Fixed bug in semi_sync replication tests.

The problem was that wait_for_slave_io_to_start reported that the io thread was ready, when it was still initializing. This caused test suite to continue too early, for example before the semi sync plugin was properly enabled. Fixed by introducing a new internal stage: "Preparing". Slave_IO_Running is now set to "Yes" only when all initializing is done and the IO thread is ready to read things from the master. The only test affected by this change is rpl_flsh_tbls, which got stuck in the preparing phase while trying to read the GTID position from a table. Fixed by having this test waiting for Preparing instead of Yes.
MariaDB · Jan 3, 2016 · 8fcc0bf · 8fcc0bf
1 parent 661a6d8
commit 8fcc0bf
Show file tree

Hide file tree

Showing 9 changed files with 37 additions and 14 deletions.
diff --git a/mysql-test/extra/rpl_tests/rpl_flsh_tbls.test b/mysql-test/extra/rpl_tests/rpl_flsh_tbls.test
@@ -45,7 +45,13 @@ drop table t1;
 connection slave;
 flush tables with read lock;
 start slave;
-source include/wait_for_slave_to_start.inc;
+
+# The IO thread will not be able to read the GTID because of flush tables
+let $slave_param= Slave_IO_Running;
+let $slave_param_value= Preparing;
+source include/wait_for_slave_param.inc;
+
+--source include/wait_for_slave_sql_to_start.inc
 --error 1192
 stop slave;
 

diff --git a/mysql-test/suite/rpl/r/rpl_row_flsh_tbls.result b/mysql-test/suite/rpl/r/rpl_row_flsh_tbls.result
@@ -22,7 +22,8 @@ include/wait_for_slave_to_stop.inc
 drop table t1;
 flush tables with read lock;
 start slave;
-include/wait_for_slave_to_start.inc
+include/wait_for_slave_param.inc [Slave_IO_Running]
+include/wait_for_slave_sql_to_start.inc
 stop slave;
 ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
 unlock tables;

diff --git a/mysql-test/suite/rpl/r/rpl_semi_sync_uninstall_plugin.result b/mysql-test/suite/rpl/r/rpl_semi_sync_uninstall_plugin.result
@@ -14,13 +14,21 @@ DROP TABLE t1;
 [connection slave]
 include/install_semisync.inc
 [connection slave]
+show global status like "Slave%_running";
+Variable_name	Value
+Slave_running	ON
+Slaves_running	1
 UNINSTALL PLUGIN rpl_semi_sync_slave;
 Warnings:
 Warning	1620	Plugin is busy and will be uninstalled on shutdown
 select plugin_name,plugin_status from information_schema.plugins where plugin_name like 'rpl_%';
 plugin_name	plugin_status
 rpl_semi_sync_slave	DELETED
 [connection master]
+show global status like "Slave%_connect%";
+Variable_name	Value
+Slave_connections	2
+Slaves_connected	1
 UNINSTALL PLUGIN rpl_semi_sync_master;
 Warnings:
 Warning	1620	Plugin is busy and will be uninstalled on shutdown

diff --git a/mysql-test/suite/rpl/r/rpl_stm_flsh_tbls.result b/mysql-test/suite/rpl/r/rpl_stm_flsh_tbls.result
@@ -22,7 +22,8 @@ include/wait_for_slave_to_stop.inc
 drop table t1;
 flush tables with read lock;
 start slave;
-include/wait_for_slave_to_start.inc
+include/wait_for_slave_param.inc [Slave_IO_Running]
+include/wait_for_slave_sql_to_start.inc
 stop slave;
 ERROR HY000: Can't execute the given command because you have active locked tables or an active transaction
 unlock tables;

diff --git a/mysql-test/suite/rpl/t/rpl_semi_sync_uninstall_plugin.test b/mysql-test/suite/rpl/t/rpl_semi_sync_uninstall_plugin.test
@@ -57,13 +57,19 @@ DROP TABLE t1;
 #  possible at this state
 --connection slave
 --echo [connection slave]
+show global status like "Slave%_running";
+
 UNINSTALL PLUGIN rpl_semi_sync_slave;
 select plugin_name,plugin_status from information_schema.plugins where plugin_name like 'rpl_%';
 
 # Step 2.3: Check that rpl_semi_sync_master uninstallation on Master is not
 #  possible at this state
 --connection master
 --echo [connection master]
+
+# The following is to catch errors if the next uninstall plugin would succeed
+show global status like "Slave%_connect%";
+
 UNINSTALL PLUGIN rpl_semi_sync_master;
 select plugin_name,plugin_status from information_schema.plugins where plugin_name like 'rpl_%';
 

diff --git a/sql/mysqld.cc b/sql/mysqld.cc
@@ -7287,8 +7287,8 @@ static int show_slave_running(THD *thd, SHOW_VAR *var, char *buff)
       get_master_info(&thd->variables.default_master_connection,
                       Sql_condition::WARN_LEVEL_NOTE);
     if (mi)
-      tmp= (my_bool) (mi->slave_running == MYSQL_SLAVE_RUN_CONNECT &&
-                      mi->rli.slave_running);
+      tmp= (my_bool) (mi->slave_running == MYSQL_SLAVE_RUN_READING &&
+                      mi->rli.slave_running != MYSQL_SLAVE_NOT_RUN);
   }
   mysql_mutex_unlock(&LOCK_active_mi);
   if (mi)

diff --git a/sql/rpl_mi.cc b/sql/rpl_mi.cc
@@ -1321,7 +1321,7 @@ bool Master_info_index::start_all_slaves(THD *thd)
       Try to start all slaves that are configured (host is defined)
       and are not already running
     */
-    if ((mi->slave_running != MYSQL_SLAVE_RUN_CONNECT ||
+    if ((mi->slave_running == MYSQL_SLAVE_NOT_RUN ||
          !mi->rli.slave_running) && *mi->host)
     {
       if ((error= start_slave(thd, mi, 1)))

diff --git a/sql/slave.cc b/sql/slave.cc
@@ -2615,6 +2615,8 @@ static bool send_show_master_info_header(THD *thd, bool full,
   DBUG_RETURN(FALSE);
 }
 
+/* Text for Slave_IO_Running */
+static const char *slave_running[]= { "No", "Connecting", "Preparing", "Yes" };
 
 static bool send_show_master_info_data(THD *thd, Master_info *mi, bool full,
                                        String *gtid_pos)
@@ -2668,9 +2670,7 @@ static bool send_show_master_info_data(THD *thd, Master_info *mi, bool full,
                     &my_charset_bin);
     protocol->store((ulonglong) mi->rli.group_relay_log_pos);
     protocol->store(mi->rli.group_master_log_name, &my_charset_bin);
-    protocol->store(mi->slave_running == MYSQL_SLAVE_RUN_CONNECT ?
-                    "Yes" : (mi->slave_running == MYSQL_SLAVE_RUN_NOT_CONNECT ?
-                             "Connecting" : "No"), &my_charset_bin);
+    protocol->store(slave_running[mi->slave_running], &my_charset_bin);
     protocol->store(mi->rli.slave_running ? "Yes":"No", &my_charset_bin);
     protocol->store(rpl_filter->get_do_db());
     protocol->store(rpl_filter->get_ignore_db());
@@ -2713,7 +2713,7 @@ static bool send_show_master_info_data(THD *thd, Master_info *mi, bool full,
       Seconds_Behind_Master: if SQL thread is running and I/O thread is
       connected, we can compute it otherwise show NULL (i.e. unknown).
     */
-    if ((mi->slave_running == MYSQL_SLAVE_RUN_CONNECT) &&
+    if ((mi->slave_running == MYSQL_SLAVE_RUN_READING) &&
         mi->rli.slave_running)
     {
       long time_diff;

diff --git a/sql/slave.h b/sql/slave.h
@@ -132,11 +132,11 @@ extern my_bool opt_replicate_annotate_row_events;
 extern ulonglong relay_log_space_limit;
 
 /*
-  3 possible values for Master_info::slave_running and
+  4 possible values for Master_info::slave_running and
   Relay_log_info::slave_running.
-  The values 0,1,2 are very important: to keep the diff small, I didn't
-  substitute places where we use 0/1 with the newly defined symbols. So don't change
-  these values.
+  The values 0,1,2,3 are very important: to keep the diff small, I didn't
+  substitute places where we use 0/1 with the newly defined symbols.
+  So don't change these values.
   The same way, code is assuming that in Relay_log_info we use only values
   0/1.
   I started with using an enum, but
@@ -145,6 +145,7 @@ extern ulonglong relay_log_space_limit;
 #define MYSQL_SLAVE_NOT_RUN         0
 #define MYSQL_SLAVE_RUN_NOT_CONNECT 1
 #define MYSQL_SLAVE_RUN_CONNECT     2
+#define MYSQL_SLAVE_RUN_READING     3
 
 #define RPL_LOG_NAME (rli->group_master_log_name[0] ? rli->group_master_log_name :\
  "FIRST")