Ceph » mgr

The mgr/prometheus module does not provide a granular view of healthchecks to Prometheus, which means some alerts rely on the generic ceph_health_status > 0 expression. This is not very informative and common source of frustration.

This feature provides a metric per encountered healthcheck, so alert rules can be customised to specific events. In addition since the healthchecks need to be emitted on each scrape, the module needs to persist healthcheck state, which opens the door to providing a healthcheck history. The history would be exposed by a new command, allowing the admin the ability to see what healthchecks have been encountered within the cluster, their frequency and the first and last seen timestamps.

The feature deliverables should include
- additional metrics
- updated prometheus rules
- updated docs for the prometheus module