I’ve found a lot of articles online about how to set up simple two node ‘example’ clusters using Corosync and Pacemaker but very little documentation about how to go beyond that, so when I finally got it working, I thought I’d share. This article assumes all of the nodes are on the same subnet, going beyond that would require routing rules and single points of failure, an issue i will address in a future article. It also assumes the virtual IP you are using is also on the same subnet. This setup is active/passive meaning that only one of the nodes claims the virtual IP at a time. When one of the nodes goes down, the others make a decision about which should get it next. There is a very brief outage between the time a node goes down and the time the virtual IP is reassigned to another node, this can be helped by tweaking timeout settings, but that’s an article for another day.  3 of the 4 nodes can go down and the virtual IP will still be accessible on the 4th.

To differentiate between commands that shoud be run on all four nodes and commands that should only be run on one of them, I will precede the commands with ‘%%%%’ and ‘%’ respectively.

For more information about Pacemaker and Corosync, I recommend the free Clusters from Scratch book, available here: http://clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Clusters_from_Scratch/

1. Install Pacemaker and the PCS daemon on all of the nodes

%%%% yum install pacemaker pcs

2. Start the PCS daemon and set it to run on boot

%%%% systemctl start pcsd
%%%% systemctl enable pcsd

This will create a hacluster user which we need to set a password for as it will be used to authenticate the nodes in the cluster

3. Set a password for the hacluster user (the same password on all four nodes)

%%%% passwd hacluster

4. On one of the nodes, set up authentication between the cluster nodes

5. On one of the nodes, run a setup command that will create a basic config file(/etc/corosync/corosync.conf) on all four nodes

6. Add needed information to the config file.

On each of the nodes, you will need to rewrite the /etc/corosync/corosync.conf file and add needed configurations. Follow the example file below.

totem { version: 2 secauth: off cluster_name: CLUSTER_NAME # Use the value from step 5 transport: udpu interface { ringnumber: 0 bindnetaddr: MY_IP_ADDRESS # Replace this with the IP address of the local machine broadcast: yes mcastport: 5405 } } nodelist { # Replace all four IP addresses here node { ring0_address: NODE1_IP_ADDRESS nodeid: 1 } node { ring0_address: NODE2_IP_ADDRESS nodeid: 2 } node { ring0_address: NODE3_IP_ADDRESS nodeid: 3 } node { ring0_address: NODE4_IP_ADDRESS nodeid: 4 } }  quorum { provider: corosync_votequorum wait_for_all: 1 }  logging { to_logfile: yes logfile: /var/log/cluster/corosync.log to_syslog: yes } 

7. On one node, start the cluster

% pcs cluster start –all

8. On one node, create the virtual IP

This assumes that VIRTUAL_IP is on the same subnet as the other 4 IP addresses, CIDR_SUBNET_MASK should be the CIDR notation form of the same subnet mask used by the other nodes (so if your subnet mask is 255.255.255.0, use CIDR_SUBNET MASK is 24)

% sudo pcs resource create Cluster_VIP ocf:heartbeat:IPaddr2 ip=VIRTUAL_IP cidr_netmask=CIDR_SUBNET_MASK op monitor interval=20s

Now we can check the status of the cluster using

pcs status

You should now be able to ping your virtual IP from any of the nodes.