MDEV-38218 : Galera test failure on galera_bf_abort_flush_for_export

janlindstrom · janlindstrom · commit e40277d29b7c · 2026-01-20T10:23:44.000+02:00
Problem was in wsrep_handle_mdl_conflict function was comparing
thd-&gt;lex-&gt;sql_command variable for granted MDL-lock.

There is two possible schedules:

    (1) FLUSH TABLES ... FOR EXPORT that will take MDL-lock (granted_thd).
        INSERT from other node is conflicting operation (request_thd)
        and sees MDL-conflict. Because granted_thd has not executed anything
        else thd-&gt;lex-&gt;sql_command == SQLCOM_FLUSH and this case was
        correctly handled in wsrep_handle_mdl_conflict i.e. INSERT needs
        to wait.

    (2) FLUSH TABLES ... FOR EXPORT that will take MDL-lock (granted_thd).
        SET SESSION wsrep_sync_wait=0; (granted_thd)
        INSERT from other node is conflicting operation (request_thd)

        However, thd-&gt;lex-&gt;sql_command is not stored to taken MDL-lock. Now
        as granted_thd is executing SET thd-&gt;lex-&gt;sql_command != SQLCOM_FLUSH
        and INSERT that is BF will abort it and that means also FTFE is
        killed and MDL-lock relesed. This is incorrect as FTFE has written
        file on filesystem and it can't be really killed.

In this fix wsrep_handle_mdl_conflict is refactored not to use
thd-&gt;lex-&gt;sql_command as a variable used for decisions. Instead
connection state can be determined also via THD members. E.g.:

    * wsrep_thd_is_toi() || wsrep_thd_is_applying - ongoing TOI or applier
    * wsrep_thd_is_BF - thread is brute force
    * wsrep_thd_is_SR - thread is streaming replication thread
    * thd-&gt;current_backup_stage != BACKUP_FINISHED - there's ongoing BACKUP
    * thd-&gt;global_read_lock.is_acquired() - ongoing FTWRL
    * thd-&gt;locked_tables_mode == LTM_LOCK_TABLES - ongoing FTFE or LOCK TABLES
diff --git a/mysql-test/suite/galera/r/MDEV-22232.result b/mysql-test/suite/galera/r/MDEV-22232.result
@@ -9,7 +9,7 @@ connection node_1;
 SET DEBUG_SYNC = 'now WAIT_FOR may_run';
 TRUNCATE TABLE t1;
 connection con1;
-ERROR 70100: Query execution was interrupted
+Got one of the listed errors
 SET DEBUG_SYNC = 'RESET';
 --- CTAS with non-empty result set ---
 INSERT INTO t1 VALUES (10), (20), (30);
@@ -19,7 +19,7 @@ connection node_1;
 SET DEBUG_SYNC = 'now WAIT_FOR may_run';
 TRUNCATE TABLE t1;
 connection con1;
-ERROR 70100: Query execution was interrupted
+Got one of the listed errors
 SET DEBUG_SYNC = 'RESET';
 DROP TABLE t1;
 disconnect con1;
diff --git a/mysql-test/suite/galera/r/MDEV-32938.result b/mysql-test/suite/galera/r/MDEV-32938.result
@@ -3,16 +3,15 @@ connection node_1;
 connect con1,127.0.0.1,root,,test,$NODE_MYPORT_1;
 call mtr.add_suppression("WSREP: ALTER TABLE isolation failure");
 CREATE TABLE t1(c1 INT PRIMARY KEY, c2 INT) ENGINE=InnoDB;
-SET DEBUG_SYNC = 'wsrep_append_fk_toi_keys_before_close_tables SIGNAL may_alter WAIT_FOR bf_abort';
+SET DEBUG_SYNC = 'wsrep_append_fk_toi_keys_before_close_tables SIGNAL may_alter';
 ALTER TABLE t1 DROP COLUMN c2;
 connection node_1;
 SET DEBUG_SYNC = 'now WAIT_FOR may_alter';
 ALTER TABLE t1 ADD COLUMN c3 INT;
 connection con1;
-ERROR 70100: Query execution was interrupted
-INSERT INTO t1 (c1, c2, c3) VALUES (1, 0, 0);
+INSERT INTO t1 (c1, c3) VALUES (1, 0);
 connection node_2;
-INSERT INTO t1 (c1, c2, c3) VALUES (2, 0, 0);
+INSERT INTO t1 (c1, c3) VALUES (2, 0);
 connection node_1;
 SET DEBUG_SYNC = 'RESET';
 DROP TABLE t1;
diff --git a/mysql-test/suite/galera/r/galera_as_slave_ctas.result b/mysql-test/suite/galera/r/galera_as_slave_ctas.result
@@ -24,8 +24,11 @@ connection node_3;
 DROP TABLE source;
 DROP TABLE target;
 connection node_2;
+call mtr.add_suppression("WSREP: Inconsistency detected: Inconsistent by consensus on");
+call mtr.add_suppression("WSREP: Failed to apply write set: gtid:");
 connection node_1;
 STOP SLAVE;
 RESET SLAVE ALL;
+call mtr.add_suppression("WSREP: unexpected async replication event: 0");
 connection node_3;
 RESET MASTER;
diff --git a/mysql-test/suite/galera/r/galera_bf_abort.result b/mysql-test/suite/galera/r/galera_bf_abort.result
@@ -12,6 +12,9 @@ connection node_2a;
 connection node_2;
 INSERT INTO t1 VALUES (2, 'node_2');
 ERROR 40001: Deadlock found when trying to get lock; try restarting transaction
+COMMIT;
 wsrep_local_aborts_increment
 1
 DROP TABLE t1;
+disconnect node_2;
+disconnect node_1;
diff --git a/mysql-test/suite/galera/r/galera_create_table_as_select.result b/mysql-test/suite/galera/r/galera_create_table_as_select.result
@@ -88,7 +88,7 @@ COUNT(*) = 5
 1
 CREATE TABLE t1 AS SELECT * FROM t2;
 connection node_1;
-ERROR 70100: Query execution was interrupted
+Got one of the listed errors
 SET GLOBAL DEBUG_DBUG = '';
 SET DEBUG_SYNC = 'RESET';
 DROP TABLE t1, t2;
diff --git a/mysql-test/suite/galera/suite.pm b/mysql-test/suite/galera/suite.pm
@@ -68,6 +68,7 @@ push @::global_suppressions,
      qr|WSREP: Trying to continue unpaused monitor|,
      qr|WSREP: Wait for gtid returned error 3 while waiting for prior transactions to commit before setting position|,
      qr|WSREP: Failed to report last committed|,
+     qr|bytes freed by|,
    );
 
 sub which($) { return `sh -c "command -v $_[0]"` }
diff --git a/mysql-test/suite/galera/t/MDEV-22232.test b/mysql-test/suite/galera/t/MDEV-22232.test
@@ -31,7 +31,7 @@ TRUNCATE TABLE t1;
 
 --connection con1
 # CTAS gets BF aborted.
---error ER_QUERY_INTERRUPTED
+--error ER_QUERY_INTERRUPTED, ER_LOCK_DEADLOCK
 --reap
 
 # Cleanup
@@ -56,7 +56,7 @@ TRUNCATE TABLE t1;
 
 --connection con1
 # CTAS gets BF aborted.
---error ER_QUERY_INTERRUPTED
+--error ER_QUERY_INTERRUPTED, ER_LOCK_DEADLOCK
 --reap
 
 # Cleanup
diff --git a/mysql-test/suite/galera/t/MDEV-32938.test b/mysql-test/suite/galera/t/MDEV-32938.test
@@ -14,7 +14,9 @@
 # - after entering TOI mode, it would later find it'd been killed, and complete with an error
 # - at the same time, the command would successfully apply on every other node except the initiator.
 #
-# Fixed by checking killed state on THD before entering TOI.
+#
+# NOTE : After MDL-lock conflict resolution rework this test behaviour has
+# changed.
 #
 
 --source include/galera_cluster.inc
@@ -28,26 +30,24 @@ call mtr.add_suppression("WSREP: ALTER TABLE isolation failure");
 CREATE TABLE t1(c1 INT PRIMARY KEY, c2 INT) ENGINE=InnoDB;
 
 # Run ALTER DROP COLUMN and hang before closing tables on adding FK keys and before entering TOI.
-# Wait until it gets BF-aborted.
-SET DEBUG_SYNC = 'wsrep_append_fk_toi_keys_before_close_tables SIGNAL may_alter WAIT_FOR bf_abort';
+SET DEBUG_SYNC = 'wsrep_append_fk_toi_keys_before_close_tables SIGNAL may_alter';
 --send
   ALTER TABLE t1 DROP COLUMN c2;
 
 --connection node_1
-# Run ALTER ADD COLUMN and BF-abort the previous ALTER DROP COLUMN.
+# Run ALTER ADD COLUMN
 SET DEBUG_SYNC = 'now WAIT_FOR may_alter';
 ALTER TABLE t1 ADD COLUMN c3 INT;
 
 --connection con1
-# ALTER DROP COLUMN gets BF aborted.
---error ER_QUERY_INTERRUPTED
+# ALTER DROP COLUMN waits
 --reap
 
-INSERT INTO t1 (c1, c2, c3) VALUES (1, 0, 0);
+INSERT INTO t1 (c1, c3) VALUES (1, 0);
 
 --connection node_2
-# ALTER DROP COLUMN must not be replicated.
-INSERT INTO t1 (c1, c2, c3) VALUES (2, 0, 0);
+# ALTER DROP COLUMN must be replicated.
+INSERT INTO t1 (c1, c3) VALUES (2, 0);
 
 # Cleanup.
 --connection node_1
diff --git a/mysql-test/suite/galera/t/galera_as_slave_ctas.test b/mysql-test/suite/galera/t/galera_as_slave_ctas.test
@@ -66,10 +66,13 @@ DROP TABLE target;
 --connection node_2
 --let $wait_condition = SELECT COUNT(*) = 0 FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_NAME = 'target';
 --source include/wait_condition.inc
+call mtr.add_suppression("WSREP: Inconsistency detected: Inconsistent by consensus on");
+call mtr.add_suppression("WSREP: Failed to apply write set: gtid:");
 
 --connection node_1
 STOP SLAVE;
 RESET SLAVE ALL;
+call mtr.add_suppression("WSREP: unexpected async replication event: 0");
 
 --connection node_3
 RESET MASTER;
diff --git a/mysql-test/suite/galera/t/galera_bf_abort.test b/mysql-test/suite/galera/t/galera_bf_abort.test
@@ -24,6 +24,7 @@ INSERT INTO t1 VALUES (1,'node_1');
 --connection node_2
 --error ER_LOCK_DEADLOCK
 INSERT INTO t1 VALUES (2, 'node_2');
+COMMIT;
 
 --let $wsrep_local_bf_aborts_after = `SELECT VARIABLE_VALUE FROM INFORMATION_SCHEMA.GLOBAL_STATUS WHERE VARIABLE_NAME = 'wsrep_local_bf_aborts'`
 
@@ -32,3 +33,4 @@ INSERT INTO t1 VALUES (2, 'node_2');
 --enable_query_log
 
 DROP TABLE t1;
+--source include/galera_end.inc
diff --git a/mysql-test/suite/galera/t/galera_create_table_as_select.test b/mysql-test/suite/galera/t/galera_create_table_as_select.test
@@ -121,7 +121,7 @@ SELECT COUNT(*) = 5 FROM t2;
 CREATE TABLE t1 AS SELECT * FROM t2;
 
 --connection node_1
---error ER_QUERY_INTERRUPTED
+--error ER_QUERY_INTERRUPTED,ER_LOCK_DEADLOCK
 --reap
 
 SET GLOBAL DEBUG_DBUG = '';
diff --git a/sql/service_wsrep.cc b/sql/service_wsrep.cc
@@ -226,6 +226,11 @@ extern "C" my_bool wsrep_thd_bf_abort(THD *bf_thd, THD *victim_thd,
    */
   if ((ret || !wsrep_on(victim_thd)) && signal)
   {
+    WSREP_DEBUG("wsrep_thd_bf_abort thread %ld sending "
+                "KILL_QUERY_HARD to %ld ret=%d signal %d",
+                bf_thd->thread_id,
+                victim_thd->thread_id,
+                ret, signal);
     victim_thd->wsrep_aborter= bf_thd->thread_id;
     victim_thd->awake_no_mutex(KILL_QUERY_HARD);
   } else {
diff --git a/sql/sql_class.cc b/sql/sql_class.cc
@@ -2182,6 +2182,10 @@ int THD::killed_errno()
     DBUG_RETURN(ER_CONNECTION_KILLED);
   case KILL_QUERY:
   case KILL_QUERY_HARD:
+#ifdef WITH_WSREP
+  if (WSREP(this))
+    wsrep_report_query_interrupted(this, __FILE__, __LINE__);
+#endif /* WITH_WSREP */
     DBUG_RETURN(ER_QUERY_INTERRUPTED);
   case KILL_TIMEOUT:
   case KILL_TIMEOUT_HARD:
@@ -8352,6 +8356,10 @@ wait_for_commit::wait_for_prior_commit2(THD *thd, bool allow_kill)
     wakeup_error= ER_QUERY_INTERRUPTED;
   my_message(wakeup_error, ER_THD(thd, wakeup_error), MYF(0));
   thd->EXIT_COND(&old_stage);
+#ifdef WITH_WSREP
+  if (WSREP(thd))
+    wsrep_report_query_interrupted(thd, __FILE__, __LINE__);
+#endif /* WITH_WSREP */
   /*
     Must do the DEBUG_SYNC() _after_ exit_cond(), as DEBUG_SYNC is not safe to
     use within enter_cond/exit_cond.
diff --git a/sql/wsrep_mysqld.cc b/sql/wsrep_mysqld.cc
diff --git a/sql/wsrep_mysqld.h b/sql/wsrep_mysqld.h
diff --git a/sql/wsrep_thd.cc b/sql/wsrep_thd.cc
diff --git a/sql/wsrep_thd.h b/sql/wsrep_thd.h

Original file line number	Diff line number	Diff line change
`@@ -12,6 +12,9 @@ connection node_2a;`
`12`	`12`	`connection node_2;`
`13`	`13`	`INSERT INTO t1 VALUES (2, 'node_2');`
`14`	`14`	`ERROR 40001: Deadlock found when trying to get lock; try restarting transaction`
	`15`	`+COMMIT;`
`15`	`16`	`wsrep_local_aborts_increment`
`16`	`17`	`1`
`17`	`18`	`DROP TABLE t1;`
	`19`	`+disconnect node_2;`
	`20`	`+disconnect node_1;`
Original file line number	Diff line number	Diff line change
`@@ -88,7 +88,7 @@ COUNT(*) = 5`
`88`	`88`	`1`
`89`	`89`	`CREATE TABLE t1 AS SELECT * FROM t2;`
`90`	`90`	`connection node_1;`
`91`		`-ERROR 70100: Query execution was interrupted`
	`91`	`+Got one of the listed errors`
`92`	`92`	`SET GLOBAL DEBUG_DBUG = '';`
`93`	`93`	`SET DEBUG_SYNC = 'RESET';`
`94`	`94`	`DROP TABLE t1, t2;`