MXS-5022 Cleanup replication setup

Replication setup is no longer retried as any failure is likely
final.

system-test/maxtest/include/maxtest/galera_cluster.hh

@@ -23,7 +23,7 @@ class GaleraCluster : public MariaDBCluster
 public:
     GaleraCluster(mxt::SharedData* shared);
 
-    bool start_replication() override;
+    bool setup_replication() override;
 
     std::string get_srv_cnf_filename(int node) override;
 

system-test/maxtest/include/maxtest/mariadb_nodes.hh

@@ -268,11 +268,12 @@ public:
     int stop_slaves();
 
     /**
-     * Start replication in manner relevant to the cluster.
+     * Start replication in manner relevant to the cluster. Is called on a blank cluster so needs to
+     * also generate users.
      *
      * @return  True on success
      */
-    virtual bool start_replication() = 0;
+    virtual bool setup_replication() = 0;
 
     /**
      * Check if the cluster is replicating or otherwise properly synced. May also attempt light fixes.

system-test/maxtest/include/maxtest/replication_cluster.hh

@@ -26,7 +26,7 @@ class ReplicationCluster : public MariaDBCluster
 public:
     ReplicationCluster(SharedData* shared);
 
-    bool start_replication() override;
+    bool setup_replication() override;
 
     /**
      * Wait until slaves are up-to-date. Only considers the first gtid-triplet.

system-test/maxtest/src/galera_cluster.cc

@@ -37,7 +37,7 @@ GaleraCluster::GaleraCluster(mxt::SharedData* shared)
 {
 }
 
-bool GaleraCluster::start_replication()
+bool GaleraCluster::setup_replication()
 {
     int local_result = stop_nodes() ? 0 : 1;
 

system-test/maxtest/src/mariadb_nodes.cc

@@ -628,46 +628,21 @@ bool MariaDBCluster::prepare_for_test()
         if (reset_servers())
         {
             log.log_msgf("%s reset. Starting replication.", namec);
-            start_replication();
-
-            int attempts = 0;
-            bool cluster_ok = false;
-
-            while (!cluster_ok && attempts < 10)
-            {
-                if (attempts > 0)
-                {
-                    log.log_msgf("Iteration %i, %s is still broken, waiting.", attempts, namec);
-                    sleep(10);
-                }
-                if (check_fix_replication())
-                {
-                    cluster_ok = true;
-                }
-                attempts++;
-            }
-
-            if (cluster_ok)
+            if (setup_replication())
             {
                 log.log_msgf("%s is replicating/synced.", namec);
                 rval = prepare_servers_for_test();
             }
-            else
-            {
-                log.add_failure("%s is still broken.", namec);
-            }
         }
         else
         {
-            logger().add_failure("Server preparation on %s failed.", name().c_str());
+            log.log_msgf("Failed to reset servers of %s.", name().c_str());
         }
     }
     else
     {
         rval = true;
     }
-
-    disconnect();
     return rval;
 }
 

system-test/maxtest/src/replication_cluster.cc

@@ -83,32 +83,34 @@ const std::string& ReplicationCluster::type_string() const
     return type_mariadb;
 }
 
-bool ReplicationCluster::start_replication()
+bool ReplicationCluster::setup_replication()
 {
     const int n = N;
     // Generate users on all nodes.
     // TODO: most users can be generated just on the master once replication is on.
+    bool gtids_reset = true;
     for (int i = 0; i < n; i++)
     {
-        create_users(i);
-    }
-
-    ping_or_open_admin_connections();
-
-    // At this point, the servers have conflicting gtids but identical data. Set gtids manually so
-    // replication can start.
-    bool reset_ok = true;
-    for (int i = 0; i < n; i++)
-    {
-        auto conn = backend(i)->admin_connection();
-        if (!conn->try_cmd("RESET MASTER;") || !conn->try_cmd("SET GLOBAL gtid_slave_pos='0-1-0'"))
+        if (create_users(i))
         {
-            reset_ok = false;
+            // The servers now have conflicting gtids but identical data. Set gtids manually so
+            // replication can start.
+            auto conn = backend(i)->admin_connection();
+            if (!conn->try_cmd("RESET MASTER;") || !conn->try_cmd("SET GLOBAL gtid_slave_pos='0-1-0'"))
+            {
+                gtids_reset = false;
+                logger().log_msgf("Gtid reset failed on %s. Cannot setup replication.",
+                                  backend(i)->vm_node().name());
+            }
+        }
+        else
+        {
+            gtids_reset = false;
         }
     }
 
     bool rval = false;
-    if (reset_ok)
+    if (gtids_reset)
     {
         bool repl_ok = true;
         // Finally, begin replication.
@@ -118,16 +120,18 @@ bool ReplicationCluster::start_replication()
             auto conn = backend(i)->admin_connection();
             if (!conn->try_cmd(change_master) || !conn->try_cmd("START SLAVE;"))
             {
+                logger().log_msgf("Failed to start replication on %s. Cannot setup replication.",
+                                  backend(i)->vm_node().name());
                 repl_ok = false;
             }
         }
 
-        if (repl_ok)
+        if (repl_ok && sync_slaves(0, 5))
         {
+            logger().log_msgf("Replication setup success on %s.", name().c_str());
             rval = true;
         }
     }
-
     return rval;
 }
 

system-test/maxtest/src/testconnections.cc

@@ -171,7 +171,7 @@ int TestConnections::prepare_for_test(int argc, char* argv[])
     if (galera && restart_galera && m_shared.settings.mdbci_test)
     {
         galera->stop_nodes();
-        galera->start_replication();
+        galera->setup_replication();
     }
 
     if (m_check_nodes)