Stats Collector Plugin

Note: this component is designated "advanced", which means that objects of this type are not expected to be created or altered in most environments. If you believe that such a change is necessary, you may want to contact support in order to understand the potential impact of that change.

This plugin captures server statistics for collection by the Data Metrics Server. This is used for profiling server performance over a long period of time, and for correlation of performance anomalies across multiple servers as a function of time. The data captured by this plugin is remotely accessed and stored by the Data Metrics Server, allowing monitoring consoles or applications to view short and long term graphs of monitored data, comparing the same data on multiple servers as a function of time when investigating performance anomalies or doing capacity planning. Unless the sample-interval or collection-interval property is set too small, this has a negligible impact on performance.

Extraction of server performance data is done by two different plugins. The Periodic Stats Logger plugin writes performance data to a local comma-delimited file, which the user must capture and process. The Stats Collector plugin collects performance data that is accessible remotely via the Metrics Backend. Both plugins capture data from the same sources, although it is suggested to leave the Stats Collector plugin enabled whenever the Data Metrics Server is deployed, while the Periodic Stats Logger plugin should only be enabled if there is a need for locally captured data.
The Stats Collector keeps performance data on disk for a window of time, expecting the Data Metrics Server to remotely fetch the data before the Stats Collector deletes the locally cached data.
The Stats Collector has a fixed set of data that it can capture, where the Periodic Stats Logger can be configured to capture any data present in any entry available from the cn=monitor backend by the configuration of custom statistics.
Stats Collector data that is captured by the Data Metrics Server is stored by the Data Metrics Server such that it can be consumed by a monitoring console or application. The Data Metrics Server initially stores the raw performance data as captured by the Stats Collector, however as the performance data 'ages out' the Data Metrics Server will aggregate the more time-granular data into coarser samples. For example if the Stats Collector is sampling every second, the Data Metrics Server will store the 1-second samples, aggregate the 1-second samples to 1-minute samples after enough 1-second samples have been captured, and then eventually delete the 1-second samples that are more than 2 hours old. This continuous aggregation and pruning means the Data Metrics Server has very granular data for the recent past, and coarser data for time further in the past. The aggregation process tends to smooth short-term spikes in the data, so if a particular sample increased by 50% for 10 seconds and then returned to normal, the absolute magnitude of the increase would be lost in the first aggregation. If performance characterization requires the long-term storage of the absolute magnitude of short-duration spikes, then use of the Periodic Stats Logger is necessary.

dsconfig Usage

To list the configured Plugins:

dsconfig list-plugins
     [--property {propertyName}] ...

To view the configuration for an existing Plugin:

dsconfig get-plugin-prop
     --plugin-name {name}
     [--tab-delimited]
     [--script-friendly]
     [--property {propertyName}] ...

To update the configuration for an existing Plugin:

dsconfig set-plugin-prop
     --plugin-name {name}
     (--set|--add|--remove) {propertyName}:{propertyValue}
     [(--set|--add|--remove) {propertyName}:{propertyValue}] ...