Merge branch '2.3' into develop

Documentation/Routers/ReadWriteSplit.md

@@ -109,12 +109,14 @@ This feature is disabled by default.
 
 	max_slave_replication_lag=<allowed lag in seconds>
 
-This applies to Master/Slave replication with MySQL monitor and
-`detect_replication_lag=1` options set. max_slave_replication_lag must be
-greater than the monitor interval.
-
-This option only affects Master-Slave clusters. Galera clusters do not have a
-concept of slave lag even if the application of write sets might have lag.
+The Readwritesplit-router does not detect the replication lag itself. A monitor
+such as the MariaDB-monitor for a Master/Slave-cluster is required. This option
+only affects Master-Slave clusters. Galera clusters do not have a concept of
+slave lag even if the application of write sets might have lag. When a server is
+disqualified from routing because of replication lag, a warning is logged. Similarly,
+when the server has caught up enough to be a valid routing target, another warning
+is logged. These messages are only logged when a query is being routed and the
+replication state changes.
 
 ### `use_sql_variables_in`
 

include/maxscale/protocol/rwbackend.hh

@@ -134,6 +134,15 @@ public:
 
     // Controlled by the session
     ResponseStat& response_stat();
+
+    /**
+     * Change server replication lag state and log warning when state changes.
+     *
+     * @param new_state New replication lag state
+     * @param max_rlag Maximum allowed lag. Used for the log message.
+     */
+    void change_rlag_state(SERVER::RLagState new_state, int max_rlag);
+
 private:
     reply_state_t    m_reply_state;
     BackendHandleMap m_ps_handles;      /**< Internal ID to backend PS handle mapping */

include/maxscale/server.hh

@@ -142,6 +142,13 @@ public:
         CLUSTRIX
     };
 
+    enum class RLagState
+    {
+        NONE,
+        BELOW_LIMIT,
+        ABOVE_LIMIT
+    };
+
     struct Version
     {
         uint64_t total = 0;     /**< The version number received from server */
@@ -193,9 +200,10 @@ public:
     unsigned long node_ts = 0;          /**< Last timestamp set from M/S monitor module */
 
     // Misc fields
-    bool master_err_is_logged = false;  /**< If node failed, this indicates whether it is logged. Only
-                                         * used by rwsplit. TODO: Move to rwsplit */
-    bool warn_ssl_not_enabled = true;   /**< SSL not used for an SSL enabled server */
+    bool master_err_is_logged = false;      /**< If node failed, this indicates whether it is logged. Only
+                                             * used by rwsplit. TODO: Move to rwsplit */
+    bool warn_ssl_not_enabled = true;       /**< SSL not used for an SSL enabled server */
+    RLagState rlag_state = RLagState::NONE; /**< Is replication lag above or under limit? Used by rwsplit. */
 
     virtual ~SERVER() = default;
 

server/modules/protocol/MySQL/rwbackend.cc

@@ -256,6 +256,31 @@ ResponseStat& RWBackend::response_stat()
     return m_response_stat;
 }
 
+void RWBackend::change_rlag_state(SERVER::RLagState new_state, int max_rlag)
+{
+    mxb_assert(new_state == SERVER::RLagState::BELOW_LIMIT || new_state == SERVER::RLagState::ABOVE_LIMIT);
+    namespace atom = maxbase::atomic;
+    auto srv = server();
+    auto old_state = atom::load(&srv->rlag_state, atom::RELAXED);
+    if (new_state != old_state)
+    {
+        atom::store(&srv->rlag_state, new_state, atom::RELAXED);
+        // State has just changed, log warning. Don't log catchup if old state was RLAG_NONE.
+        if (new_state == SERVER::RLagState::ABOVE_LIMIT)
+        {
+            MXS_WARNING("Replication lag of '%s' is %is, which is above the configured limit %is. "
+                        "'%s' is excluded from query routing.",
+                        srv->name(), srv->rlag, max_rlag, srv->name());
+        }
+        else if (old_state == SERVER::RLagState::ABOVE_LIMIT)
+        {
+            MXS_WARNING("Replication lag of '%s' is %is, which is below the configured limit %is. "
+                        "'%s' is returned to query routing.",
+                        srv->name(), srv->rlag, max_rlag, srv->name());
+        }
+    }
+}
+
 mxs::SRWBackends RWBackend::from_servers(SERVER_REF* servers)
 {
     SRWBackends backends;

server/modules/routing/readwritesplit/rwsplit_route_stmt.cc

@@ -605,14 +605,25 @@ RWBackend* RWSplitSession::get_slave_backend(int max_rlag)
             && counts.second < m_router->max_slave_count();
 
         bool master_or_slave = backend->is_master() || backend->is_slave();
-        bool is_useable = backend->in_use() || can_take_slave_into_use;
-        bool not_a_slacker = rpl_lag_is_ok(backend, max_rlag);
+        bool is_usable = backend->in_use() || can_take_slave_into_use;
+        bool rlag_ok = rpl_lag_is_ok(backend, max_rlag);
 
-        bool server_is_candidate = master_or_slave && is_useable && not_a_slacker;
-
-        if (server_is_candidate)
+        if (master_or_slave && is_usable)
         {
-            candidates.push_back(backend);
+            if (rlag_ok)
+            {
+                candidates.push_back(backend);
+                if (max_rlag > 0)
+                {
+                    // Replication lag discrimination is on and the server passed.
+                    backend->change_rlag_state(SERVER::RLagState::BELOW_LIMIT, max_rlag);
+                }
+            }
+            else
+            {
+                // The server is otherwise usable except it's lagging too much.
+                backend->change_rlag_state(SERVER::RLagState::ABOVE_LIMIT, max_rlag);
+            }
         }
     }