[Linux-HA] Three node cluster had never worked!!!

[Abel Martín]

On 11/1/06, Alan Alexandre da Silva <alan at dextra.com.br> wrote:
> The documentation is confusing.
> Based on my informations about the servers, could you help me to generate crm and cib.xml?
> I just want three servers in the cluster with an IP Virtual and Apache. If Apache or server fail, the other nodes will take the Ip Virtual and Apache.
>
> I'm very confused.
>
> Thanks
>
> --
> Alan A. da Silva
>
>
Hi.

A long time ago I managed to configure an asymmetric cluster with
three nodes, but I lost the cib.xml config file. A good start point
would be generating a cib.xml file from your haresorce file. You can
accomplish this with the script /usr/lib/heartbeat/haresources2cib.py
(the location may vary depending on your distro). You might need to
read the documentation too, as this script doesn't work perfectly for
me now, at least with the version I'm using (Debian package
heartbeat-2 2.07).

For using the new configuration don't forget to add the line "crm
true" to your ha.cf. Start with a two-cluster node and one IP as
shared resource, to get used to the new syntax of cib.xml. And watch
out the owners of the cib.xml file, which must be appropriate in order
to make the cluster work.

One more thing: the example provided in
http://linux-ha.org/v2/Examples/Simple is not fully functional, as it
lacks an additional <rule ...> section with a different score inside
<rsc_location> in order to migrate the resource to other node in case
a failover occurs.

I hope this helps you. Good luck.