Skip to content

Metric segmentation & access by project

OSM Metric segmentation & access by project

Proposers

  • Gianpietro Lavado (Whitestack)
  • Gerardo Garcia (Telefonica)
  • Francisco Javier Ramon (Telefonica)

Type

Feature

Target MDG/TF

SA

Description

Metrics in Prometheus are stored in key-value pairs, grouped per name and differentiated by labels. Current design keeps metrics grouped by name (i.e. cpu_usage) and allows to differentiate based on label filtering (i.e. ns_id), so for example, a Grafana dashboard can be created for a single NS ID, filtering by the metric's label "NS_ID".

Still, current design has some limitations:

  • VIM metrics have different names by default and refer to the same resource for any NS/VNF, but VNF Metrics (Indicators) using the same name, could have different semantics depending on the VNF, or worse, different types (gauge and counter for example), and may not be supported.
  • There is no role-based authentication or tenant awareness. Prometheus has some limitations, but Grafana may solve some of ths issues at a higher level by automating the creation of dashboards.

This feature presents an opportunity to revisit metric segmentation, authentication, and dashboard automation.

Demo or definition of done

OSM users should be able to see the metrics beloging to their own NS, automatically, in Grafana.