mgr/prometheus: add missing 'deep' state to PG_STATES in ceph-mgr prometheus plugin

without this the endpoint throws 500s when any PG is doing a deep scrub.

Signed-off-by: Peter Woodman <peter@shortbus.org>
Fixes: http://tracker.ceph.com/issues/22116
(cherry picked from commit 8c93c0eeaff406af0878916ef5b9ce958896d586)

diff --git a/src/pybind/mgr/prometheus/module.py b/src/pybind/mgr/prometheus/module.py
index 80e8313d56ec889b0395eb5b402ef65cacab40b9..07f53a6373915521077686f4335029031256b6f5 100644 (file)
--- a/src/pybind/mgr/prometheus/module.py
+++ b/src/pybind/mgr/prometheus/module.py
@@ -42,7 +42,7 @@ def health_status_to_number(status):
     elif status == 'HEALTH_ERR':
         return 2
 
-PG_STATES = ['creating', 'active', 'clean', 'down', 'scrubbing', 'degraded',
+PG_STATES = ['creating', 'active', 'clean', 'down', 'scrubbing', 'deep', 'degraded',
         'inconsistent', 'peering', 'repair', 'recovering', 'forced-recovery',
         'backfill', 'forced-backfill', 'wait-backfill', 'backfill-toofull',
         'incomplete', 'stale', 'remapped', 'undersized', 'peered']
@@ -266,12 +266,15 @@ class Module(MgrModule):
             try:
                 self.metrics[path].set(value)
             except KeyError:
-                self.log.warn('Skipping unknown PG State {}'.format(state))
+                self.log.warn("skipping pg in unknown state {}".format(state))
         reported_states = [s[0] for s in reported_pg_s]
         for state in PG_STATES:
             path = 'pg_{}'.format(state)
             if state not in reported_states:
-                self.metrics[path].set(0)
+                try:
+                    self.metrics[path].set(0)
+                except KeyError:
+                    self.log.warn("skipping pg in unknown state {}".format(state))
 
     def get_metadata_and_osd_status(self):
         osd_map = self.get('osd_map')