There are several ways to monitor the cluster. The web interface is the easiest way to get general information about cluster load, configuration, and how each node is doing. Also, significant events like nodes coming online or failing service checks will be automatically emailed to the monitoring email address specified in setup.

The monitoring email address can be found in the web interface on the Status / Monitoring and Alerts Configuration page. It will be listed under each service. To change the email address for a given service, edit the configuration file and restart the integrated Mon monitoring system:

vi /cluster/mon/mon.cf
service mon restart

To view the current status of any node in the cluster, detailed information can be found under Status / Performance Graphs. Clicking there will bring up a new web page that shows the integrated open source Ganglia monitoring system.

Ganglia collects information from every node and stores it in a round robin database, and converts this raw data into graphical format that can be displayed in a browser. Statistics will be shown for every node that include CPU, memory, and disk usage, and many other performance indicators.

2. Command Line

For detailed information, it may be best to directly read the system log on the master server. All of the compute nodes send their logs to this server, so in this one file you can see the syslog messages from every node in real time. Information will be logged here whenever nodes come on or offline, or when they are added or removed from load balancing.

Terracotta is installed by default on ClusterMaker servers, but is not necessarily used unless you want to build a java web application cluster. It can be monitored by launching the Terracotta Admin Console on the master server or on a remote workstation. After connecting to the Terracotta server instance (either on localhost or by entering the IP address of the master server), a variety of information can be checked.

The server instance itself can be monitored, as well as each node participating in the Terracotta cluster. Graphical displays are presented for many indicators, such as transactions / sec, cache hits, memory in use, etc.