======Pacemaker on CentOS >6.5======
=====Installation=====
====Packages====
  yum install pacemaker pcs ccs resource-agents cman rng-tools

=====Configuration=====
====Iptables=====
<code>
-A INPUT -m state --state NEW -m udp -p udp --dport 5405 -j ACCEPT
</code>

====Random number generator====
<code - /etc/sysconfig/rngd>
EXTRAOPTIONS="-r /dev/urandom"
</code>
  chkconfig rngd on
  service rngd start

====Empty cluster=====
===Legacy CMAN stuff===
do the following on the first node:
<code>
corosync-keygen
ccs -f /etc/cluster/cluster.conf --createcluster cluster
ccs -f /etc/cluster/cluster.conf --addnode node1.mngt.bh.helux.nl
ccs -f /etc/cluster/cluster.conf --addalt node1.mngt.bh.helux.nl node1.local.bh.helux.nl
ccs -f /etc/cluster/cluster.conf --addnode node2.mngt.bh.helux.nl
ccs -f /etc/cluster/cluster.conf --addalt node2.mngt.bh.helux.nl node2.local.bh.helux.nl
ccs -f /etc/cluster/cluster.conf --addfencedev pcmk agent=fence_pcmk
ccs -f /etc/cluster/cluster.conf --addmethod pcmk-redirect node1.mngt.bh.helux.nl
ccs -f /etc/cluster/cluster.conf --addmethod pcmk-redirect node2.mngt.bh.helux.nl
ccs -f /etc/cluster/cluster.conf --addfenceinst pcmk node1.mngt.bh.helux.nl pcmk-redirect port=node1.mngt.bh.helux.nl delay=15
ccs -f /etc/cluster/cluster.conf --addfenceinst pcmk node2.mngt.bh.helux.nl pcmk-redirect port=node2.mngt.bh.helux.nl delay=15
ccs -f /etc/cluster/cluster.conf --setcman keyfile="/etc/corosync/authkey" transport="udpu" port="5405" two_node=1 expected_votes=1
ccs -f /etc/cluster/cluster.conf --settotem rrp_mode="active" 
ccs_config_validate -f /etc/cluster/cluster.conf
</code>
install basic config to other node:
  scp /etc/cluster/cluster.conf node2:/etc/cluster
  scp /etc/corosync/authkey node2:/etc/corosync
configure on all nodes
  chkconfig cman on
  chkconfig pacemaker on
  service cman start
check the rings on all nodes
  corosync-objctl | fgrep members
check secauth, rrp_mode, transport etc. on all nodes
  corosync-objctl | egrep ^totem

===Pacemaker stuff===
do the following on all nodes:
  service pacemaker start

do the following on one node:

validate everything
  crm_mon -Arf1
=====Fencing=====
====create VMware fencing====
  fence_vmware_soap -z -l root@localos -p <password> -a vcenter.mngt.rtd.helux.nl -o list |grep node
  node2.mngt,421ceeb6-b7d2-00a5-0be4-81dadde2106f
  node1.mngt,421ce817-6a23-84c9-da3a-7e9f9324ff39

  pcs property set no-quorum-policy=ignore

  pcs stonith create fence_node1 fence_vmware_soap pcmk_host_map="node1.mngt.bh.helux.nl:node1.mngt"  login="root@localos" passwd=<password> action="off" ipaddr="vcenter.mngt.rtd.helux.nl" ssl=1

  pcs stonith create fence_node2 fence_vmware_soap pcmk_host_map="node2.mngt.bh.helux.nl:node2.mngt" login="root@localos" passwd=<password> action="off" ipaddr="vcenter.mngt.rtd.helux.nl" ssl=1
=====Create cluster resources=====
====IP address====
  pcs resource create ClusterIP ocf:heartbeat:IPaddr2 ip=172.16.2.26 cidr_netmask=24 op monitor interval=30s
====LVM====
edit /etc/lvm/lvm.conf and define which volume groups to enable at startup. exclude cluster VGs. in this example the cluster VG is data.
<code - /etc/lvm/lvm.conf>
.
.  
#volume_list = [ "vg1", "vg2/lvol1", "@tag1", "@*" ]
volume_list = [ "system" ]
.
.
</code>
<code>
pcs resource create lvm_data LVM volgrpname=data exclusive=true
pcs resource create fs_lvol1 Filesystem device="/dev/data/lvol1" directory="/mnt" fstype="ext4" options="rw,noatime"
pcs constraint order lvm_data then fs_lvol1
</code>

====Application====
the following will create a linux startup script application (/etc/init.d) postfix to be a clustered application
  pcs resource create postfix lsb:postfix
put them in a resource group
  pcs resource group add postfix_group lvm_data fs_lvol1 ClusterIP postfix
=====Move cluster resource to a node=====
if you want to patch/reboot a cluster node, you have to fix the cluster resource to run on one node. in this example it is node1.

first, we check:
  pcs constraint --full
output:  
  Location Constraints:
  Ordering Constraints:
    start lvm_data then start fs_lvol1 (Mandatory) (id:order-lvm_data-fs_lvol1-mandatory)
  Colocation Constraints:

fixate resource on node1:
  pcs constraint location postfix_group prefers node1.mngt.bh.helux.nl=INFINITY

when done with work on node2 and node2 is available again we remove the fixation:
  pcs constraint remove location-postfix_group-node1.mngt.bh.helux.nl-INFINITY


{{tag>centos}}