Ceph » mgr

The Prometheus exporter module (ceph: src/pybind/mgr/prometheus/module.py) has a bug where the list of "known" placement group states (PG_STATES) has values that do not entirely match what comes from Ceph.

This may be because the values match an older version of Ceph - I currently don't have visibility of this.

I have been fixing them in a local version to improve monitoring of my cluster, and have thus far have made the following changes:

1. Original, broken PG_STATES
   #PG_STATES = ['creating', 'active', 'clean', 'down', 'scrubbing', 'degraded',
2. 'inconsistent', 'peering', 'repair', 'recovering', 'forced-recovery',
3. 'backfill', 'forced-backfill', 'wait-backfill', 'backfill-toofull',
4. 'incomplete', 'stale', 'remapped', 'undersized', 'peered']

1. PG_STATES, adjusted as new issues are found
   PG_STATES = ['creating', 'active', 'clean', 'down', 'scrubbing', 'degraded',
   'inconsistent', 'peering', 'repair', 'recovering',
   'incomplete', 'stale', 'remapped', 'undersized', 'peered',
   1. The following states are incorrectly named in the original list
      'forced_recovery',
      'backfilling', 'forced_backfill', 'backfill_wait', 'backfill_toofull',
   1. the following states are missing, but documented on
   2. http://docs.ceph.com/docs/master/rados/operations/pg-states/
      'deep',
   1. The following states are undefined on
   2. http://docs.ceph.com/docs/master/rados/operations/pg-states/
      'recovery_wait'
      ]