osd/PG: discover unfound objects if PG is degraded

When a PG is remapped from OSD `a` to OSD `b`, the objects are
backfilled. When OSD `a` is restarted, objects become degraded
as `a` is no longer queried or considered as a backfill source.

The discovery is done with `PG::discover_all_missing`, which is
called from `PG::activate` (which is not called since the PG is
already active), from `PG::find_unfound` (which is not called
as there are no unfound objects). It's also called from
`PG::RecoveryState::Active::react(const ActMap&)` and
`PG::RecoveryState::Active::react(const MNotifyRec& notevt)`
but only if there are unfound objects.

This patch introduces a call to `discover_all_missing` when
the PG is degraded. The degraded objects can be restored.

Fixes: https://tracker.ceph.com/issues/37439
Signed-off-by: Jonas Jelten <jj@stusta.net>

src/osd/PG.cc

@@ -8505,7 +8505,7 @@ boost::statechart::result PG::RecoveryState::Active::react(const ActMap&)
   ldout(pg->cct, 10) << "Active: handling ActMap" << dendl;
   ceph_assert(pg->is_primary());
 
-  if (pg->have_unfound()) {
+  if (pg->have_unfound() || pg->is_degraded()) {
     // object may have become unfound
     pg->discover_all_missing(*context< RecoveryMachine >().get_query_map());
   }
@@ -8558,7 +8558,7 @@ boost::statechart::result PG::RecoveryState::Active::react(const MNotifyRec& not
 		       << dendl;
     pg->proc_replica_info(
       notevt.from, notevt.notify.info, notevt.notify.epoch_sent);
-    if (pg->have_unfound()) {
+    if (pg->have_unfound() || pg->is_degraded()) {
       pg->discover_all_missing(*context< RecoveryMachine >().get_query_map());
     }
   }