MDEV-30423 Deadlock on Replica during BACKUP STAGE BLOCK_COMMIT on XA…

… transactions The user XA commit execution branch was caught not have been covered with MDEV-21953 fixes. The XA involved deadlock is resolved now to apply the former fixes pattern. Along the fixes the following changes have been implemented. - MDL lock attribute correction - dissociation of the externally completed XA from the current thread's xid_state in the error branches - cleanup_context() preseves the prepared XA - wait_for_prior_commit() is relocated to satisfy both the binlog ON (log-slave-updates and skip-log-bin) and OFF slave execution branches.
MariaDB · Jan 23, 2023 · dc646c2 · dc646c2
1 parent 647a723
commit dc646c2
Show file tree

Hide file tree

Showing 17 changed files with 784 additions and 35 deletions.
diff --git a/extra/wolfssl/wolfssl b/extra/wolfssl/wolfssl
diff --git a/libmariadb b/libmariadb
diff --git a/mysql-test/include/master-slave.inc b/mysql-test/include/master-slave.inc
@@ -11,6 +11,7 @@
 #   [--let $rpl_skip_start_slave= 1]
 #   [--let $rpl_debug= 1]
 #   [--let $slave_timeout= NUMBER]
+#   [--let $rpl_server_skip_log_bin= 1]
 #   --source include/master-slave.inc
 #
 # Parameters:

diff --git a/mysql-test/include/rpl_init.inc b/mysql-test/include/rpl_init.inc
@@ -73,6 +73,7 @@
 #      before CHANGE MASTER and START SLAVE.  RESET MASTER and RESET
 #      SLAVE are suppressed if $rpl_skip_reset_master_and_slave is
 #      set.
+#      Also see $rpl_server_skip_log_bin.
 #
 #    $rpl_skip_change_master
 #      By default, this script issues CHANGE MASTER so that all slaves
@@ -94,6 +95,10 @@
 #      Timeout used when waiting for the slave threads to start.
 #      See include/wait_for_slave_param.inc
 #
+#    $rpl_server_skip_log_bin
+#      When $rpl_skip_reset_master_and_slave is not set
+#      RESET MASTER does not report ER_FLUSH_MASTER_BINLOG_CLOSED
+#      on any server.
 #
 # ==== Side effects ====
 #
@@ -161,7 +166,16 @@ while ($_rpl_server)
   USE test;
   if (!$rpl_skip_reset_master_and_slave)
   {
-    RESET MASTER;
+    if (!$rpl_server_skip_log_bin)
+    {
+     --error 0
+       RESET MASTER;
+    }
+    if ($rpl_server_skip_log_bin)
+    {
+     --error 0,ER_FLUSH_MASTER_BINLOG_CLOSED
+       RESET MASTER;
+    }
     SET GLOBAL gtid_slave_pos= "";
     RESET SLAVE;
   }

diff --git a/mysql-test/suite/rpl/r/parallel_backup.result b/mysql-test/suite/rpl/r/parallel_backup.result
@@ -7,6 +7,8 @@ include/master-slave.inc
 connection master;
 CREATE TABLE t1 (a INT PRIMARY KEY) ENGINE = innodb;
 connection slave;
+call mtr.add_suppression("Deadlock found when trying to get lock");
+call mtr.add_suppression("Commit failed due to failure of an earlier commit");
 include/stop_slave.inc
 SET @old_parallel_threads= @@GLOBAL.slave_parallel_threads;
 SET @old_parallel_mode   = @@GLOBAL.slave_parallel_mode;
@@ -30,6 +32,167 @@ connection backup_slave;
 BACKUP STAGE END;
 connection slave;
 include/diff_tables.inc [master:t1,slave:t1]
+# MDEV-30423: dealock XA COMMIT vs BACKUP
+#
+# Normal XA COMMIT
+connection slave;
+include/stop_slave.inc
+connection master;
+connection aux_slave;
+BEGIN;
+INSERT INTO t1 VALUES (4);
+connection master;
+XA START '1';
+INSERT INTO  t1 VALUES (3);
+XA END '1';
+XA PREPARE '1';
+connection master1;
+INSERT INTO t1 VALUES (4);
+connection master;
+XA COMMIT '1';
+include/save_master_gtid.inc
+connection slave;
+include/start_slave.inc
+connection aux_slave;
+# Xid '1' must be in the output:
+XA RECOVER;
+formatID	gtrid_length	bqual_length	data
+1	1	0	1
+connection backup_slave;
+BACKUP STAGE START;
+BACKUP STAGE BLOCK_COMMIT;
+connection aux_slave;
+ROLLBACK;
+connection backup_slave;
+BACKUP STAGE END;
+connection slave;
+include/sync_with_master_gtid.inc
+include/stop_slave.inc
+#
+# Normal XA ROLLBACK
+connection slave;
+include/stop_slave.inc
+Warnings:
+Note	1255	Slave already has been stopped
+connection master;
+connection aux_slave;
+BEGIN;
+INSERT INTO t1 VALUES (6);
+connection master;
+XA START '1';
+INSERT INTO  t1 VALUES (5);
+XA END '1';
+XA PREPARE '1';
+connection master1;
+INSERT INTO t1 VALUES (6);
+connection master;
+XA ROLLBACK '1';
+include/save_master_gtid.inc
+connection slave;
+include/start_slave.inc
+connection aux_slave;
+# Xid '1' must be in the output:
+XA RECOVER;
+formatID	gtrid_length	bqual_length	data
+1	1	0	1
+connection backup_slave;
+BACKUP STAGE START;
+BACKUP STAGE BLOCK_COMMIT;
+connection aux_slave;
+ROLLBACK;
+connection backup_slave;
+BACKUP STAGE END;
+connection slave;
+include/sync_with_master_gtid.inc
+include/stop_slave.inc
+#
+# Errored out XA COMMIT
+connection slave;
+include/stop_slave.inc
+Warnings:
+Note	1255	Slave already has been stopped
+connection master;
+connection aux_slave;
+BEGIN;
+INSERT INTO t1 VALUES (8);
+connection master;
+XA START '1';
+INSERT INTO  t1 VALUES (7);
+XA END '1';
+XA PREPARE '1';
+connection master1;
+INSERT INTO t1 VALUES (8);
+connection master;
+XA COMMIT '1';
+include/save_master_gtid.inc
+connection slave;
+SET @sav_innodb_lock_wait_timeout  = @@global.innodb_lock_wait_timeout;
+SET @sav_slave_transaction_retries = @@global.slave_transaction_retries;
+SET @@global.innodb_lock_wait_timeout =1;
+SET @@global.slave_transaction_retries=0;
+include/start_slave.inc
+connection aux_slave;
+# Xid '1' must be in the output:
+XA RECOVER;
+formatID	gtrid_length	bqual_length	data
+1	1	0	1
+connection backup_slave;
+BACKUP STAGE START;
+BACKUP STAGE BLOCK_COMMIT;
+connection aux_slave;
+ROLLBACK;
+connection backup_slave;
+BACKUP STAGE END;
+connection slave;
+include/stop_slave.inc
+SET @@global.innodb_lock_wait_timeout = @sav_innodb_lock_wait_timeout;
+SET @@global.slave_transaction_retries= @sav_slave_transaction_retries;
+connection slave;
+include/start_slave.inc
+include/sync_with_master_gtid.inc
+#
+# Errored out XA ROLLBACK
+connection slave;
+include/stop_slave.inc
+connection master;
+connection aux_slave;
+BEGIN;
+INSERT INTO t1 VALUES (10);
+connection master;
+XA START '1';
+INSERT INTO  t1 VALUES (9);
+XA END '1';
+XA PREPARE '1';
+connection master1;
+INSERT INTO t1 VALUES (10);
+connection master;
+XA ROLLBACK '1';
+include/save_master_gtid.inc
+connection slave;
+SET @sav_innodb_lock_wait_timeout  = @@global.innodb_lock_wait_timeout;
+SET @sav_slave_transaction_retries = @@global.slave_transaction_retries;
+SET @@global.innodb_lock_wait_timeout =1;
+SET @@global.slave_transaction_retries=0;
+include/start_slave.inc
+connection aux_slave;
+# Xid '1' must be in the output:
+XA RECOVER;
+formatID	gtrid_length	bqual_length	data
+1	1	0	1
+connection backup_slave;
+BACKUP STAGE START;
+BACKUP STAGE BLOCK_COMMIT;
+connection aux_slave;
+ROLLBACK;
+connection backup_slave;
+BACKUP STAGE END;
+connection slave;
+include/stop_slave.inc
+SET @@global.innodb_lock_wait_timeout = @sav_innodb_lock_wait_timeout;
+SET @@global.slave_transaction_retries= @sav_slave_transaction_retries;
+connection slave;
+include/start_slave.inc
+include/sync_with_master_gtid.inc
 connection slave;
 include/stop_slave.inc
 SET @@global.slave_parallel_threads= @old_parallel_threads;