MDEV-26: Global transaction ID.

Add tests crashing the slave in the middle of replication and checking that
replication picks-up again on restart in a crash-safe way.

Fix silly code that causes crash by inserting uninitialised data into a hash.

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

@@ -56,5 +56,36 @@ a
 1
 2
 3
+*** Test crashing slave at various points and check that it recovers crash-safe. ***
+include/stop_slave.inc
+SET GLOBAL debug_dbug="+d,inject_crash_before_write_rpl_slave_state";
+START SLAVE;
+INSERT INTO t1 VALUES (4);
+include/stop_slave.inc
+SET GLOBAL debug_dbug="+d,crash_commit_before";
+START SLAVE;
+INSERT INTO t1 VALUES (5);
+include/stop_slave.inc
+SET GLOBAL debug_dbug="+d,crash_commit_after";
+START SLAVE;
+INSERT INTO t1 VALUES (6);
+include/stop_slave.inc
+SET GLOBAL debug_dbug="+d,inject_crash_before_flush_rli";
+START SLAVE;
+INSERT INTO t1 VALUES (7);
+include/stop_slave.inc
+SET GLOBAL debug_dbug="+d,inject_crash_after_flush_rli";
+START SLAVE;
+INSERT INTO t1 VALUES (8);
+SELECT * FROM t1 ORDER BY a;
+a
+1
+2
+3
+4
+5
+6
+7
+8
 DROP TABLE t1;
 include/rpl_end.inc

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

@@ -124,6 +124,140 @@ SHOW BINLOG EVENTS IN 'master-bin.000004' LIMIT 1,1;
 --sync_with_master
 SELECT * FROM t1 ORDER BY a;
 
+
+--echo *** Test crashing slave at various points and check that it recovers crash-safe. ***
+
+# Crash the slave just before updating mysql.rpl_slave_state table.
+--source include/stop_slave.inc
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+wait
+EOF
+# We do not have to save @@GLOBAL.debug_dbug, it is reset when slave crashes.
+SET GLOBAL debug_dbug="+d,inject_crash_before_write_rpl_slave_state";
+START SLAVE;
+
+--connection server_1
+INSERT INTO t1 VALUES (4);
+
+--connection server_2
+--source include/wait_until_disconnected.inc
+
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+restart: --skip-slave-start=0
+EOF
+
+--enable_reconnect
+--source include/wait_until_connected_again.inc
+
+--let $wait_condition= SELECT COUNT(*) = 4 FROM t1
+--source include/wait_condition.inc
+
+# Crash the slave just before committing.
+--source include/stop_slave.inc
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+wait
+EOF
+SET GLOBAL debug_dbug="+d,crash_commit_before";
+START SLAVE;
+
+--connection server_1
+INSERT INTO t1 VALUES (5);
+
+--connection server_2
+--source include/wait_until_disconnected.inc
+
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+restart: --skip-slave-start=0
+EOF
+
+--enable_reconnect
+--source include/wait_until_connected_again.inc
+
+--let $wait_condition= SELECT COUNT(*) = 5 FROM t1
+--source include/wait_condition.inc
+
+# Crash the slave just after committing.
+--source include/stop_slave.inc
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+wait
+EOF
+SET GLOBAL debug_dbug="+d,crash_commit_after";
+START SLAVE;
+
+--connection server_1
+INSERT INTO t1 VALUES (6);
+
+--connection server_2
+--source include/wait_until_disconnected.inc
+
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+restart: --skip-slave-start=0
+EOF
+
+--enable_reconnect
+--source include/wait_until_connected_again.inc
+
+--let $wait_condition= SELECT COUNT(*) = 6 FROM t1
+--source include/wait_condition.inc
+
+# Crash the slave just before updating relay-log.info
+--source include/stop_slave.inc
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+wait
+EOF
+SET GLOBAL debug_dbug="+d,inject_crash_before_flush_rli";
+START SLAVE;
+
+--connection server_1
+INSERT INTO t1 VALUES (7);
+
+--connection server_2
+--source include/wait_until_disconnected.inc
+
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+restart: --skip-slave-start=0
+EOF
+
+--enable_reconnect
+--source include/wait_until_connected_again.inc
+
+--let $wait_condition= SELECT COUNT(*) = 7 FROM t1
+--source include/wait_condition.inc
+
+# Crash the slave just after updating relay-log.info
+--source include/stop_slave.inc
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+wait
+EOF
+SET GLOBAL debug_dbug="+d,inject_crash_after_flush_rli";
+START SLAVE;
+
+--connection server_1
+INSERT INTO t1 VALUES (8);
+
+--connection server_2
+--source include/wait_until_disconnected.inc
+
+--remove_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+--write_file $MYSQLTEST_VARDIR/tmp/mysqld.2.expect
+restart: --skip-slave-start=0
+EOF
+
+--enable_reconnect
+--source include/wait_until_connected_again.inc
+
+--let $wait_condition= SELECT COUNT(*) = 8 FROM t1
+--source include/wait_condition.inc
+
+
+# Check that everything was replicated correctly.
+SELECT * FROM t1 ORDER BY a;
+
+
 --connection server_1
 DROP TABLE t1;
 

sql/rpl_gtid.cc

@@ -329,6 +329,7 @@ rpl_slave_state::record_gtid(THD *thd, const rpl_gtid *gtid, uint64 sub_id,
   table->field[1]->store(sub_id, true);
   table->field[2]->store((ulonglong)gtid->server_id, true);
   table->field[3]->store(gtid->seq_no, true);
+  DBUG_EXECUTE_IF("inject_crash_before_write_rpl_slave_state", DBUG_SUICIDE(););
   if ((err= table->file->ha_write_row(table->record[0])))
       goto end;
 

sql/rpl_rli.cc

@@ -1246,7 +1246,9 @@ void Relay_log_info::stmt_done(my_off_t event_master_log_pos,
         DBA aware of the problem in the error log.
       */
     }
+    DBUG_EXECUTE_IF("inject_crash_before_flush_rli", DBUG_SUICIDE(););
     flush_relay_log_info(this);
+    DBUG_EXECUTE_IF("inject_crash_after_flush_rli", DBUG_SUICIDE(););
     /*
       Note that Rotate_log_event::do_apply_event() does not call this
       function, so there is no chance that a fake rotate event resets
@@ -1453,6 +1455,10 @@ rpl_load_gtid_slave_state(THD *thd)
       entry= (struct local_element *)rec;
       if (entry->sub_id >= sub_id)
         continue;
+      entry->sub_id= sub_id;
+      DBUG_ASSERT(entry->gtid.domain_id == domain_id);
+      entry->gtid.server_id= server_id;
+      entry->gtid.seq_no= seq_no;
     }
     else
     {
@@ -1462,17 +1468,16 @@ rpl_load_gtid_slave_state(THD *thd)
         err= 1;
         goto end;
       }
+      entry->sub_id= sub_id;
+      entry->gtid.domain_id= domain_id;
+      entry->gtid.server_id= server_id;
+      entry->gtid.seq_no= seq_no;
       if ((err= my_hash_insert(&hash, (uchar *)entry)))
       {
         my_free(entry);
         goto end;
       }
     }
-
-    entry->sub_id= sub_id;
-    entry->gtid.domain_id= domain_id;
-    entry->gtid.server_id= server_id;
-    entry->gtid.seq_no= seq_no;
   }
 
   rpl_global_gtid_slave_state.lock();