberd@node1

I get the same error. What would cause that error?

Re: Clustering two servers

sander — Tue, 21 Feb 2006 19:24:01 +0000

bbarnes wrote:

It does occur to me that node1 is running as -name ejabberd@node1 and my test and node2 are running as -sname, would that cause this problem?

I guess so.

Clustering two servers

bbarnes — Tue, 21 Feb 2006 18:41:53 +0000

It does occur to me that node1 is running as -name ejabberd@node1 and my test and node2 are running as -sname, would that cause this problem?

Re: Clustering two servers

bbarnes — Tue, 21 Feb 2006 18:37:06 +0000

If I run erlang:get_cookie(). on both nodes they are identical. If I attempt to run a remote session to my -detached session on node1 from node1 I get the same error.

Re: Clustering two servers

sander — Tue, 21 Feb 2006 17:29:06 +0000

bbarnes wrote:

** Connection attempt from disallowed node ejabberd@node2 **

So, my nodes can in fact see each other. I have read and followed the instructions in http://www.ejabberd.im/interconnect-erl-nodes, so what am I doing wrong that the nodes don't trust each other?

I have no experience with clustering, but I guess it might help if you verify that the Erlang cookies on each node match eachother.

Clustering two servers

bbarnes — Tue, 21 Feb 2006 15:53:22 +0000

After doing additional troubleshooting, I have discovered in the ejabberd.log of the first machine the following line:

** Connection attempt from disallowed node ejabberd@node2 **

So, my nodes can in fact see each other. I have read and followed the instructions in http://www.ejabberd.im/interconnect-erl-nodes, so what am I doing wrong that the nodes don't trust each other?