Ceph » mgr » Dashboard

Each ceph daemon exposes a collection of performance counters. Currently if someone wants to inspect the counters, they need to use the CLI "ceph daemon <type>.<id> perf schema" on a daemon (for example "ceph daemon mon.a perf schema"), and then the similar "perf dump" command to get the actual values.

Fortunately, ceph-mgr has an internal copy of all the counters for all the services, and their schema. This is available to python modules via MgrModule.get_perf_schema (get all the perf counter metadata for a service), and MgrModule.get_counter (to get a single counter).

We should have a dashboard page with a URL like "/perf_counters/<service type>/<service id>", which shows the performance counters, including their names, description strings, and current values. We can link to it from other places where the services are shown in the UI, e.g. to each appropriate /perf_counters/osd/<id> from the list of OSDs.

This page will not replace existing detail views of services, but will provide an extra detailed view for engineers who want to explore what performance data is available for a particular service.