On Tue, Jul 15, 2008 at 1:26 PM, Chase Simms <[EMAIL PROTECTED]> wrote:
> OK.  I thought pingd was used to test connectivity and could put a node
> in a degraded state.
>
> I thought suicide was valid because it was listed in the output from
> stonith -L.
>
> So, if pingd does not control killing a node, and STONITH does not
> support suicide; how does a node know to shut down when there is a
> network problem?
>
> Is the idea that if node "A" loses it's network connection, no clients
> can reach it, so it's OK for both nodes to be active?

It's not Ok to have both sides Active when you have a network problem.
And that's where STONITH comes in, but you have to have a reliable
link to a STONITH device, i.e. a device that can power off a
disconnected node.

I don't know what would be the good choice of a STONITH device for
your cases where you have node clusters located in geographically
separated places. Usually people use Heartbear for building HA
clusters on one site, not for DR.

>
> If I do give up my dreams of DR and use a cross-over, will a node fail
> automatically when it's public interface fails?
>
> Thanks,
> Chase
>
>
>>>> "Serge Dubrouski" <[EMAIL PROTECTED]> 7/15/2008 2:14 PM >>>
> On Tue, Jul 15, 2008 at 11:40 AM, Chase Simms <[EMAIL PROTECTED]>
> wrote:
>> If it is the link between locations, the server that is not located
> with
>> the 3rd party address used by pingd would no longer be able to reach
> it.
>
> pingd has nothing to do with STONITH. pingd can control where resource
> shall be started but it can't control what server has to die.
>
>>
>>
>>>>> "Serge Dubrouski" <[EMAIL PROTECTED]> 7/15/2008 11:34 AM >>>
>> On Tue, Jul 15, 2008 at 9:04 AM, Chase Simms <[EMAIL PROTECTED]>
>> wrote:
>>> I have a cluster set up and working except STONITH.  Which means
>> it's
>>> unmanageable and not fault tolerant.  I have multiple fibre
>> connections
>>> between two geographically separated locations.  I want to have one
>> node
>>> at each location for disaster recovery.  This means I cannot use a
>>> cross-over or serial connection.  I'm really hoping to parlay this
>> into
>>> a proof of concept to sell the idea of using the other site as a
> hot
>>> site for critical systems.
>>>
>>> Right now I have them set up as:
>>>
>>> server 1 vlan 7 -> server 2 vlan 7 (public interface, shared IP)
>>> server 1 vlan 8 -> server 2 vlan 8 (DRBD replication, heartbeat)
>>>
>>> How can I configure STONITH to suicide when a connection goes down?
>>
>> I don't think that this is possible. How would you know which one of
>> the servers has to suicide if your network connection dies?
>>
>>> Right now, when I take down one interface, the ping still gets
>> through
>>> the other interface.
>>>
>>> It occurred to me to create a VLAN that will not route for the
>> private
>>> interfaces.  But that will be very difficult to push through.  Is
>> there
>>> another way?
>>>
>>> Thank you,
>>> Chase
>>>
>>>
>>>
>>> The information in this email is intended for the sole use of the
>>> addressees and may be confidential and subject to protection under
>> the
>>> law. If you are not the intended recipient, you are hereby notified
>> that
>>> any distribution or copying of this email is strictly prohibited.
> If
>> you
>>> have received this message in error, please reply and delete your
>> copy.
>>>
>>> _______________________________________________
>>> Linux-HA mailing list
>>> [email protected]
>>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>>> See also: http://linux-ha.org/ReportingProblems
>>>
>>
>>
>>
>> --
>> Serge Dubrouski.
>> _______________________________________________
>> Linux-HA mailing list
>> [email protected]
>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> See also: http://linux-ha.org/ReportingProblems
>>
>>
>> The information in this email is intended for the sole use of the
>> addressees and may be confidential and subject to protection under
> the
>> law. If you are not the intended recipient, you are hereby notified
> that
>> any distribution or copying of this email is strictly prohibited. If
> you
>> have received this message in error, please reply and delete your
> copy.
>>
>> _______________________________________________
>> Linux-HA mailing list
>> [email protected]
>> http://lists.linux-ha.org/mailman/listinfo/linux-ha
>> See also: http://linux-ha.org/ReportingProblems
>>
>
>
>
> --
> Serge Dubrouski.
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
>
> The information in this email is intended for the sole use of the
> addressees and may be confidential and subject to protection under the
> law. If you are not the intended recipient, you are hereby notified that
> any distribution or copying of this email is strictly prohibited. If you
> have received this message in error, please reply and delete your copy.
>
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>



-- 
Serge Dubrouski.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to