[Pacemaker] Andrew and Lars please confirm this.

Andrew Beekhof beekhof at gmail.com
Tue Mar 3 12:50:33 UTC 2009


On Tue, Mar 3, 2009 at 06:32, Romi Verma <romi3rdfeb at gmail.com> wrote:
>
>
> On Mon, Mar 2, 2009 at 1:22 PM, Romi Verma <romi3rdfeb at gmail.com> wrote:
>>
>> ok, so let me summarize what i understood.
>> we can form a openais/pacemaker cluster whose nodes span across multiple
>> sites. these sites can be as close as just adjacent rooms or they can be
>> across continent  also.  only problem is in case of communication failure ,
>> openais-pacemaker does not have any solution yet.

Not quite.

Comms failures lead to split-brains (regardless of where the servers
are or even whether openais or heartbeat is used).

Pacemaker has various options for how it behaves during a split-brain, but
1) Its more desirable to avoid them in the first place
2) None of the strategies are (yet) really suited to failing over from
one physical location to another


>> That's why it is
>> recommend to have a local cluster  ( nodes sharing same rack).
>>
>> am i right??
>>
>> Thanks & Regards,
>>
>> On Mon, Mar 2, 2009 at 12:47 PM, Andrew Beekhof <beekhof at gmail.com> wrote:
>>>
>>> On Mon, Mar 2, 2009 at 04:27, Romi Verma <romi3rdfeb at gmail.com> wrote:
>>> > Thanks Andrew and Lars,
>>> > I would like  to understand it bit more. , is there any distance
>>> > parameter
>>> > here? i mean to say suppose i have two nodes in one room and 1 node in
>>> > adjacent room  , cant i form three nodes openais/pacemaker cluster
>>> > here.
>>>
>>> of course.
>>> i believe the point is that as the distance increases, so does the
>>> number of switches/telcos involved and therefor the probability of a
>>> comms failure is between the two sites is quite high.
>>> as is the latency...
>>>
>>> >  i was under impression that as far as multicast ip works , nothing
>>> > stops us
>>> > in forming a cluster.
>>> >
>>> > Regrards,
>>> >
>>> >
>>> > On Sat, Feb 28, 2009 at 12:04 AM, Lars Marowsky-Bree <lmb at suse.de>
>>> > wrote:
>>> >>
>>> >> On 2009-02-27T13:38:02, Romi Verma <romi3rdfeb at gmail.com> wrote:
>>> >>
>>> >> > Thanks for reply Andrew,
>>> >> > ok, then how openais-pacemaker cluster handles the cluster whose
>>> >> > nodes
>>> >> > span
>>> >> > across multiple sites.
>>> >>
>>> >> This is not currently handled.
>>> >>
>>> >> But deploying a split-site setup with quorumd never worked either; it
>>> >> is
>>> >> not integrated with fencing, and it is impossible to build a working
>>> >> split-site cluster with it. This was, I'm afraid to say, only a 40%
>>> >> implementation of the solution even in heartbeat/CCM land.
>>> >>
>>> >>
>>> >> Having proper split-site support is on the roadmap, as soon as the
>>> >> openAIS integration stabilizes.
>>> >>
>>> >>
>>> >> Regards,
>>> >>    Lars
>>> >>
>>> >> --
>>> >> Teamlead Kernel, SuSE Labs, Research and Development
>>> >> SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
>>> >> "Experience is the name everyone gives to their mistakes." -- Oscar
>>> >> Wilde
>>> >>
>>> >>
>>> >> _______________________________________________
>>> >> Pacemaker mailing list
>>> >> Pacemaker at oss.clusterlabs.org
>>> >> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>> >
>>> >
>>> > _______________________________________________
>>> > Pacemaker mailing list
>>> > Pacemaker at oss.clusterlabs.org
>>> > http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>> >
>>> >
>>>
>>> _______________________________________________
>>> Pacemaker mailing list
>>> Pacemaker at oss.clusterlabs.org
>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>
>
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
>




More information about the Pacemaker mailing list