MXS-2158 Relax requirements for manual rejoin

The operation is now allowed even if the rejoining server has empty gtid:s.
Auto-rejoin keeps the safeties on.

Documentation/Monitors/MariaDB-Monitor.md

@@ -368,12 +368,17 @@ operations.
 ### Manual activation
 
 Cluster operations can be activated manually through the REST API, MaxCtrl or
-MaxAdmin. The commands are only performed when MaxScale is in active mode. All
+MaxAdmin. The commands are only performed when MaxScale is in active mode. The
-commands require the monitor instance name as the first parameter. Failover
+commands generally match their automatic versions. The exception is _rejoin_, in
+which the manual command allows rejoining even when the joining server has empty
+gtid:s. This rule allows the user to force a rejoin on a server without binary
+logs.
+
+All commands require the monitor instance name as the first parameter. Failover
 selects the new master server automatically and does not require additional
 parameters. Rejoin requires the name of the joining server as second parameter.
-Replication reset accepts the name of the new master server as second
+Replication reset accepts the name of the new master server as second parameter.
-parameter. If not given, the current master is selected.
+If not given, the current master is selected.
 
 Switchover takes one to three parameters. If only the monitor name is given,
 switchover will autoselect both the slave to promote and the current master as
@@ -401,8 +406,8 @@ to demote (OldMasterServ). For rejoin, the server to join (OldMasterServ) is
 required. Replication reset requires the server to promote (NewMasterServ).
 
 It is safe to perform manual operations even with automatic failover, switchover
-or rejoin enabled since the automatic operations cannot happen simultaneously
+or rejoin enabled since automatic operations cannot happen simultaneously
-with the manual one.
+with manual ones.
 
 If a switchover or failover fails, automatic failover is disabled to prevent
 master changes to a possibly malfunctioning cluster. Automatic failover can be

server/modules/monitor/mariadbmon/cluster_manipulation.cc

@@ -115,17 +115,42 @@ bool MariaDBMonitor::manual_rejoin(SERVER* rejoin_server, json_t** output)
         if (mon_slave_cand)
         {
             MariaDBServer* slave_cand = get_server_info(mon_slave_cand);
-
             if (server_is_rejoin_suspect(slave_cand, output))
             {
                 string gtid_update_error;
                 if (m_master->update_gtids(&gtid_update_error))
                 {
+                    // The manual version of rejoin does not need to be as careful as the automatic one.
+                    // The rules are mostly the same, the only difference is that a server with empty gtid:s
+                    // can be rejoined manually.
+                    // TODO: Add the warning to JSON output.
                     string no_rejoin_reason;
-                    if (slave_cand->can_replicate_from(m_master, &no_rejoin_reason))
+                    bool safe_rejoin = slave_cand->can_replicate_from(m_master, &no_rejoin_reason);
+                    bool empty_gtid = slave_cand->m_gtid_current_pos.empty();
+                    bool rejoin_allowed = false;
+                    if (safe_rejoin)
                     {
-                        ServerArray joinable_server;
+                        rejoin_allowed = true;
-                        joinable_server.push_back(slave_cand);
+                    }
+                    else
+                    {
+                        if (empty_gtid)
+                        {
+                            rejoin_allowed = true;
+                            MXB_WARNING("gtid_curren_pos of %s is empty. Manual rejoin is unsafe "
+                                        "but allowed.", rejoin_serv_name);
+                        }
+                        else
+                        {
+                            PRINT_MXS_JSON_ERROR(output, "%s cannot replicate from master server %s: %s",
+                                                 rejoin_serv_name, m_master->name(),
+                                                 no_rejoin_reason.c_str());
+                        }
+                    }
+
+                    if (rejoin_allowed)
+                    {
+                        ServerArray joinable_server = {slave_cand};
                         if (do_rejoin(joinable_server, output) == 1)
                         {
                             rval = true;
@@ -136,12 +161,6 @@ bool MariaDBMonitor::manual_rejoin(SERVER* rejoin_server, json_t** output)
                             PRINT_MXS_JSON_ERROR(output, "Rejoin attempted but failed.");
                         }
                     }
-                    else
-                    {
-                        PRINT_MXS_JSON_ERROR(output,
-                                             "%s cannot replicate from cluster master %s: %s.",
-                                             rejoin_serv_name, m_master->name(), no_rejoin_reason.c_str());
-                    }
                 }
                 else
                 {
@@ -153,8 +172,7 @@ bool MariaDBMonitor::manual_rejoin(SERVER* rejoin_server, json_t** output)
         }
         else
         {
-            PRINT_MXS_JSON_ERROR(output,
+            PRINT_MXS_JSON_ERROR(output, "The given server '%s' is not monitored by this monitor.",
-                                 "The given server '%s' is not monitored by this monitor.",
                                  rejoin_serv_name);
         }
     }
@@ -1862,4 +1880,4 @@ MariaDBMonitor::FailoverParams::FailoverParams(const ServerOperation& promotion,
     , demotion_target(demotion_target)
     , general(general)
 {
-}
+}