Summary:

We have different components involved in this so need to explore each of those for  the implementation, as such:-

• Each of the service for which we have to explore metric collection runs differently so have to be checked if those should have common or different mechanisms for metrics collection, like
  • ovn-controller -> as daemonset
  • ovs-vswitchd/ovsdb-server ->as daemonset
  • ovn nb and sb db server -> as statefulset
  • ovn northd -> as deployment
• For OpenShift Cluster as also running with OVN-Kubernetes as networking backend, there are some metrics being already collected for those. Different prometheus instance i.e not managed by telemetry operator being used in this case.
• Rabbitmq/ceilometer/alertmanager/kube-state running on Control plane and for these metrics are also being collected by prometheus instance managed by telemetry-operator. Have to be checked on how these being managed.

Goal:

• To finalize:-
  • what will be implementation method for all of these OVN/OVS services
  • What all metrics have to be collected for each of these services
  • If some parts of OVN/OVS metrics collector from OCP can be reused
  • To check possibility to use openstack-network-exporter also for collecting Metrics from control plane side or a new collector/repo have to be created.
  • Whether to enable/disable by default these collectors and consider initial operator update impact on already running workload
  • Integration with telemetry operator

TimeBox:

• 1 Week

Deliverables/Outcomes:

• Outline of the implementation details, previous attempt can be used as a baseline for this analysis https://docs.google.com/document/d/1W-JVIpO-4XuRpBKehYX6SeBh00qWvrAYViC5uO7WJk0