- Initialize member cluster ID only on connection to cluster and forget

ayurchen · ayurchen · commit 488c259e4ed6 · 2018-11-08T14:16:39.000+02:00
it on disconnect. - Don't rely on own index from the view because the view may come from another member (IST/SST), instead always determine own index from own ID. Refs #13
diff --git a/dbsim/db_server.cpp b/dbsim/db_server.cpp
@@ -27,15 +27,14 @@
 
 db::server::server(simulator& simulator,
                    const std::string& name,
-                   const std::string& server_id,
                    const std::string& address)
     : simulator_(simulator)
     , storage_engine_(simulator_.params())
     , mutex_()
     , cond_()
     , server_service_(*this)
     , server_state_(*this, server_service_,
-                    name, server_id, address, "dbsim_" + name + "_data")
+                    name, address, "dbsim_" + name + "_data")
     , last_client_id_(0)
     , last_transaction_id_(0)
     , appliers_()
diff --git a/dbsim/db_server.hpp b/dbsim/db_server.hpp
@@ -41,7 +41,6 @@ namespace db
     public:
         server(simulator& simulator,
                const std::string& name,
-               const std::string& id,
                const std::string& address);
         void applier_thread();
         void start_applier();
diff --git a/dbsim/db_server_state.hpp b/dbsim/db_server_state.hpp
@@ -35,15 +35,13 @@ namespace db
         server_state(db::server& server,
                      wsrep::server_service& server_service,
                      const std::string& name,
-                     const std::string& server_id,
                      const std::string& address,
                      const std::string& working_dir)
             : wsrep::server_state(
                 mutex_,
                 cond_,
                 server_service,
                 name,
-                server_id,
                 "",
                 address,
                 working_dir,
diff --git a/dbsim/db_simulator.cpp b/dbsim/db_simulator.cpp
@@ -109,7 +109,6 @@ void db::simulator::start()
                         std::make_unique<db::server>(
                             *this,
                             name_os.str(),
-                            id_os.str(),
                             address_os.str()))));
         if (it.second == false)
         {
diff --git a/include/wsrep/server_state.hpp b/include/wsrep/server_state.hpp
@@ -296,7 +296,7 @@ namespace wsrep
          * A method which will be called when the server
          * has been joined to the cluster
          */
-        void on_connect(const wsrep::gtid& gtid);
+        void on_connect(const wsrep::view& view);
 
         /**
          * A method which will be called when a view
@@ -540,7 +540,6 @@ namespace wsrep
                      wsrep::condition_variable& cond,
                      wsrep::server_service& server_service,
                      const std::string& name,
-                     const std::string& id,
                      const std::string& incoming_address,
                      const std::string& address,
                      const std::string& working_dir,
@@ -565,7 +564,7 @@ namespace wsrep
             , streaming_appliers_()
             , provider_()
             , name_(name)
-            , id_(id)
+            , id_(wsrep::id::undefined())
             , incoming_address_(incoming_address)
             , address_(address)
             , working_dir_(working_dir)
diff --git a/include/wsrep/view.hpp b/include/wsrep/view.hpp
@@ -30,6 +30,7 @@
 #include "seqno.hpp"
 #include "gtid.hpp"
 #include <vector>
+#include <iostream>
 
 namespace wsrep
 {
@@ -111,6 +112,8 @@ namespace wsrep
             return (members_.empty() && own_index_ == -1);
         }
 
+        void print(std::ostream& os) const;
+
     private:
         wsrep::gtid state_id_;
         wsrep::seqno view_seqno_;
@@ -120,6 +123,12 @@ namespace wsrep
         int protocol_version_;
         std::vector<wsrep::view::member> members_;
     };
+
+    static inline
+    std::ostream& operator<<(std::ostream& os, const wsrep::view& v)
+    {
+        v.print(os); return os;
+    }
 }
 
 #endif // WSREP_VIEW
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
@@ -11,6 +11,7 @@ add_library(wsrep-lib
   logger.cpp
   provider.cpp
   seqno.cpp
+  view.cpp
   server_state.cpp
   transaction.cpp
   wsrep_provider_v26.cpp)
diff --git a/src/server_state.cpp b/src/server_state.cpp
@@ -565,14 +565,38 @@ wsrep::server_state::causal_read(int timeout) const
     return provider_->causal_read(timeout);
 }
 
-void wsrep::server_state::on_connect(const wsrep::gtid& gtid)
+void wsrep::server_state::on_connect(const wsrep::view& view)
 {
+    // Sanity checks
+    if (id_.is_undefined() == false)
+    {
+        wsrep::log_warning() << "Unexpected connection in connected state. "
+                             << "Received view: " << view
+                             << "Previous ID: " << id_;
+        assert(0);
+    }
+
+    if (view.own_index() < 0 ||
+        size_t(view.own_index()) >= view.members().size())
+    {
+        std::ostringstream os;
+        os << "Invalid view on connect: own index out of range: " << view;
+        wsrep::log_error() << os.str();
+        assert(0);
+        throw wsrep::runtime_error(os.str());
+    }
+
+    id_ = view.members()[view.own_index()].id();
+
     wsrep::log_info() << "Server "
                       << name_
                       << " connected to cluster at position "
-                      << gtid;
+                      << view.state_id()
+                      << " with ID "
+                      << id_;
+
     wsrep::unique_lock<wsrep::mutex> lock(mutex_);
-    connected_gtid_ = gtid;
+    connected_gtid_ = view.state_id();
     state(lock, s_connected);
 }
 
@@ -599,22 +623,7 @@ void wsrep::server_state::on_view(const wsrep::view& view,
     if (view.status() == wsrep::view::primary)
     {
         wsrep::unique_lock<wsrep::mutex> lock(mutex_);
-        if (view.own_index() >= 0)
-        {
-            if (id_.is_undefined())
-            {
-                // No identifier was passed during server state initialization
-                // and the ID was generated by the provider.
-                id_ = view.members()[view.own_index()].id();
-            }
-            else
-            {
-                // Own identifier must not change between views.
-                // assert(id_ == view.members()[view.own_index()].id());
-            }
-        }
         assert(view.final() == false);
-
         //
         // Reached primary from connected state. This may mean the following
         //
@@ -703,6 +712,7 @@ void wsrep::server_state::on_view(const wsrep::view& view,
             {
                 close_transactions_at_disconnect(*high_priority_service);
             }
+            id_ = id::undefined();
             state(lock, s_disconnected);
         }
         else if (state_ != s_disconnecting)
diff --git a/src/view.cpp b/src/view.cpp
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) 2018 Codership Oy <info@codership.com>
+ *
+ * This file is part of wsrep-lib.
+ *
+ * Wsrep-lib is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation, either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * Wsrep-lib is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with wsrep-lib.  If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include "wsrep/view.hpp"
+
+void wsrep::view::print(std::ostream& os) const
+{
+    os << "  id: " << state_id() << "\n"
+       << "  status: " << status() << "\n"
+       << "  prococol_version: " << protocol_version() << "\n"
+       << "  final: " << final() << "\n"
+       << "  own_index: " << own_index() << "\n"
+       << "  members(" << members().size() << "):\n";
+
+    for (std::vector<wsrep::view::member>::const_iterator i(members().begin());
+         i != members().end(); ++i)
+    {
+        os << "\t" << (i - members().begin()) /* ordinal index */
+           << ") id: " << i->id()
+           << ", name: " << i->name() << "\n";
+    }
+}
diff --git a/src/wsrep_provider_v26.cpp b/src/wsrep_provider_v26.cpp
@@ -287,7 +287,8 @@ namespace
     {
         return capabilities;
     }
-    wsrep::view view_from_native(const wsrep_view_info& view_info)
+    wsrep::view view_from_native(const wsrep_view_info& view_info,
+                                 const wsrep::id& own_id)
     {
         std::vector<wsrep::view::member> members;
         for (int i(0); i < view_info.memb_num; ++i)
@@ -303,6 +304,25 @@ namespace
                         sizeof(view_info.members[i].incoming)));
             members.push_back(wsrep::view::member(id, name, incoming));
         }
+
+        int own_idx(-1);
+        if (own_id.is_undefined())
+        {
+            // If own ID is undefined, obtain it from the view. This is
+            // the case on the initial connect to cluster.
+            own_idx = view_info.my_idx;
+        }
+        else
+        {
+            // If the node has already obtained its ID from cluster,
+            // its position in the view (or lack thereof) must be determined
+            // by the ID.
+            for (size_t i(0); i < members.size(); ++i)
+            {
+                if (own_id == members[i].id()) { own_idx = i; break; }
+            }
+        }
+
         return wsrep::view(
             wsrep::gtid(
                 wsrep::id(view_info.state_id.uuid.data,
@@ -311,7 +331,7 @@ namespace
             wsrep::seqno(view_info.view),
             map_view_status_from_native(view_info.status),
             map_capabilities_from_native(view_info.capabilities),
-            view_info.my_idx,
+            own_idx,
             view_info.proto_ver,
             members);
     }
@@ -325,12 +345,14 @@ namespace
         const wsrep_view_info_t* view_info)
     {
         assert(app_ctx);
-        wsrep::view view(view_from_native(*view_info));
         wsrep::server_state& server_state(
             *reinterpret_cast<wsrep::server_state*>(app_ctx));
+        assert(server_state.id().is_undefined());
+        wsrep::view view(view_from_native(*view_info, server_state.id()));
+        assert(view.own_index() >= 0);
         try
         {
-            server_state.on_connect(view.state_id());
+            server_state.on_connect(view);
             return WSREP_CB_SUCCESS;
         }
         catch (const wsrep::runtime_error& e)
@@ -354,7 +376,7 @@ namespace
             reinterpret_cast<wsrep::high_priority_service*>(recv_ctx));
         try
         {
-            wsrep::view view(view_from_native(*view_info));
+            wsrep::view view(view_from_native(*view_info, server_state.id()));
             server_state.on_view(view, high_priority_service);
             return WSREP_CB_SUCCESS;
         }
diff --git a/test/client_state_fixture.hpp b/test/client_state_fixture.hpp
diff --git a/test/mock_server_state.hpp b/test/mock_server_state.hpp
diff --git a/test/server_context_test.cpp b/test/server_context_test.cpp

Original file line number	Diff line number	Diff line change
`@@ -109,7 +109,6 @@ void db::simulator::start()`
`109`	`109`	`std::make_unique<db::server>(`
`110`	`110`	`*this,`
`111`	`111`	`name_os.str(),`
`112`		`- id_os.str(),`
`113`	`112`	`address_os.str()))));`
`114`	`113`	`if (it.second == false)`
`115`	`114`	`{`