Please it is very urgent: about cluster suite on Dapper

[Fabio Massimo Di Nitto]

carlopmart wrote:
> Hi Fabio,
> 
> 
> Fabio Massimo Di Nitto wrote:
>> carlopmart wrote:
>>> Hi Fabio,
>>>
>>>  My questions:
>>>
>>>  a) gfs init script. In previous message, I have do it this question.
>>> Today I receive your answer, but I think that is not possible to do as
>>> you say:
>>>
>>>   - If I put an entry on fstab to mount a gfs filesystem, when system
>>> shutdowns it can not umount this filesystem. Due to this, I think that
>>> is necessary to include a gfs init script on gfs-tools package like
>>> RedHat/CentOS does.
>>>
>>>   - If I setup as a resource, system services managed by rgmanger can
>>> fail (in my configuration httpd and postfix fails due to not gfs
>>> filesystem is mounted).
>>>
>>
>> SCORE! I totally miss the init script while converting all of them.
>> I will fix it within today.
>> (NOTE that the init script from redhat is still based on fstab entries
>> set to noauto!)
> 
> Correct, On my fstab I put noauto entry (like RedHat gfs init script
> requires), but without init script all services configured failed to start.

I uploaded a redhat cluster suite in dapper (i assume we are talking dapper
here) to ship the gfs init script. I did check again
and the noauto must be removed. The script does respect that.
In Ubuntu gfs volumes are not mounted by default, and they will mounted by
the script.

>>
>>>   c) fence devices. I am using fence-gnbd as a fence device. When one
>>> node start up and the other is off, fence fails. Is this a bug?.
>>
>> I will need to check this one specifically. Could you please
>> make your config available somewhere? or send it (only the config)
>> to private email?
> 
> OK, as soon as possible I will send you ..
> 

I did check your config but could you be so kind to paste the exact error?

As far as i know you need the 3rd node to join the cluster and the fence domain
otherwise the nodes do not know how to tell saruman how to fence gnbd, but it
doesn't need any other service.

It is also ideal anyway to have 3 nodes voting in the cluster. In your
case with a quorum of 2.
The 3rd node will sort of keep an eye on the other 2 that are doing the real
work without taking services.

Fabio

-- 
I'm going to make him an offer he can't refuse.