Ceph » mgr

The ceph-mgr influx module, as with all the metric related modules, calls get_all_perf_counters() to collect all daemon stats but it appears that the cache tiering related stats are not in the list? From what I can tell in the code they are all set under PRIO_DEBUG instead of PRIO_USEFUL (https://github.com/ceph/ceph/blob/mimic/src/osd/OSD.cc#L3055).

I would argue that the cache tier stats should be considered 'useful' rather than 'debug'. For example, they show up with other already 'useful' stats under ceph -s so it seems like it shouldn't be any different as far as the mgr stat collection goes. Personally I'd rather not have to start collecting all the other debug prio stats just to get cache tier stats.

Example 'ceph -s':
cache: 794 MiB/s flush, 495 MiB/s evict, 163 op/s promote

It seems like we'd want all the stats prefixed tier_ or osd_tier_ to be able to accurately monitor cache tier performance.