[Linux-cluster] CLVMD hangs on 2nd node startup and hangs all gfs nodes.

[Tracey Flanders]

I tested creating  a GFS disk with 2 nodes started in the cluster without using LVMs and CLVMD stop.
I mounted the disk on the first node but when I mounted the 2nd node it did the same thing. So it seems its something 
other than CLVMD. I've attached my cluster.conf. It's kind of dumbed down because I was troubleshooting. So I removed the GFS mount for a services, etc.
This the config I used.



        
        
                
                        
                                
                                        
                                
                        
                
                
                        
                                
                                        
                                
                        
                
                
                        
                                
                                        
                                
                        
                
        
        
        
                
                
                
        
        
                
                        
                                
                                
                                
                        
                        
                                
                                
                                
                        
                        
                                
                                
                                
                        
                
        



> If you suspect a problem with clvmd, you could simply remove it from  
> the equation and retest, right?
> 
> You could just use the underlying iSCSI device and mkfs.gfs on  
> that.... at least for testing if clvmd is the problem.
> 
> I suppose you could also test if clvmd is the problem by testing the  
> logical volumes without GFS in the mix.  IOW, create some LVs and read/ 
> write to them at the same time from different machines.  If this is  
> working, the file system should work.  If the file system doesn't,  
> then the problem is probably higher up than clvmd.
> 
>   brassow
> 
> On Apr 25, 2008, at 11:34 AM, Tracey Flanders wrote:
> 
>> I've been trying to setup a 3 server cluster with GFS mounted over  
>> iSCSI on Qemu Virtual Machines. A 4th server acts as a iSCSI Target.  
>> I found and article that explains my issue, but I can't seem to  
>> figure out what the solution is. QUOTED from :http://kbase.redhat.com/faq/FAQ_51_10923.shtm 
>>  After successfully setting up a cluster, cman_tool shows the  
>> cluster is healthy. Mounting the gfs mount on the first node works  
>> successfully. However, when mounting gfs on the second node, the  
>> mount command hangs. Writing to a file on the first node also hangs.  
>> On the second node, the following error is seen in /var/log/ 
>> messages: Jul 18 14:49:27 blade3 kernel: Lock_Harness 2.6.9-72.2  
>> (built Apr 24 2007 12:45:55) installed Jul 18 14:49:27 blade3  
>> kernel: GFS 2.6.9-72.2 (built Apr 24 2007 12:46:12) installed Jul 18  
>> 14:52:53 blade3 kernel: GFS: Trying to join cluster "lock_dlm",  
>> "vcomcluster:testgfs" Jul 18 14:52:53 blade3 kernel: Lock_DLM (built  
>> Apr 24 2007 12:45:57) installed Jul 18 14:52:53 blade3 kernel: dlm:  
>> connect from non cluster node Jul 18 14:52:53 blade3 kernel: dlm:  
>> connect from non cluster node END QUOTE My Virtual Machines only  
>> have one interface so I still can't figure out why this is  
>> happening. I can successfully mount the GFS partition on any one  
>> node but as soon as I try to start the clvmd on a 2nd node it hangs  
>> the whole cluster. I'm wondering if its a Qemu VM network issue?  
>> Each host can ping each other by name and ip. The cluster works fine  
>> but I cant get GFS to work on th VMs. Is it possible to debug the  
>> clvmd to see what IP Address it is sending? Thanks, Tracey Flanders
>> In a rush? Get real-time answers with Windows Live Messenger. --
>> Linux-cluster mailing list
>> Linux-cluster at redhat.com
>> https://www.redhat.com/mailman/listinfo/linux-cluster
> 
_________________________________________________________________
Spell a grand slam in this game where word skill meets World Series. Get in the game.
http://club.live.com/word_slugger.aspx?icid=word_slugger_wlhm_admod_april08