I’m a big fan of the Nagios network monitoring system and rely on it to tell me if something goes wrong with the systems for which I am responsible. I have made a large investment in time configuring Nagios to monitor exactly what I am interested in, and this effort would be wasted if Nagios detected a problem, but failed to communicate that problem to me. To make Nagios more robust, I wanted to make sure that its alerting mechanism did not depend on connections to the Internet—this would include the physical connection itself and internal and external services, such as e-mail, routing and DNS.

I have relied on e-mail-based systems in the past to deliver alerts; however, my dilemma was that if I was not getting e-mail, I did not know if this meant everything was okay or if there was some problem preventing me from getting the e-mail alerts, such as a down Internet connection or another kind of e-mail failure. I found that I became uneasy after long periods of silence and felt compelled to “poll” the system to make sure everything was okay.

On the other hand, I felt that if my alerting system was robust and I could trust it, my thinking would become “no news is good news”, and the absence of alerts would mean everything was fine.

Read Full Article from the March, 2010 Issue

Buy This Issue