[Pacemaker] spilit brain situation

Romi Verma romi3rdfeb at gmail.com
Fri Feb 6 04:29:40 EST 2009


On Fri, Feb 6, 2009 at 2:47 PM, Dominik Klein <dk at in-telegence.net> wrote:

> Romi Verma wrote:
> > On Fri, Feb 6, 2009 at 2:28 PM, Andrew Beekhof <beekhof at gmail.com>
> wrote:
> >
> >> On Feb 6, 2009, at 9:53 AM, Romi Verma wrote:
> >>
> >> Thanks Dominic,
> >> i have two questions now.
> >>
> >> 1) what does no-quorum-policy= suicide means then??  does it remove the
> >> resource completely.
> >>
> >>
> >> no, the node kills itself and any other node in the partition
> >> this makes no sense in a 2 node cluster because both nodes will do this
> >>
> >
> > i assume the partition having less nodes will loose quorum and if
> > no-quorum-policy is set to suicide  then then they will commit suicide.
> >  like if in 3 nodes cluster , if one node looses communication to other
> > nodes then there will be two partitions . one will contain 2 nodes and
> > second will contain 1 node. partition having 2 nodes will be having
> quorum
> > and will not be affected. partition having 1 node will loose quourm and
> and
> > it will kill itself.  is my understanding is right??
>
> correct
>
> > i want the partition without quorum to reset the nodes instead of killing
> .
> > is it possible.
>
> define the difference between reset node and kill node?


 According to me  , killing a node means powering it off but reseting a node
means rebooting it.

>
>
> >> 2) why each node is thinking itsef as DC as Andrew said after spilit
> brain
> >> election happens and one node is selected as DC.
> >>
> >>
> >> no, i said after the split-brain is _repaired_ an election occurs.
> >> clearly this can't happen during a split-brian because by definition
> they
> >> can't communicate.
> >>
> >
> > ok got it  , so how do we repair this spilit brain condition . by setting
> > no-quorum-policy to reset??  or is there any other way also.
>
> stonith would reboot the node. This means, in case of a clustersoftware
> failure that led to loss of communication, the node reboots, restarts
> the cluster software and everything should be fine again.


 so you mean stonith should be used to handle spilit brain situation .  but
.. but  dont you think in two nodes cluster if we get spilit brain situation
both node will try to stonith each other.

>
>
> If there's a network problem, you would of course have to fix that ;)


:)

>
>
> Regards
> Dominik
>
> >> This is not happening in my case.
> >> i dont have any stonith configured in my cluster . do i need stonith to
> >> handle spilit brain situation.
> >>
> >>
> >>
> >> On Fri, Feb 6, 2009 at 1:59 PM, Dominik Klein <dk at in-telegence.net>
> wrote:
> >>
> >>> Romi Verma wrote:
> >>>> Thanks for fast reply ,
> >>>> Ok, Let me explain the situation. i have two nodes cluster . i pulled
> >>> out
> >>>> the network cable of one
> >>>> node which produced spilit brain situation. this time both nodes are
> >>>> thinking that other one is dead.  each node is thinking itself as DC
> and
> >>> on
> >>>> each node cluster is up and running without quorum.
> >>>>
> >>>> i am new to openais/pacemaker so dont know much but according to some
> >>>> documents it seems by default no-quorum-policy is  to "stop" the
> >>> cluster. i
> >>>> have not specified any no-quorum-policy that's why i expect that my
> >>> cluster
> >>>> should stop if it looses quorum somehow.
> >>> The "stop" refers to the resources. policy=stop on a node with no
> quorum
> >>> means: do not run any resources.
> >>>
> >>> "ignore" would mean: run resources even though we don't have quorum
> >>> (like the old heartbeat behaviour would be)
> >>>
> >>> "freeze" would mean: run and manage what you did run up to this point,
> >>> but don't aquire any other resources.
> >>>
> >>> Regards
> >>> Dominik
> >>>
> >>>> But in present spilit brain situation , on each node cluster is up and
> >>>> running without quorum.   could you please explain why this is
> >>> happening.
> >>>> Romi
> >>>>
> >>>>
> >>>> On Fri, Feb 6, 2009 at 12:52 PM, Andrew Beekhof <beekhof at gmail.com>
> >>> wrote:
> >>>>> Well the no-quorum-policy option applies during the split and an
> >>> election
> >>>>> is held to determine the DC when the partitions reform.
> >>>>> Can you be more specific please?
> >>>>> On Feb 6, 2009, at 4:54 AM, Romi Verma wrote:
> >>>>>
> >>>>>
> >>>>>
> >>>>> hi all,
> >>>>>> how does openais + pacemaker (suse 11) cluster handles spilit brain
> >>>>>> situation .  can any one explain.
> >>>>>>
> >>>>>> Thanks,
> >>>>>> Romi.
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20090206/c2e46bd3/attachment-0001.html>


More information about the Pacemaker mailing list