Server Monitoring

Too often proper monitoring is overlooked as a part of critical infrastructure. Service such as Pingdom and the like are not able to tell you what is going on with the specifics of your server, they are only able to tell you if it's responding or not. While useful this information is extremely lacking, it doesn't allow an administrator to take preventative action to prevent an outage in the first place.

Full server monitoring involves installing some software on your server that reports back to a central server with the health status of the various parts of your system. These checks range from basic checks such as hard disk space usage, low ram, and service outages such as MySQL, Apache, or PHP, right through to comprehensive checks of network usage, database performance, and hardware health.

These metrics are then stored so that a history of the server's health can be reviewed allowing the administrator to make a good judgment on changes that may be needed, and where bottlenecks might be in the future as your business grows. Alarms and warnings are also generated before outages occur allowing the administrator to take action and correct the issue before there is an interruption to services.

A good example of what monitoring can be used to diagnose is poor database performance. In one case we had a client that had poor database performance during peak periods, the database was on a separate dedicated server. MySQL logs and status indicated that the database was indeed saturated, but the server load showed that it wasn't very busy. When looking at the metrics collected from the monitoring service it was very quickly determined that the network was saturated between the hosts, and after upgrading to a 10Gbit link the issue was completely resolved.