[Linux-HA] ping groups and heartbeat

[Andreas Kurz]

Hello,

I solved a similiar problem by using v2 configuration style and a
third node in a different location....of course it would also be
possible, but inadvisable, to have one site with two and the other
site with one node.

This third node never runs resources and only a node in the nodegroup
that has a quorum (a group with at least two nodes) is able to run
resources.

Nodes that have no quorum stop all resources. I think in  heartbeat
2.0.8 there is a special quorum service included, so there is no need
to configure the quorum node like a "normal" node.

Regards,
Andreas

On 1/8/07, Nick Woolley <nick at kudoswebsolutions.com> wrote:
> Hi,
>
>
>
> This is a repost of a message originally sent out on the 20th December.
> There haven't been any replies yet and we were wondering if it was missed
> during the festive season.
>
>
>
> We're using heartbeat over two Ethernet links on two individual NICs. We
> can't use comm communication as the nodes are in two physically different
> locations.
>
>
>
> We have our setup working great for all failure scenarios apart from when
> both NIC links fail.  In this instance it is possible for us to get a
> split-brain between nodes.
>
>
>
> Ideally we need it so that if connectivity to the internet is broken on both
> NIC one and NIC two, heartbeat would always be secondary - no one can get to
> it, so I want to force it to become a secondary node - kind of like STONITH,
> but self inflicted.
>
>
>
> We are using heartbeat version 2.0.7 but using configuration options as
> defined for version 1.  Our configuration is shown below:
>
>
>
> logfacility local0
>
> keepalive 2
>
> warntime 10
>
> deadtime 120
>
> initdead 120
>
> ucast eth0 10.16.1.C
>
> ucast eth0 10.16.1.D
>
> ucast eth1 Y.Y.Y.P
>
> ucast eth1 X.X.X.Q
>
> node node1
>
> node node2
>
> ping Y.Y.Y.Z X.X.X.W 10.16.1.E
>
> respawn hacluster /usr/lib64/heartbeat/ipfail udpport 694 auto_failback off
>
>
>
> We are thinking that we could add a ping group of on node 1:
>
>
>
> pinggroup other_node Y.Y.Y.P 10.16.1.C Y.Y.Y.Z X.X.X.W
>
>
>
> And on node 2:
>
>
>
> pinggroup other_node Y.Y.Y.Q 10.16.1.D Y.Y.Y.Z X.X.X.W
>
>
>
> This is enough to monitor connectivity to the other nodes and the internet
> but we want this to trigger a "service heartbeat restart" rather than use
> ipfail.  Ipfail won't work as connectivity between the two nodes of the
> cluster will have been lost, and there is no-other communicable node to
> ipfail against.
>
>
>
> Is it possible to setup a pinggroup to run a different script (say one that
> sets heartbeat secondary) on failure of all nodes?  We still need ipfail to
> work for the other ping nodes.
>
>
>
> I basically need to ensure that an isolated node - one that is isolated from
> the other heartbeat node, and the both internet connections is always
> secondary.
>
>
>
> Is this possible or do I have to use some third party software to do it like
> mon?
>
>
>
> Any help would be much appreciated.
>
>
>
> Many thanks,
>
> Nick Woolley
>
>
>
>
> --
> No virus found in this outgoing message.
> Checked by AVG Free Edition.
> Version: 7.5.432 / Virus Database: 268.16.7/619 - Release Date: 07/01/2007
> 18:29
>
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>