[Linux-cluster] GFS and samba problem in Fedora, again

[sandra-llistes]

Hi Ahbi,

I've proved to install kernel 2.6.10 and GFS from cvs on Fedora 5 
because if this software worked for RHES4 It will work also for Fedora.
I have had to recompile the kernel and software, and have some 
problems because Fedora has gcc 4 and Red Hat gcc 3.4.6. Now I 
succesfully installed it but I'm getting odd errors with ccsd:

Nov  7 16:44:39 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 120 seconds.
Nov  7 16:45:09 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 150 seconds.
Nov  7 16:45:40 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 180 seconds.
Nov  7 16:46:10 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 210 seconds.
Nov  7 16:46:40 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 240 seconds.
Nov  7 16:47:10 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 270 seconds.
Nov  7 16:47:40 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 300 seconds.
Nov  7 16:48:10 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 330 seconds.
Nov  7 16:48:40 nocilla ccsd[15826]: Unable to connect to cluster 
infrastructure after 360 seconds.

cman starts ok and quorum is regained, but ccsd fails. I tried:
[root ~]# ccs_test connect
ccs_connect failed: Connection refused
[root ~]# ccs_test connect force
Force is set.
Connect successful.
  Connection descriptor = 60

Also I tried to start ccsd with -4 and -I parameters with the same 
results:
  #ccsd -4 -I 127.0.0.0.1

What could the problem be?
Thanks,

Sandra


Abhijith Das wrote:
> Hi Sandra,
> 
>> Hi Abhi,
>>
>> Your mail astonished me. The only difference between your environment 
>> and our is that you've RHEL4 and we've Fedora 5 with GFS.
> 
> I'll get FC5 installed on a test cluster and try it out.
> 
>> I'm sorry about this but I have to insist. Are you completelly sure 
>> that your samba access was simoultaneous? because if you probe one 
>> client, and then another it isn't the same.
>> I found people that complain about the same problem we have:
>> http://www.redhat.com/archives/linux-cluster/2004-November/msg00065.html
>> http://www.centos.org/modules/newbb/viewtopic.php?topic_id=4997
> 
> Yes, the access was simultaneous. If you have a specific test, (specific 
> types of files or windows programs) I can try that too. I'm not denying 
> that you're seeing problems with GFS+samba :-). If I can reproduce your 
> problem, I can chase it down and we'll know why this is happening.
> 
>> Well, I had finally compiled GFS2 but I was uncapable to start cman 
>> daemons.
>> [root at server2 ~]#  /etc/init.d/cman start
>> Starting cluster:
>>    Loading modules... done
>>    Mounting configfs... done
>>    Starting ccsd... done
>>    Starting cman... failed
>> /usr/sbin/cman_tool: aisexec daemon didn't start
>>
>> And I obtain this when I try: # strace /usr/sbin/cman_tool -t 120 -w join
>> ....
>> connect(5, {sa_family=AF_FILE, path="/var/run/cman_admin"}, 110) = -1 
>> ENOENT (No such file or directory)
>> close(5)
>> write(2, "/usr/sbin/cman_tool: ", 21/usr/sbin/cman_tool: )   = 21
>> write(2, "aisexec daemon didn\'t start\n", 28aisexec daemon didn't start
>> ) = 28
>> exit_group(1)
>>
>> So I'm stalled by that way.
> 
> Looks like the openais component of the cluster is missing. The cluster 
> architecture has changed quite a bit for GFS2. Please refer to this for 
> help installing GFS2:
> http://sources.redhat.com/cgi-bin/cvsweb.cgi/cluster/doc/usage.txt?rev=1.35&content-type=text/x-cvsweb-markup&cvsroot=cluster 
> 
> 
>> I'm downloading RHEL 30-days evaluating version and I'm also trying 
>> samba 3.0.23c compiling with cluster support.
>> I will tell you something as soon as I finish all test.
>> Regards,
>>
>> Sandra
> 
> Thanks,
> --Abhi