[corosync] Two node cluster not connected to each other

Steven Dake sdake at redhat.com
Thu Dec 1 18:06:08 GMT 2011


The fact that broadcast works implies your multicast is not working as
you expect.  If broadcast isn't a good option, you might try out UDPU
(which avoids the hassles of multicast switch configuration).

I'd recommend following up with your second question on the
linux-cluster ml.

Regards
-stee

On 12/01/2011 09:00 AM, David Golovan wrote:
> Hi,
> 
> I have switched to broadcast mode and the cluster works fine until I try
> to setup quorum device.
> 
> Starting cman don't start qdisk, as the check cman script for quorum
> divice fails:
> 
> [root at clit1-p /]# ccs_tool query /cluster/quorumd
> Entity: line 25: parser error : Input is not proper UTF-8, indicate
> encoding !
> Bytes: 0xC0 0x22 0xF4 0x8A
> <cman expected_votes="1" broadcast="�"�" two_node="0" nodename="cli
>                                                 ^
> Entity: line 25: parser error : attributes construct error
> <cman expected_votes="1" broadcast="�"�" two_node="0" nodename="cli
>                                                   ^
> Entity: line 25: parser error : Couldn't find end of Start Tag cman line 25
> <cman expected_votes="1" broadcast="�"�" two_node="0" nodename="cli
>                                                   ^
> Entity: line 26: parser error : Opening and ending tag mismatch: cluster
> line 2 and cman
> </cman>
>        ^
> Entity: line 27: parser error : Extra content at the end of the document
> <fencedevices>
> ^
> Query failed: No such file or directory
> 
> 
> I can change cman script to start qdisk always and everything works, but
> it's not the right thing to use the cluster imo...
> 
> David Golovan
> 
> 
> On 11/30/2011 08:36 PM, Steven Dake wrote:
> 
>> try broadcast mode
>>
>> On 11/30/2011 10:24 AM, David Golovan wrote:
>>> All setting set to default(multicast address and port) now and same
>>> still problem.
>>>
>>> The multicast looks like working, I can't check the switch, but we ran
>>> multicast test on both servers(using mdump tool) and it works fine.
>>> Also, using tcpdump I can see that both servers communicate with each
>>> other using cluster multicast address and port.
>>>
>>> David Golovan
>>>
>>> On 11/30/2011 06:44 PM, Digimer wrote:
>>>> On 11/30/2011 11:39 AM, David Golovan wrote:
>>>>> Hi,
>>>>>
>>>>> Thanks for the help.
>>>>>
>>>>> I have removed the multicast address(now the cluster choose the address)
>>>>> and tried to change the default port for multicast (from 5404 to 5400),
>>>>> but no luck, still same problem...
>>>>>
>>>>> While changing this settings I have noticed something:
>>>>>
>>>>> I have removed to multicast address from the cluster(using luci) and
>>>>> reboot only one node. When the node started it couldn't connect to the
>>>>> cluster(as they had different multicast addresses), so it created
>>>>> cluster of his own with same ring id as the second node's cluster. And
>>>>> this time it didn't entered the loop of creating cluster generation number
>>>>>
>>>>> David Golovan
>>>> If you can avoid changing the port or the IP and just let the cluster
>>>> use the default multicast, it would be best. Also make sure that, if you
>>>> have a managed switch, that it is handling multicast properly.
>>>>
>>>
>>> _______________________________________________
>>> discuss mailing list
>>> discuss at corosync.org
>>> http://lists.corosync.org/mailman/listinfo/discuss
> 



More information about the discuss mailing list