Invalid deadlock error when inserting concurrently into an AUTO_INCREMENT column

[Philip Stoev]

Concurrent INSERTs with an auto_increment column produce an unexpected "Deadlock found when trying to get lock; try restarting transaction" error.

This does not happen if the column is not auto_increment.

Test case:

--source include/galera_cluster.inc
--source include/have_innodb.inc

# Create a second connection to node1 so that we can run transactions concurrently
--let $galera_connection_name = node_1a
--let $galera_server_number = 1
--source include/galera_connect.inc

--connection node_1
SET GLOBAL wsrep_sync_wait = 7;
CREATE TABLE ten (f1 INTEGER);
INSERT INTO ten VALUES (1),(2),(3),(4),(5),(6),(7),(8),(9),(10);

CREATE TABLE t1 (f1 INTEGER AUTO_INCREMENT PRIMARY KEY, f2 INTEGER) Engine=InnoDB;
--connection node_2
SET GLOBAL wsrep_sync_wait = 7;

--connection node_1
--send INSERT INTO t1 (f2) SELECT 1 FROM ten AS a1, ten AS a2, ten AS a3;

--connection node_1a
--send INSERT INTO t1 (f2) SELECT 1 FROM ten AS a1, ten AS a2, ten AS a3;

--connection node_2
--send INSERT INTO t1 (f2) SELECT 1 FROM ten AS a1, ten AS a2, ten AS a3;

--connection node_1
--reap

--connection node_1a
--reap

--connection node_2
--reap
SELECT COUNT(*) = 30000 FROM t1;
SELECT COUNT(DISTINCT f1) = 30000 FROM t1;
SELECT COUNT(*) = 5 FROM INFORMATION_SCHEMA.PROCESSLIST WHERE USER = 'system user';

SET GLOBAL wsrep_slave_threads = 1;
SET GLOBAL wsrep_sync_wait = 0;

--connection default
SET GLOBAL wsrep_sync_wait = 0;
DROP TABLE t1;
DROP TABLE ten;