The Grafana dashboards we're using to replace Ganglia need some usability / visibility improvements, e.g.
- The main per-datacenter dashboard with per-cluster breakdown should be tagged featured
- The cluster breakdown dashboard should have graphs of the selected view (e.g. memory/cpu) for all hosts in the cluster
- Ideally related breakdowns (memory/load) should have the same scale to do easy comparisons
- From a "overview" dashboard it should be possible to select a specific cluster and go to its "drilldown" dashboard easily
- The Prometheus-related dashboard don't really need to have "prometheus" in the name. This in itself is easy to do, though it changes the dashboard URL and therefore breaks existing links. See also Grafana issue #7043 for dashboard redirects.