osd/PG: normalize query processing in Stray and ReplicaActive

A stray PG may end up in ReplicaActive if it is participating in backfill.
However, whether it is or isn't, we should treat queries the same.
Otherwise we end up with weird behaviors like:

 - osd's stray pg moves to ReplicaActive (gets info+log from primary)
 - osd goes down and back up
 - primary restarts peering, request FULLLOG to find missing objects
 - osd ignores FULLLOG because it is ReplicaActive and not Stray

Fixes: http://tracker.ceph.com/issues/24373
Reported-by: Kouya Shimura <kouya@jp.fujitsu.com>
Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit 4e14a380becd61d823a1c86e2dbb41dc8e3a5834)

diff --git a/src/osd/PG.cc b/src/osd/PG.cc
index d6f8c4b7dc4de5ab84f002f5cc33f3ebbee4a3a3..967884be0068e845a2084c0fc31a96a018dffd93 100644 (file)
--- a/src/osd/PG.cc
+++ b/src/osd/PG.cc
@@ -5469,6 +5469,26 @@ void PG::fulfill_log(
   osd->send_message_osd_cluster(mlog, con.get());
 }
 
+void PG::fulfill_query(const MQuery& query, RecoveryCtx *rctx)
+{
+  if (query.query.type == pg_query_t::INFO) {
+    pair<pg_shard_t, pg_info_t> notify_info;
+    update_history(query.query.history);
+    fulfill_info(query.from, query.query, notify_info);
+    rctx->send_notify(
+      notify_info.first,
+      pg_notify_t(
+       notify_info.first.shard, pg_whoami.shard,
+       query.query_epoch,
+       get_osdmap()->get_epoch(),
+       notify_info.second),
+      past_intervals);
+  } else {
+    update_history(query.query.history);
+    fulfill_log(query.from, query.query, query.query_epoch);
+  }
+}
+
 void PG::check_full_transition(OSDMapRef lastmap, OSDMapRef osdmap)
 {
   bool changed = false;
@@ -7802,13 +7822,11 @@ boost::statechart::result PG::RecoveryState::ReplicaActive::react(const ActMap&)
   return discard_event();
 }
 
-boost::statechart::result PG::RecoveryState::ReplicaActive::react(const MQuery& query)
+boost::statechart::result PG::RecoveryState::ReplicaActive::react(
+  const MQuery& query)
 {
   PG *pg = context< RecoveryMachine >().pg;
-  if (query.query.type == pg_query_t::MISSING) {
-    pg->update_history(query.query.history);
-    pg->fulfill_log(query.from, query.query, query.query_epoch);
-  } // else: from prior to activation, safe to ignore
+  pg->fulfill_query(query, context<RecoveryMachine>().get_recovery_ctx());
   return discard_event();
 }
 
@@ -7899,21 +7917,7 @@ boost::statechart::result PG::RecoveryState::Stray::react(const MInfoRec& infoev
 boost::statechart::result PG::RecoveryState::Stray::react(const MQuery& query)
 {
   PG *pg = context< RecoveryMachine >().pg;
-  if (query.query.type == pg_query_t::INFO) {
-    pair<pg_shard_t, pg_info_t> notify_info;
-    pg->update_history(query.query.history);
-    pg->fulfill_info(query.from, query.query, notify_info);
-    context< RecoveryMachine >().send_notify(
-      notify_info.first,
-      pg_notify_t(
-       notify_info.first.shard, pg->pg_whoami.shard,
-       query.query_epoch,
-       pg->get_osdmap()->get_epoch(),
-       notify_info.second),
-      pg->past_intervals);
-  } else {
-    pg->fulfill_log(query.from, query.query, query.query_epoch);
-  }
+  pg->fulfill_query(query, context<RecoveryMachine>().get_recovery_ctx());
   return discard_event();
 }
 

diff --git a/src/osd/PG.h b/src/osd/PG.h
index e042a8afd0a0fed901a5379477fe6866253bdb09..e8acafae53cc00f21b78475be2d3a049fcbb0f6f 100644 (file)
--- a/src/osd/PG.h
+++ b/src/osd/PG.h
@@ -2620,7 +2620,7 @@ public:
   void fulfill_info(pg_shard_t from, const pg_query_t &query,
                    pair<pg_shard_t, pg_info_t> &notify_info);
   void fulfill_log(pg_shard_t from, const pg_query_t &query, epoch_t query_epoch);
-
+  void fulfill_query(const MQuery& q, RecoveryCtx *rctx);
   void check_full_transition(OSDMapRef lastmap, OSDMapRef osdmap);
 
   bool should_restart_peering(