Hi,
we are migrating from heartbeat v1 to v2 and also want to use pingd
instead of ipfail. Our two cluster nodes are connected to redundant
networks (two switches etc.) to communicate to the rest of the world.
With v1 ipfail compared the number of successful pings of both cluster
nodes to let the resources run on the node with best network quality. If
one switch was faulty, both nodes had less, but equal pings and the
resources were kept just where they are. In v2 pingd is running and
configured as follows:
...
<nvpair id="nvpair-63c37f35-716a-4541-ac79-aca7252d27d7"
name="default-resource-stickiness" value="INFINITY"/>
...
<rsc_location id="rsc_location_MyRcs" node="node0" rsc="MyRcs"
score="1"/>
<rsc_location rsc="MyRcs" id="rsc_location_MyRcs_pingd">
<rule boolean-op="or" id="rsc_location_MyRcs_pingd-rule"
score-attribute="-INFINITY">
<expression attribute="pingd"
id="rsc_location_MyRcs_pingd-expression-0" operation="not_defined"/>
<expression attribute="pingd"
id="rsc_location_MyRcs_pingd-expression-1" operation="lt" value="2"/>
</rule>
</rsc_location>
If one node has less connectivity the resource switch to the other. But
if one of the switches is faulty, none of my cluster nodes will take the
resources. (Maybe because the score is -INFINITY for both??)
Has anyone an idea how to configure crm to let the resources stay as in
v1? Any hint is welcome.
Regards,
Patrick
P.S.: I've found some links to www.linux-ha.org/pingd but the page is
not found?!?
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
