MySQL Bugs: #98288: xa commit crash lead mysql replication error

Bug #98288	xa commit crash lead mysql replication error
Submitted:	20 Jan 2020 0:57	Modified:	13 Feb 2020 8:23
Reporter:	phoenix Zhang (OCA)	Email Updates:
Status:	Duplicate	Impact on me:	None
Category:	MySQL Server: XA transactions	Severity:	S3 (Non-critical)
Version:	5.7.26	OS:	Any
Assigned to:		CPU Architecture:	Any
Tags:	replication, xa

Description:
Consider an mysql replication, master-A and slave-B.

In master-A, execute below:
XA START '1';
INSERT INTO t1 VALUES (1);
XA END '1';
XA PREPARE '1';
XA COMMIT '1';

Now, mysqld server crash when do XA COMMIT, one case is, the binlog already write and send to slave-B, while it does not commit in engine.

Cause slave-B receive the binlog event, the trx finally commit in slave-B. However, when master-A restart, it will not auto commit this trx, it still in prepared state. Then, master and slave share data inconsistent. If master-A do XA COMMIT '1' again, then the slave thread will broken in slave-B.

How to repeat:
according the description. kill -9 master-A mysqld server process when do xa commit, it will repeat. 

Suggested fix:
When master-A restart, if external xa transaction being prepared state, and it can find that xid in binlog event, do auto commit when recover.

Here is my patch to do auto recover when mysqld restart. Do commit when engine prepared, and write XA COMMIT event in binlog

Attachment: xa_commit_recover.diff (text/x-patch), 9.55 KiB.

Hi,

Thanks for the report. This one was easy to reproduce, I appreciate the test case.

If you did not already signed OCA:

In order to submit contributions you must first sign the Oracle Contribution Agreement (OCA).
For additional information please check http://www.oracle.com/technetwork/community/oca-486395.html.
If you have any questions, please contact the MySQL community team. 

much appreciated
Bogdan

Posted by developer:
 
This is a duplicate to the effect that "XA statements are not atomic in the server (engine + binlog)".

Duplicate of

Bug#76233 XA prepare is logged ahead of engine prepare