Details
-
Bug
-
Status: Open (View Workflow)
-
Critical
-
Resolution: Unresolved
-
12.1
-
Can result in unexpected behaviour
Description
A regression in 12.1 for the error 123, however the pseudo_slave_mode/XA bug exists in all versions. We can split the bug if needed.
SET pseudo_slave_mode=1; |
CREATE TABLE t (c1 INT,c2 INT AUTO_INCREMENT PRIMARY KEY,KEY(c1)) ENGINE=InnoDB; |
XA START 'a'; |
INSERT INTO t VALUES (0,0); |
XA END 'a'; |
XA PREPARE 'a'; |
XA START 'b'; |
SELECT * FROM t; |
INSERT INTO t SELECT * FROM t; |
Leads to:
|
CS 12.1.0 4b79d7b8ee557d53a859aedec839b8673585b514 (Optimized) Build 06/06/2025, CLI |
12.1.0-opt>INSERT INTO t SELECT * FROM t;
|
ERROR 1020 (HY000): Record has changed since last read in table 't'
|
|
CS 12.1.0 4b79d7b8ee557d53a859aedec839b8673585b514 (Optimized) Build 06/06/2025, Error log |
2025-06-07 20:30:35 4 [ERROR] Got error 123 when reading table './test/t'
|
In contrast with 12.1 (immediate error), in 10.6 we see the query remaining in Sending data state:
|
CS 10.6.22 643319a7fb1e273797c2a1e46d76cfac0fa1da8f (Optimized) Build 06/06/2025 |
10.6.22-opt>SHOW PROCESSLIST;
|
+----+------+-----------+------+---------+------+--------------+-------------------------------+----------+
|
| Id | User | Host | db | Command | Time | State | Info | Progress |
|
+----+------+-----------+------+---------+------+--------------+-------------------------------+----------+
|
| 4 | root | localhost | test | Query | 49 | Sending data | INSERT INTO t SELECT * FROM t | 0.000 |
|
| 5 | root | localhost | test | Query | 0 | starting | SHOW PROCESSLIST | 0.000 |
|
+----+------+-----------+------+---------+------+--------------+-------------------------------+----------+
|
2 rows in set (0.000 sec)
|
Till we get a lock timeout:
|
CS 10.6.22 643319a7fb1e273797c2a1e46d76cfac0fa1da8f (Optimized) Build 06/06/2025 |
10.6.22-opt>INSERT INTO t SELECT * FROM t;
|
ERROR 1205 (HY000): Lock wait timeout exceeded; try restarting transaction
|
There are many oddities besides. For example, XA START 'b'; should fail as a PREPARE is in progress. And this indeed is the case when we remove enabling pseudo slave mode (in all versions):
|
CS 12.1.0 4b79d7b8ee557d53a859aedec839b8673585b514 (Optimized) Build 06/06/2025 |
12.1.0-opt>XA START 'b';
|
ERROR 1399 (XAE07): XAER_RMFAIL: The command cannot be executed when global transaction is in the PREPARED state
|
12.1.0-opt>SELECT * FROM t;
|
ERROR 1399 (XAE07): XAER_RMFAIL: The command cannot be executed when global transaction is in the PREPARED state
|
12.1.0-opt>INSERT INTO t SELECT * FROM t;
|
ERROR 1399 (XAE07): XAER_RMFAIL: The command cannot be executed when global transaction is in the PREPARED state
|
Another example is when we remove only AUTO_INCREMENT from the testcase, but leave the enabling of pseudo_slave_mode in place. In this case, all versions go into the 'Sending data' query state for the INSERT till a lock wait timeout is reached.
Attachments
Issue Links
- relates to
-
-
- Open
-