The keys to good database monitoring are:
- Understanding what is relevant
- Eliminating unnecessary noise
We developed our proprietary monitoring platform, Monocle, because we found most commercial and open source database monitoring packages completely inadequate. Most of them consisted of a small collection of arbitrary threshold alerts or irrelevant ratio monitors that could neither reliably predict database problems nor provide meaningful, actionable alerts.
Monocle was designed to be lightweight, reliable, and quickly adaptable to changing monitoring needs. Database state data is quickly collected one time, then processed in the Monocle repository against a series of relevant metrics that account for both immediate problems and troubling trends. For example, Monocle would not alarm on a 99% full Oracle tablespace if that tablespace had been 99% full for the last 6 months; it would alarm if more space was suddenly consumed, indicating a deviation from the known trend.
By keeping monitoring relevant and alerts meaningful, Blue Gecko DBAs are not hampered by “babysitting” unnecessary alarms, or asking less experienced admins to sort through the mess and determine what may or may not be important, hoping they won’t miss something. Good monitoring means our customers’ systems are more stable and our DBAs are able to focus on what is important.