Monitoring of NFVI-level VNF metrics from Prometheus TSDB

Proposers

• Alejandro García (Whitestack)
• Gianpietro Lavado (Whitestack)

Target MDG/TF

MON, Common

Description

Current OpenStack telemetry system does not scale for large environments, so some VIMs are already moving away from Gnocchi to Prometheus as TSDB. The proposal is to add a new collector or extend the existing OpenStack collector, in order to enable metrics collection from Prometheus. This requires the addition/modification of the existing collector, as well as a new metric mapping to be able to keep the same naming conventions used today at OSM. There is also an opportunity to migrate away from a periodic descriptor lookup to a notification-based approach, when looking for descriptors specifying metrics collection at MongoDB.

More details at this presentation.

Demo or definition of done

Adding a VIM that does not have Gnocchi but a Prometheus TSDB for storing VM metrics, should allow for the same functionality that OSM has today with Gnocchi collection, including the storage of those metrics in OSM's Prometheus under the same names, as well as the autoscaling and other capabilities. Robot tests SA-01, SA-02 and SA-07 must pass with a VIM that uses Prometheus as Telemetry system.