In preparation for {T228380}, The [[ https://wikitech.wikimedia.org/wiki/SRE/Observability | SRE Observability team ]] wishes to improve overall reporting pending sources to migrate before we can comfortably turn things off without affecting any production system.
Currently tracking via overall ingest on both systems in the [[ https://grafana.wikimedia.org/d/nCxX65cSk/mediawiki-statslib-migration?orgId=1 | Graphite vs Prometheus metrics Dashboard]], we have identified the need to track **in use metrics** separately as a better indicator of when we can effectively turn Graphite read-only as an initial step towards deprecation.
For the sake of this effort, we are defining "in-use" metrics as data points emitted to graphite and captured in some dashboard or alert (making them useful). Metrics stored and un-accessed will be flagged for sunsetting and deprecation.
With this definition in mind, tracking dashboards/panels in Grafana that are backed by graphite metrics will be our key migration metric to indicate migration progress at the later stages.
The [[ https://grafana-rw.wikimedia.org/d/K6DEOo5Ik/grafana-graphite-datasource-utilization | Grafana Datasource Utilization ]] dashboard should then only count queries to the `graphite` data source. Other data sources (e.g., `graphite-synthetic-testing`) are not within the scope of graphite deprecation effort.