To support the Unified Observability next-generation initiative, statistics and events need to be exposed from OVN to OpenShift Observability, which is based on prometheus for data collection, so this should take the form of a Prometheus exporter.

The focus of this Epic is to implement a basic Prometheus exporter for OVS/OVN and integrate it with RHOSO. Metrics collection for each of the individual metrics categories and metrics identification are defined in separate epics.

Tasks are:

1.) Partner with the Fast Data Path team to incorporate their input on how to monitor OVN.
2.) Evaluate the currently available Prometheus exporter(s) [1] and determine their suitability for productization within OpenStack. If not, then determine an alternative strategy.
3.) Verify the validity of the metrics/events produced by the exporter against a test cloud. Include OVN Octavia load balancers, DPDK, and any other potentially unusual circumstances.
4.) Create a list of data from the OVN exporter that could be used to answer the question "When a customer says they are broken, what do I need to look at in order to debug?"
4a.) Virtual switch stats like packet loss, latency, etc.
4b.) What events to store - for example, port creation/minute may be needed in a super-high-port-count situation (i.e. Services Australia)
4c.) What logs to store?

Note: Podified control plane is out of scope as a consideration for this.

https://docs.google.com/document/d/1ZGqQtMkXOh2m8JcxPaiW39TCR3Ekld-n8t7plkXQqeU/edit

[1] https://github.com/greenpau/ovn_exporter/ is one example but there may be more.