Behaviour

• Stores each computed KPI value into the datastore as a timestamped row, tagged so it can be found and read back later.
• Commits queued rows on a configurable flush interval (or as the Metric calculations drive it) to keep logger traffic low.
• Serves the stored KPI values back over a logger data server.
• Reports a backend write failure for the affected Metric only, so one Metric can flag a problem while the others keep progressing.

Configuration

In normal use the sink is added to a MetricsManager through the MetricLoggerDataSink recipe, a LoggerDataSink pre-configured with a local compact datastore for storing the MetricsManager's KPI values. The settings are:

• Datastore - the database backend the KPI values are stored in; the recipe uses a local CDPCompactDatastore.
• FlushIntervalSec - how often queued rows are committed to the datastore; 0 commits as the Metric calculations drive it.
• ServerInterface, ServerPort - the network endpoint on which the stored KPI values are served for reading back.
• Debug - diagnostic verbosity; higher values log more detail about registrations, writes, and flushes.
• ServerMetadataCacheDurationSec, TotalFramesWritten, TotalSamplesWritten - read-only server cache duration and write-throughput counters.