Web Monitoring
There are quite a few free monitoring software packages available for networks and clusters. Most of these require a web server to be installed. Often this is run on the head node, and then IPTables can be changed to forward web traffic to that node, just like SSH traffic is forwarded to it.
On my cluster, I run both Nagios and Ganglia. Both of these require a web server to run on. There are walkthroughs for installing and configuring both of these.
Web Server
Setting up a web server is relatively easy with Debian. There are a few optional things that can then be done to secure the web server.
Nagios
Nagios is often used to monitor whole networks, not just clusters. It is typically used to monitor services. For instance, it comes with options to monitor web servers, SSH servers, and do ping tests to make sure hosts are alive, amongst other options. These checks are typically run from one Nagios host, with no installation necessary on the other hosts.
- Nagios
- Nagios Installation and Configuration
- Creating Your Own Nagios Plugin
- Nagios NRPE Addon Installation and Configuration - a plugin used to execute remote commands
Ganglia
Ganglia needs to be installed on each machine to be monitored. One node then runs a meta daemon and collects the information from all of the clients. Information provided by Ganglia includes cluster load, node load, memory use, and network traffic. More information about individual hosts is also available.
