[Linux-cluster] fencing issue in 2 nodes cluster

[Heiko Nardmann]

Am 02.08.2012 17:02, schrieb Digimer:
> On 08/02/2012 08:42 AM, Heiko Nardmann wrote:
>> Am 02.08.2012 14:19, schrieb AKIN ÿffffffffffd6ZTOPUZ:
>>> Hi
>>> I have fencing problem in 2 nodes cluster  <cman expected_votes="1"
>>> two_node="1"/> )
>>> fence device  agent  is    like that :
>>> <fencedevice agent="fence_ipmilan" ipaddr="***********" lanplus="1"
>>> login="clsfenceadmin" method="cycle" name="fence_node2"
>>> passwd="**************" power_wait="4"/>
>>> when I run fence_node   nodename    command  on  host   , Related node
>>> goes to down but   I am taking errors in /var/log/messages  :
>>> Aug  2 14:55:31 sapclsn2 fenced[6714]: fencing node 
>>> "sapclsn1.edase.com"
>>> Aug  2 14:55:32 sapclsn2 fenced[6714]: agent "fence_ipmilan" reports:
>>> Rebooting machine @ IPMI:192.168.11.68...Done
>>> Aug  2 14:55:32 sapclsn2 fenced[6714]: fence "sapclsn1.edase.com" 
>>> failed
>>> you have any ideas?
>>>
>>>
>>> -- 
>>> Linux-cluster mailing list
>>> Linux-cluster at redhat.com
>>> https://www.redhat.com/mailman/listinfo/linux-cluster
>>
>> Without looking at it: how is this agent implemented? Maybe you can
>> easily debug it to see why it returns a non zero exit code?
>>
>>
>> Kind regards,
>>
>>      Heiko
>
> It's written in C;
>
> http://git.fedorahosted.org/cgit/fence-agents.git/tree/fence/agents/ipmilan/ipmilan.c 
>
>

To what I see from the source it supports a '-v' option to increase 
verbosity.

@Akin: are you able to run the command manually with success? Every 
entry inside the cluster.conf has an equivalent concerning the command 
line options of the fencing agent (should be /usr/sbin/...). So if you 
follow digimers advice and pass only those parameters you need together 
with the '-v' then maybe you easily find out what is wrong.


Kind regards,

     Heiko