Real World Computing

Once you are happy with all these features of Nagios, there are two other feature sets to look at. First, there is the ability to escalate a service problem that is not being handled. For example, you might decide that if a service has been issuing warning events for a number of hours and whoever it told is not fixing it, you need to tell someone else. Service escalations let you do this and can be set up with various thresholds. Also, if your network is complex enough, you are able to set up 'service dependencies', which are useful when one service depends on others and one of those fails. Service dependencies enable Nagios to decide what the root cause of a problem is and report on that, rather than on the symptoms it may cause elsewhere.

One last thing you can do is produce pretty graphs of your monitored variables. Nagios does not do this itself, but most Nagios plug-ins provide performance information that Nagios passes on to another piece of software. There are various packages to do this, but we use PerfParse, which enables you to call up graphs of particular variables and to watch how a service performs over time. This is not only important for diagnosing problems, but also lets you fine-tune the thresholds on warning and critical events.

We first got Nagios running about a month ago, and now use it to monitor all our Unix and Windows servers and networking equipment. We also use it to check services at our clients' premises and whether our clients are seeing our services correctly. We have set it up to employ instant messaging via AIM to let us know of any problems and, of course, I can check the services on my mobile phone from the pub.

Find out more at the main Nagios site (www.nagios.org) and get the Nagios plug-ins at nagiosplug.sourceforge.net. The community website is at www.nagiosexchange.org and the PerfParse graphing package at perfparse.sourceforge.net