Merge pull request #23507 from ukernel/luminous-26858

luminous: mds: reset heartbeat map at potential time-consuming places Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
ceph · Sep 17, 2018 · 786e77e · 786e77e
2 parents 7180037 + 5bf6312
commit 786e77e
Show file tree

Hide file tree

Showing 4 changed files with 25 additions and 2 deletions.
diff --git a/src/mds/MDSRank.cc b/src/mds/MDSRank.cc
@@ -1020,6 +1020,11 @@ utime_t MDSRank::get_laggy_until() const
   return beacon.get_laggy_until();
 }
 
+double MDSRank::get_dispatch_queue_max_age(utime_t now) const
+{
+  return messenger->get_dispatch_queue_max_age(now);
+}
+
 bool MDSRank::is_daemon_stopping() const
 {
   return stopping;
@@ -1827,7 +1832,7 @@ void MDSRankDispatcher::handle_mds_map(
       list<MDSInternalContextBase*> ls;
       ls.swap(p->second);
       waiting_for_mdsmap.erase(p++);
-      finish_contexts(g_ceph_context, ls);
+      queue_waiters(ls);
     }
   }
 

diff --git a/src/mds/MDSRank.h b/src/mds/MDSRank.h
@@ -301,6 +301,10 @@ class MDSRank {
       finished_queue.splice( finished_queue.end(), ls );
       progress_thread.signal();
     }
+    void queue_waiters_front(std::list<MDSInternalContextBase*>& ls) {
+      finished_queue.splice(finished_queue.begin(), ls);
+      progress_thread.signal();
+    }
 
     MDSRank(
         mds_rank_t whoami_,
@@ -355,6 +359,7 @@ class MDSRank {
     void damaged_unlocked();
 
     utime_t get_laggy_until() const;
+    double get_dispatch_queue_max_age(utime_t now) const;
 
     void send_message_mds(Message *m, mds_rank_t mds);
     void forward_message_mds(Message *req, mds_rank_t mds);

diff --git a/src/mds/Migrator.cc b/src/mds/Migrator.cc
@@ -2606,6 +2606,15 @@ void Migrator::import_remove_pins(CDir *dir, set<CDir*>& bounds)
   }
 }
 
+class C_MDC_QueueContexts : public MigratorContext {
+public:
+  list<MDSInternalContextBase*> contexts;
+  C_MDC_QueueContexts(Migrator *m) : MigratorContext(m) {}
+  void finish(int r) override {
+    // execute contexts immediately after 'this' context
+    get_mds()->queue_waiters_front(contexts);
+  }
+};
 
 /*
  * note: this does teh full work of reversing and import and cleaning up
@@ -2633,7 +2642,7 @@ void Migrator::import_reverse(CDir *dir)
 
   cache->adjust_subtree_auth(dir, stat.peer);
 
-  C_ContextsBase<MDSInternalContextBase, MDSInternalContextGather> *fin = new C_ContextsBase<MDSInternalContextBase, MDSInternalContextGather>(g_ceph_context);
+  auto fin = new C_MDC_QueueContexts(this);
   if (!dir->get_inode()->is_auth() &&
       !dir->get_inode()->has_subtree_root_dirfrag(mds->get_nodeid())) {
     dir->get_inode()->clear_scatter_dirty();

diff --git a/src/mds/Server.cc b/src/mds/Server.cc
@@ -707,7 +707,10 @@ void Server::find_idle_sessions()
   // timeout/stale
   //  (caps go stale, lease die)
   utime_t now = ceph_clock_now();
+  double queue_max_age = mds->get_dispatch_queue_max_age(now);
+
   utime_t cutoff = now;
+  cutoff -= queue_max_age;
   cutoff -= g_conf->mds_session_timeout;  
   while (1) {
     Session *session = mds->sessionmap.get_oldest_session(Session::STATE_OPEN);
@@ -729,6 +732,7 @@ void Server::find_idle_sessions()
 
   // autoclose
   cutoff = now;
+  cutoff -= queue_max_age;
   cutoff -= g_conf->mds_session_autoclose;
 
   // don't kick clients if we've been laggy