Configuring Galera arbitrator

Galera arbitrator comes with the galera cluster and it’s a daemon to avoid an split brain situation for a mysql cluster typically configured with two nodes. An split brain in clustering is a situation where one node of a cluster is not be able to communicate and synchronize with other node on the cluster, and this situation behave to lock the resources of the cluster. To avoid this garb will running in an other server, a good allocation may be the load balancing server (HaProxy recommended) and this node will connect with the group member of the galera cluster, believing that this server is a another member of the cluster with available resources.

Download and install galera binary package

– 32 bits:

# wget https://launchpad.net/codership-mysql/5.5/5.5.28-23.7/+download/mysql-server-wsrep-5.5.28-23.7-i386.deb && dpkg -i mysql-server-wsrep-5.5.28-23.7-i386.deb

– 64 bits:

# wget https://launchpad.net/codership-mysql/5.5/5.5.28-23.7/+download/mysql-server-wsrep-5.5.28-23.7-amd64.deb && dpkg -i mysql-server-wsrep-5.5.28-23.7-amd64.deb

– Running galera arbitrator daemon:

# garbd -a gcomm://192.168.1.138:4567 -g my_wsrep_cluster -l /var/log/garbd.log -d

Configuring persistent configuration for garbd

# vi /etc/default/garb

# A space-separated list of node addresses (address[:port]) in the cluster
GALERA_NODES=" 192.168.1.138:4567"
# Galera cluster name, should be the same as on the rest of the nodes.
GALERA_GROUP="my_wsrep_cluster"
# Log file for garbd. Optional, by default logs to syslog
LOG_FILE="/var/log/garbd.log"

Start garbd at boot time

# update-rc.d garb defaults

Checks
– Now the number of nodes of the cluster has to show as three:

mysql> show status like 'wsrep%';

| wsrep_cluster_size         | 3        |

 Sources

http://www.codership.com/wiki/doku.php?id=galera_arbitrator