[Gluster-users] connection refused with AFR single process (1.3.8pre)

[Andrew McGill]

On Wednesday 19 November 2008 17:06:34 Ilan Berci wrote:
> Dear Glusterfs experts,
>
> Situation: I am here sitting in the server room which is colder than my
> fridge surrounded by loud windows machines all around me that keep blinking
> their obnoxious flashing lights at me just to drive me crazy.  
I once spent a day in a server room where some obnoxious id10-t had installed 
UV lighting.  I don't care how clever that might be - it's stupid -- but read 
on ...

> I am a server side developer that works on linux and know next to nothing
> about ops (or administering linux) but have somehow found myself in here
> deploying linux servers for various tasks (we are a small company).  I am
> looking for a simple afr between some of my application servers for various
> data that we don't keep in our dbs.
>
> Here is the situation
>
> - glusterfs 1.3.8pre
> - running on Debian 4.0r5
> - retrieved and installed from apt-get with NO modifications
>
>
> - running the script AFR single process :
> http://www.gluster.org/docs/index.php/AFR_single_process - ONLY
> modifications to the script are the machine names.
> - I run glusterfs EXACTLY as it states below the example:  glusterfs -f
> client.vol /mnt/glusterfs
>
> I am getting a connection refused to the other host when I check the logs..
> The FAQ on this error informs me to telnet from machine1 to machine2 on
> port 6996 whereby I get the same error.  (But if I use the regular port of
> 23 I get in fine) so I assume that process is not binding to the port
> properly.
If you telnet to machine2 and get connection refused that means that either:
 * there is no service running on the remote machine
 * the firewall on the remote machine refuses the connection with -j 
REJECT --reject-with tcp-reset (or something like that)

On the "server" side:
 * Check if the service is running -- 
	netstat -lntp | grep 699
   Failing this review /var/log/glusterfsd.log
 * Review the firewall rules 
	iptables-save -c | less
   The counters will change if you hit a rule.

&:-)

>
> - The mount is fine on both machines and fuse (also deployed with apt-get)
> is loaded when glusterfs starts. - The files are moving fine between the
> posix directories and the mount points.
>
> - I wiped the firewalls on both boxes and tracerouted to make sure that
> there was indeed only ONE hop between both machines.
>
> I then started grasping at straws and then tried running glusterfs and
> glusterfsd (which to my understanding is not needed with builds >= 1.3.8)
> and I get the same results.
>
> If I do an nmap on localhost, I don't see anything sticking to port 6996,
> If I do an nmap from the outside interface, I still don't see anything on
> port 6996 (my understanding is that the daemon is listening on all
> interfaces by default so this doesn't surprise me)
>
> - I tried specifying the port and interface explicitly in the .vol file
> only to get the same results.
>
> Besides this, I think that glusterfs is a work of art.. It's very straight
> forward, well documented, and a joy to work with..   The fact that I don't
> have to compile anything into the kernel is great as well.. :)
>
> Any help on this would be greatly appreciated..
>
> ilan
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users