[Pacemaker] Help on defining resource-stickness

Andrew Beekhof andrew at beekhof.net
Tue Mar 12 01:08:19 UTC 2013


On Mon, Mar 11, 2013 at 5:01 PM, Paul Sun <paul.sun at syniverse.com> wrote:
> Hi
>
>
>
> I am currently learning how to use pacemaker. I have created a three nodes
> configuration, and two IP resources. My target to setup the IP to follow
> only two nodes, and this is able to achieve by using the score in location
> command. My next target is to disable the failover back when node is
> resumed. But I am confused how I can define the resource-stickness
> accordingly to different resources.
>
>
>
> Below is my configuration.
>
>
>
> crm(live)configure# show
>
> node HA_NODE1
>
> node HA_NODE2
>
> node HA_NODE3
>
> primitive ClusterIP ocf:heartbeat:IPaddr2 \
>
>       params ip="192.168.206.161" cidr_netmask="32" \
>
>       op monitor interval="30s"
>
> primitive PDNS_IP ocf:heartbeat:IPaddr2 \
>
>       params ip="192.168.206.160" cidr_netmask="32" nic="eth2"
>
> location ClusterIP-prefer1 ClusterIP 500: HA_NODE1
>
> location ClusterIP-prefer2 ClusterIP 300: HA_NODE2
>
> location ClusterIP-prefer3 ClusterIP -inf: HA_NODE3
>
> location PDNS_IP-prefer1 PDNS_IP -inf: HA_NODE1
>
> location PDNS_IP-prefer2 PDNS_IP 500: HA_NODE2
>
> location PDNS_IP-prefer3 PDNS_IP 300: HA_NODE3
>
> property $id="cib-bootstrap-options" \
>
>       dc-version="1.1.7-6.el6-148fccfd5985c5590cc601123c6c16e966b85d14" \
>
>       cluster-infrastructure="openais" \
>
>       expected-quorum-votes="3" \
>
>       stonith-enabled="false" \
>
>       no-quorum-policy="ignore" \
>
>       last-lrm-refresh="1362737116"
>
> rsc_defaults $id="rsc-options" \
>
>       resource-stickiness="150"
>
>
>
> Target is to
>
> 1.       ClusterIP will flow between HA_NODE1 and HA_NODE2 only.
>
> 2.       PDNS_IP will flow between HA_NODE2 and HA_NODE3 only.
>
>
>
> Question:
>
> How I can setup the resources “CLUSTERIP” not failover back to HA_NODE1 when
> it is resumed, and resources “PDNS_IP” failover back to HA_NODE2 when it is
> resumed?

Set the default resource-stickiness to 0.
Then, for CLUSTERIP set the resource-stickiness meta attribute to 201
(1 +  ClusterIP-prefer1.score - ClusterIP-prefer2.score).




More information about the Pacemaker mailing list