Fetchlets

A managed resource can expose performance indicators through various interfaces like JMX, SNMP, and so on. For example, WebLogic Server performance indicators are exposed through a JMX interface. Enterprise Manager provides executables that can access performance data from most of the standard interfaces. These executables are called fetchlets. There is one fetchlet for each data access mechanism. Some useful fetchlets are SNMP, JMX, and URLXML fetchlets, which can be used to access data from SNMP interface; JMX interface, and URL interfaces respectively.

For a complete list of available fetchlets, please refer to the Enterprise Manager documentation.

Metrics definition

Metrics is the performance data that is collected from each target at fixed intervals and can also be collected on demand. The metrics definition for a target type defines all performance indicators that need to be collected and what fetchlet should be used to collect those indicators. This is defined in a template file. This template file is very important and we'll be referring to this file by the term target metadata.

For a given target type, the collection frequency of the metrics is defined in another template and we refer to that as a target collection file.

Metric collection and aggregation

To monitor performance indicators of a given target, metrics are collected by an agent, as per the definition in the target metadata file, and are collected as per the frequency defined in the target collection file.

After collection, the data is kept at a staging location; from there it is uploaded into the repository at fixed intervals. All performance metric data is persisted in the repository and can be used to analyze a system's performance over a period of time. For coarse views, these metrics are rolled up every hour and every 24 hours. Data aggregated every hour is kept for 31 days and data aggregated for 24 hours is kept for 365 days. Raw data received from the agent is only kept for seven days. Data purge policies can be configured to suit user needs.

The following screenshot shows one such example, detailing the average execution for one of the web services deployed on a WebLogic Server. From this view, the user can see real-time data or historical data.

Metric alerts

Enterprise Manager can use metric data that is persisted in the repository in two ways. Firstly, the user can see and analyze data over a period of time. Secondly, any time a particular data value crosses acceptable limits; Enterprise Manager can keep track of such instances.

The user can define thresholds to in turn define the acceptable limits of performance metrics. Enterprise Manager supports two levels of thresholds for the metrics — warning and critical threshold. For example, a CPU % usage metric can have 70% as warning and 90% as critical threshold. Enterprise Manager provides default thresholds for many metrics that the user can customize. For metrics where no thresholds are defined, the users can define them.

Whenever a metric data value crosses the threshold, an alert is generated and stored in the Management repository. If a warning threshold is violated; a warning alert is generated, and if a critical threshold is violated; a critical alert is generated. These alerts can be used for generating some form of notification.

Monitoring templates

Monitoring template is a construction, using which, one can define the thresholds for a target type, and can apply this template to multiple targets of the same type. When there are multiple targets of same type, setting thresholds for all such targets can be a tedious task. Monitoring templates can make it simpler by providing one central place to define thresholds for all targets of the same type.

System Administrator can also define default-monitoring template for a given target type. Once a template is marked as default for a target type, all new targets of the same type will automatically get the same thresholds as defined in the template.