Whether to capture points before and after event

time-series
influxdb
#1

Is point overloading* recommended for capturing metrics before and after a particular event?

* I say point overloading in referring to how InfluxDB handles writing duplicate points since it is a mechanism to add new fields to an existing point.

Let me paint a picture:

  • Application with asynchronous queue processing (e.g. Kafka)
  • Want metrics around events picked up for processing and events having been processed
  • Measurement would be concrete class with single purpose containing multiple metrics (e.g. latency, number of records, number successfully processed, number failed, etc)

I realize this would require emitting multiple points for the same measurement + tag set + timestamp. The only other ways I can see this information being captured is 1) multiple measurements or 2) tag with multiple values.

Thoughts?