[Linux-cluster] [Q] Good documentation about command line interface??

[Hiroyuki Sato]

Hello Digimer and Kaloyan

Thank you for your information.

I'll set gfs1, gfs2 and gfs3 with lowest ID (ex, 1,2,3).

I found the following Notes in fenced/recover.c

recover.c

   Notes:
   - When fenced is started, the complete list is initialized to all
   the nodes in cluster.conf.
   - fence_victims actually only runs on one of the nodes in the domain
   so that a victim isn't fenced by everyone.
   - The node to run fence_victims is the node with lowest id that's in both
   complete and prev lists.
   - This node will never be a node that's just joining since by definition
   the joining node wasn't in the last complete group.
   - An exception to this is when there is just one node in the group
   in which case it's chosen even if it wasn't in the last complete group.
   - There's also a leaving list that parallels the victims list but are
   not fenced.


Here is call procedures.

recover.c
  do_recovery
    fence_victims
      dispatch_fence_agent

agent.c
  dispatch_fence_agent
    use_device
      run_agent
        exec fence_XXXX

Regards.



2011/5/31 Digimer <linux at alteeve.com>:
> On 05/30/2011 11:30 AM, Kaloyan Kovachev wrote:
>>>
>>> actually sends the fence command. I do know that it has to come from a
>>> machine with quorum, and I do believe it is handled by the cluster
>>> manager. It's not like pacemaker where a DC is clearly defined.
>>>
>>> I'll try to sort out how the internals work and will let you know.
>>
>> Not sure where i got this information from (i think it was on this list),
>> but for sure: the node with the lowest ID, which is quorate, will take the
>> responsibility to call the fencing script
>
> Indeed, you are right. :)
>
> --
> Digimer
> E-Mail:              digimer at alteeve.com
> Freenode handle:     digimer
> Papers and Projects: http://alteeve.com
> Node Assassin:       http://nodeassassin.org
> "I feel confined, only free to expand myself within boundaries."
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>



-- 
Hiroyuki Sato