MDEV-17515: GTID Replication in optimistic mode deadlock

Problem:
=======
In slave_parallel_mode=optimistic configuration, when admin commands and
DML operation on the same table are scheduled simultaneously for execution,
it results in lock conflict and slave server either hangs due to
deadlock or goes down with an assert.

Analysis:
========
Admin commands OPTIMIZE, REPAIR and ANALYZE are written to binary log as
ordinary transactions. When 'slave_parallel_mode' is 'optimistic' DMLs are
allowed to run in parallel. But these locks are not detected by parallel
replication deadlock detection-and-handling mechanism. At times they result
in deadlock or assertion.

Fix:
===
Flag admin commands as DDL in Gtid_log_event at the time of writing to
binary log. Add a new bit EXECUTED_TABLE_ADMIN_CMD to
'm_unsafe_rollback_flags'. During 'mysql_admin_table' command execution it
accepts a list of tables to be processed and executes them in a loop. Upon
successful execution enable 'EXECUTED_TABLE_ADMIN_CMD' bit in
thd->transaction.stmt_unsafe_rollback_flags. Gtid_log_event constructor
will notice this flag and mark the current transaction with 'FL_DDL' flag.
Gtid_log_events marked as FL_DDL will not be scheduled parallel execution,
on the slave. They will execute in isolation to prevent deadlocks.

Note: Removed the call to 'trans_commit_implicit' from 'mysql_admin_table'
function as 'mysql_execute_command' will take care of invoking
'trans_commit_implicit'.

mysql-test/include/commit.inc

@@ -770,7 +770,7 @@ call p_verify_status_increment(2, 0, 2, 0);
 commit;
 call p_verify_status_increment(0, 0, 0, 0);
 check table t1, t2, t3;
-call p_verify_status_increment(6, 0, 6, 0);
+call p_verify_status_increment(4, 0, 4, 0);
 commit;
 call p_verify_status_increment(0, 0, 0, 0);
 drop view v1;

mysql-test/r/commit_1innodb.result

@@ -873,7 +873,7 @@ Table	Op	Msg_type	Msg_text
 test.t1	check	status	OK
 test.t2	check	status	OK
 test.t3	check	status	OK
-call p_verify_status_increment(6, 0, 6, 0);
+call p_verify_status_increment(4, 0, 4, 0);
 SUCCESS
 
 commit;

mysql-test/suite/rpl/r/rpl_mark_optimize_tbl_ddl.result

+include/rpl_init.inc [topology=1->2]
+connection server_1;
+FLUSH TABLES;
+ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
+connection server_2;
+SET @save_slave_parallel_threads= @@GLOBAL.slave_parallel_threads;
+SET @save_slave_parallel_mode= @@GLOBAL.slave_parallel_mode;
+include/stop_slave.inc
+SET GLOBAL slave_parallel_threads=2;
+SET GLOBAL slave_parallel_mode=optimistic;
+include/start_slave.inc
+connection server_1;
+CREATE TABLE t1(a INT) ENGINE=INNODB;
+OPTIMIZE TABLE t1;
+Table	Op	Msg_type	Msg_text
+test.t1	optimize	note	Table does not support optimize, doing recreate + analyze instead
+test.t1	optimize	status	OK
+INSERT INTO t1 VALUES(1);
+INSERT INTO t1  SELECT 1+a FROM t1;
+INSERT INTO t1  SELECT 2+a FROM t1;
+connection server_2;
+#
+# Verify that following admin commands are marked as ddl
+# 'OPTIMIZE TABLE', 'REPAIR TABLE' and 'ANALYZE TABLE'
+#
+connection server_1;
+OPTIMIZE TABLE t1;
+Table	Op	Msg_type	Msg_text
+test.t1	optimize	note	Table does not support optimize, doing recreate + analyze instead
+test.t1	optimize	status	OK
+REPAIR TABLE t1;
+Table	Op	Msg_type	Msg_text
+test.t1	repair	note	The storage engine for the table doesn't support repair
+ANALYZE TABLE t1;
+Table	Op	Msg_type	Msg_text
+test.t1	analyze	status	OK
+FLUSH LOGS;
+FOUND 1 /GTID 0-1-8 ddl/ in mysqlbinlog.out
+FOUND 1 /GTID 0-1-9 ddl/ in mysqlbinlog.out
+FOUND 1 /GTID 0-1-10 ddl/ in mysqlbinlog.out
+#
+# Clean up
+#
+DROP TABLE t1;
+connection server_2;
+FLUSH LOGS;
+#
+# Check that ALTER TABLE commands with ANALYZE, OPTIMIZE and REPAIR on
+# partitions will be marked as DDL in binary log.
+#
+connection server_1;
+CREATE TABLE t1(id INT) PARTITION BY RANGE (id) (PARTITION p0 VALUES LESS THAN (100),
+PARTITION pmax VALUES LESS THAN (MAXVALUE));
+INSERT INTO t1 VALUES (1), (10), (100), (1000);
+ALTER TABLE t1 ANALYZE PARTITION p0;
+Table	Op	Msg_type	Msg_text
+test.t1	analyze	status	OK
+ALTER TABLE t1 OPTIMIZE PARTITION p0;
+Table	Op	Msg_type	Msg_text
+test.t1	optimize	status	OK
+ALTER TABLE t1 REPAIR PARTITION p0;
+Table	Op	Msg_type	Msg_text
+test.t1	repair	status	OK
+FLUSH LOGS;
+FOUND 1 /GTID 0-1-14 ddl/ in mysqlbinlog.out
+FOUND 1 /GTID 0-1-15 ddl/ in mysqlbinlog.out
+FOUND 1 /GTID 0-1-16 ddl/ in mysqlbinlog.out
+#
+# Clean up
+#
+DROP TABLE t1;
+connection server_2;
+include/stop_slave.inc
+SET GLOBAL slave_parallel_threads= @save_slave_parallel_threads;
+SET GLOBAL slave_parallel_mode= @save_slave_parallel_mode;
+include/start_slave.inc
+include/rpl_end.inc

mysql-test/suite/rpl/t/rpl_mark_optimize_tbl_ddl.test

+# ==== Purpose ====
+#
+# Test verifies that there is no deadlock or assertion in
+# slave_parallel_mode=optimistic configuration while applying admin command
+# like 'OPTIMIZE TABLE', 'REPAIR TABLE' and 'ANALYZE TABLE'.
+#
+# ==== Implementation ====
+#
+# Steps:
+#    0 - Create a table, execute OPTIMIZE TABLE command on the table followed
+#        by some DMLS.
+#    1 - No assert should happen on slave server.
+#    2 - Assert that 'OPTIMIZE TABLE', 'REPAIR TABLE' and 'ANALYZE TABLE' are
+#        marked as 'DDL' in the binary log.
+#
+# ==== References ====
+#
+# MDEV-17515: GTID Replication in optimistic mode deadlock
+#
+--source include/have_partition.inc
+--source include/have_innodb.inc
+--let $rpl_topology=1->2
+--source include/rpl_init.inc
+
+--connection server_1
+FLUSH TABLES;
+ALTER TABLE mysql.gtid_slave_pos ENGINE=InnoDB;
+
+--connection server_2
+SET @save_slave_parallel_threads= @@GLOBAL.slave_parallel_threads;
+SET @save_slave_parallel_mode= @@GLOBAL.slave_parallel_mode;
+--source include/stop_slave.inc
+SET GLOBAL slave_parallel_threads=2;
+SET GLOBAL slave_parallel_mode=optimistic;
+--source include/start_slave.inc
+
+--connection server_1
+CREATE TABLE t1(a INT) ENGINE=INNODB;
+OPTIMIZE TABLE t1;
+INSERT INTO t1 VALUES(1);
+INSERT INTO t1  SELECT 1+a FROM t1;
+INSERT INTO t1  SELECT 2+a FROM t1;
+--save_master_pos
+
+--connection server_2
+--sync_with_master
+
+--echo #
+--echo # Verify that following admin commands are marked as ddl
+--echo # 'OPTIMIZE TABLE', 'REPAIR TABLE' and 'ANALYZE TABLE'
+--echo #
+--connection server_1
+
+OPTIMIZE TABLE t1;
+--let optimize_gtid= `SELECT @@GLOBAL.gtid_binlog_pos`
+
+REPAIR TABLE t1;
+--let repair_gtid= `SELECT @@GLOBAL.gtid_binlog_pos`
+
+ANALYZE TABLE t1;
+--let analyze_gtid= `SELECT @@GLOBAL.gtid_binlog_pos`
+
+let $binlog_file= query_get_value(SHOW MASTER STATUS, File, 1);
+FLUSH LOGS;
+
+--let $MYSQLD_DATADIR= `select @@datadir`
+--exec $MYSQL_BINLOG $MYSQLD_DATADIR/$binlog_file > $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+
+--let SEARCH_PATTERN= GTID $optimize_gtid ddl
+--let SEARCH_FILE= $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--source include/search_pattern_in_file.inc
+
+--let SEARCH_PATTERN= GTID $repair_gtid ddl
+--let SEARCH_FILE= $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--source include/search_pattern_in_file.inc
+
+--let SEARCH_PATTERN= GTID $analyze_gtid ddl
+--let SEARCH_FILE= $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--source include/search_pattern_in_file.inc
+
+--echo #
+--echo # Clean up
+--echo #
+DROP TABLE t1;
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--save_master_pos
+
+--connection server_2
+--sync_with_master
+FLUSH LOGS;
+
+--echo #
+--echo # Check that ALTER TABLE commands with ANALYZE, OPTIMIZE and REPAIR on
+--echo # partitions will be marked as DDL in binary log.
+--echo #
+--connection server_1
+CREATE TABLE t1(id INT) PARTITION BY RANGE (id) (PARTITION p0 VALUES LESS THAN (100),
+ PARTITION pmax VALUES LESS THAN (MAXVALUE));
+INSERT INTO t1 VALUES (1), (10), (100), (1000);
+
+ALTER TABLE t1 ANALYZE PARTITION p0;
+--let analyze_gtid= `SELECT @@GLOBAL.gtid_binlog_pos`
+
+ALTER TABLE t1 OPTIMIZE PARTITION p0;
+--let optimize_gtid= `SELECT @@GLOBAL.gtid_binlog_pos`
+
+ALTER TABLE t1 REPAIR PARTITION p0;
+--let repair_gtid= `SELECT @@GLOBAL.gtid_binlog_pos`
+
+let $binlog_file= query_get_value(SHOW MASTER STATUS, File, 1);
+FLUSH LOGS;
+
+--exec $MYSQL_BINLOG $MYSQLD_DATADIR/$binlog_file > $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+
+--let SEARCH_PATTERN= GTID $analyze_gtid ddl
+--let SEARCH_FILE= $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--source include/search_pattern_in_file.inc
+
+--let SEARCH_PATTERN= GTID $optimize_gtid ddl
+--let SEARCH_FILE= $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--source include/search_pattern_in_file.inc
+
+--let SEARCH_PATTERN= GTID $repair_gtid ddl
+--let SEARCH_FILE= $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--source include/search_pattern_in_file.inc
+
+--echo #
+--echo # Clean up
+--echo #
+DROP TABLE t1;
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqlbinlog.out
+--save_master_pos
+
+--connection server_2
+--sync_with_master
+
+--source include/stop_slave.inc
+SET GLOBAL slave_parallel_threads= @save_slave_parallel_threads;
+SET GLOBAL slave_parallel_mode= @save_slave_parallel_mode;
+--source include/start_slave.inc
+
+--source include/rpl_end.inc

sql/handler.h

@@ -1481,7 +1481,17 @@ struct THD_TRANS
   static unsigned int const DROPPED_TEMP_TABLE= 0x04;
   static unsigned int const DID_WAIT= 0x08;
   static unsigned int const DID_DDL= 0x10;
+  static unsigned int const EXECUTED_TABLE_ADMIN_CMD= 0x20;
 
+  void mark_executed_table_admin_cmd()
+  {
+    DBUG_PRINT("debug", ("mark_executed_table_admin_cmd"));
+    m_unsafe_rollback_flags|= EXECUTED_TABLE_ADMIN_CMD;
+  }
+  bool trans_executed_admin_cmd()
+  {
+    return (m_unsafe_rollback_flags & EXECUTED_TABLE_ADMIN_CMD) != 0;
+  }
   void mark_created_temp_table()
   {
     DBUG_PRINT("debug", ("mark_created_temp_table"));

sql/log_event.cc

@@ -7561,8 +7561,10 @@ Gtid_log_event::Gtid_log_event(THD *thd_arg, uint64 seq_no_arg,
     flags2|= FL_WAITED;
   if (thd_arg->transaction.stmt.trans_did_ddl() ||
       thd_arg->transaction.stmt.has_created_dropped_temp_table() ||
+      thd_arg->transaction.stmt.trans_executed_admin_cmd() ||
       thd_arg->transaction.all.trans_did_ddl() ||
-      thd_arg->transaction.all.has_created_dropped_temp_table())
+      thd_arg->transaction.all.has_created_dropped_temp_table() ||
+      thd_arg->transaction.all.trans_executed_admin_cmd())
     flags2|= FL_DDL;
   else if (is_transactional && !is_tmp_table)
     flags2|= FL_TRANSACTIONAL;

sql/sql_admin.cc

@@ -434,7 +434,8 @@ static bool mysql_admin_table(THD* thd, TABLE_LIST* tables,
                               int (handler::*operator_func)(THD *,
                                                             HA_CHECK_OPT *),
                               int (view_operator_func)(THD *, TABLE_LIST*,
-                                                       HA_CHECK_OPT *))
+                                                       HA_CHECK_OPT *),
+                              bool is_cmd_replicated)
 {
   TABLE_LIST *table;
   List<Item> field_list;
@@ -1147,6 +1148,13 @@ static bool mysql_admin_table(THD* thd, TABLE_LIST* tables,
         break;
       }
     }
+    /*
+      Admin commands acquire table locks and these locks are not detected by
+      parallel replication deadlock detection-and-handling mechanism. Hence
+      they must be marked as DDL so that they are not scheduled in parallel
+      with conflicting DMLs resulting in deadlock.
+    */
+    thd->transaction.stmt.mark_executed_table_admin_cmd();
     if (table->table && !table->view)
     {
       if (table->table->s->tmp_table)
@@ -1182,9 +1190,7 @@ static bool mysql_admin_table(THD* thd, TABLE_LIST* tables,
     }
     else
     {
-      if (trans_commit_stmt(thd) ||
-          (stmt_causes_implicit_commit(thd, CF_IMPLICIT_COMMIT_END) &&
-           trans_commit_implicit(thd)))
+      if (trans_commit_stmt(thd))
         goto err;
     }
     close_thread_tables(thd);
@@ -1209,6 +1215,11 @@ static bool mysql_admin_table(THD* thd, TABLE_LIST* tables,
     if (protocol->write())
       goto err;
   }
+  if (is_cmd_replicated && !thd->lex->no_write_to_binlog)
+  {
+    if (write_bin_log(thd, TRUE, thd->query(), thd->query_length()))
+      goto err;
+  }
 
   my_eof(thd);
   thd->resume_subsequent_commits(suspended_wfc);
@@ -1270,7 +1281,7 @@ bool mysql_assign_to_keycache(THD* thd, TABLE_LIST* tables,
   check_opt.key_cache= key_cache;
   DBUG_RETURN(mysql_admin_table(thd, tables, &check_opt,
 				"assign_to_keycache", TL_READ_NO_INSERT, 0, 0,
-				0, 0, &handler::assign_to_keycache, 0));
+				0, 0, &handler::assign_to_keycache, 0, false));
 }
 
 
@@ -1297,7 +1308,7 @@ bool mysql_preload_keys(THD* thd, TABLE_LIST* tables)
   */
   DBUG_RETURN(mysql_admin_table(thd, tables, 0,
 				"preload_keys", TL_READ_NO_INSERT, 0, 0, 0, 0,
-				&handler::preload_keys, 0));
+				&handler::preload_keys, 0, false));
 }
 
 
@@ -1315,15 +1326,7 @@ bool Sql_cmd_analyze_table::execute(THD *thd)
   WSREP_TO_ISOLATION_BEGIN_WRTCHK(NULL, NULL, first_table);
   res= mysql_admin_table(thd, first_table, &m_lex->check_opt,
                          "analyze", lock_type, 1, 0, 0, 0,
-                         &handler::ha_analyze, 0);
-  /* ! we write after unlocking the table */
-  if (!res && !m_lex->no_write_to_binlog)
-  {
-    /*
-      Presumably, ANALYZE and binlog writing doesn't require synchronization
-    */
-    res= write_bin_log(thd, TRUE, thd->query(), thd->query_length());
-  }
+                         &handler::ha_analyze, 0, true);
   m_lex->select_lex.table_list.first= first_table;
   m_lex->query_tables= first_table;
 
@@ -1346,7 +1349,7 @@ bool Sql_cmd_check_table::execute(THD *thd)
     goto error; /* purecov: inspected */
   res= mysql_admin_table(thd, first_table, &m_lex->check_opt, "check",
                          lock_type, 0, 0, HA_OPEN_FOR_REPAIR, 0,
-                         &handler::ha_check, &view_check);
+                         &handler::ha_check, &view_check, false);
 
   m_lex->select_lex.table_list.first= first_table;
   m_lex->query_tables= first_table;
@@ -1371,15 +1374,7 @@ bool Sql_cmd_optimize_table::execute(THD *thd)
     mysql_recreate_table(thd, first_table, true) :
     mysql_admin_table(thd, first_table, &m_lex->check_opt,
                       "optimize", TL_WRITE, 1, 0, 0, 0,
-                      &handler::ha_optimize, 0);
-  /* ! we write after unlocking the table */
-  if (!res && !m_lex->no_write_to_binlog)
-  {
-    /*
-      Presumably, OPTIMIZE and binlog writing doesn't require synchronization
-    */
-    res= write_bin_log(thd, TRUE, thd->query(), thd->query_length());
-  }
+                      &handler::ha_optimize, 0, true);
   m_lex->select_lex.table_list.first= first_table;
   m_lex->query_tables= first_table;
 
@@ -1404,16 +1399,8 @@ bool Sql_cmd_repair_table::execute(THD *thd)
                          TL_WRITE, 1,
                          MY_TEST(m_lex->check_opt.sql_flags & TT_USEFRM),
                          HA_OPEN_FOR_REPAIR, &prepare_for_repair,
-                         &handler::ha_repair, &view_repair);
+                         &handler::ha_repair, &view_repair, true);
 
-  /* ! we write after unlocking the table */
-  if (!res && !m_lex->no_write_to_binlog)
-  {
-    /*
-      Presumably, REPAIR and binlog writing doesn't require synchronization
-    */
-    res= write_bin_log(thd, TRUE, thd->query(), thd->query_length());
-  }
   m_lex->select_lex.table_list.first= first_table;
   m_lex->query_tables= first_table;
 

sql/sql_class.h

@@ -4530,7 +4530,8 @@ class THD :public Statement,
     transaction.all.m_unsafe_rollback_flags|=
       (transaction.stmt.m_unsafe_rollback_flags &
        (THD_TRANS::DID_WAIT | THD_TRANS::CREATED_TEMP_TABLE |
-        THD_TRANS::DROPPED_TEMP_TABLE | THD_TRANS::DID_DDL));
+        THD_TRANS::DROPPED_TEMP_TABLE | THD_TRANS::DID_DDL |
+        THD_TRANS::EXECUTED_TABLE_ADMIN_CMD));
   }
   /*
     Reset current_linfo