Initial Metrics for Device Services
The device service SDKs shall collect and report the following metrics
1) Metrics defined for all EdgeX services in ADR #0006
Service uptime
Cumulative API requests succeeded / failed / invalid
Average response time for API requests
Average and Maximum request size
Note: The API requests of interest are the device accesses. For Device service purposes "request size" is to be read as "event size"
2) Metrics defined for Device Services in ADR #0006
Number of devices managed by this service
Device requests
Note: For "Device requests" we shall provide:
total number of events generated
total number of readings generated
number of events generated in the last sampling period (configurable with a default of 1 minute)
data volume of events generated in the last sampling period.
3) Metrics which will replace the lastConnected field in v3
Last successful read or write access (per device)
In addition the SDKs shall provide a facility for the device service implementations to collect and report custom metrics for that service.
Comments
Iain's idea of handling the metrics for the REST request in the REST handler middleware is something that we should look into. Might be a bit tricky, but would be nice to have is a central reusable code.
Per WG meeting 6/6/22:
Service uptime → better handled by querying the environment eg docker, ps etc
Number of devices managed → can be obtained from core-metadata
Requests success/fail/invalid → count fail+invalid only
Note the sampling to get number of events generated or data volume can be derived from the raw data be the receiver.