[Linux-HA] pingd failover in active/standby cluster - SOLVED

[Andrew Beekhof]

On 10/4/07, Matt Zagrabelny <mzagrabe at d.umn.edu> wrote:
> On Wed, 2007-10-03 at 14:37 -0500, Matt Zagrabelny wrote:
> > Update:
> >
> > I ran ptest on 'cody', the primary node. It looks like it is computing
> > the values incorrectly for pingd. The multiplier seems to be only "1",
> > when in fact in the cib.xml file it is "100". Thus I am getting the
> > following comparison in the output of the ptest command:
> >
> > debug: native_assign_node: Color external_VIP, Node[0] cody: 51
> > debug: native_assign_node: Color external_VIP, Node[1] tim: 2
> >
> > which is wrong, it should be:
> >
> > debug: native_assign_node: Color external_VIP, Node[0] cody: 150
> > debug: native_assign_node: Color external_VIP, Node[1] tim: 200
>
> The problem was that I had the pingd multiplier in the wrong attributes
> section. It was in the clone's attribute section, not the pingd
> primitive's section.
>
>
> > and 'tim' should then get the resources. Also, shouldn't it be computing
> > the values for the group which is "monolith_resources" not
> > "external_VIP"?

sort of but not really.
the preferences of other resources in the group are taken into account
when placing the first resource of the group.

have a look at the colocation docs at the link below:
   oss.beekhof.net/~beekhof/heartbeat/docs/

> Though I still don't understand the following output. Any explanations
> would be great.
>
> > With regard to the last question, I do not understand why the following
> > resource (internal_VIP) is getting different values from above (51 and
> > 2), it is in the same resource group (monolith_resources) as
> > external_VIP:
> >
> > debug: native_assign_node: Color internal_VIP, Node[0] cody: 1000000
> > debug: native_assign_node: Color internal_VIP, Node[1] tim: -1000000
> >