Bug #110094 When slave replays the STOP EVENT, the stop slave will cause a dead cycle

• Submitted: 16 Feb 2023 7:34
• Modified: 22 Feb 2023 13:55
• Reporter: GONGTUI FU
• Status: Can't repeat
• Category: MySQL Server: Replication
• Severity: S2 (Serious)
• Version: 8.0.25
• OS: Any
• Assigned to: MySQL Verification Team
• CPU Architecture: Any

[16 Feb 2023 7:34] GONGTUI FU

Description:
When slave gets a GTID EVENT from master,then stop the server.The STOP EVENT will be recorded in the relaylog file.
If we change the server-id and restart the server,executing START SLAVE SQL_THREAD and STOP SLAVE SQL_THREAD will casuse a endless loop.

How to repeat:
slave server uses DEBUG version.
1. slave server executes: SET GLOBAL DEBUG='d,stop_io_after_reading_gtid_log_event';(only receive GTID EVENT)
2. maser server executes any dml statement.
3. stop slave server. This will generate a STOP EVENT after GTID EVENT in relaylog file.
4. stop master server. It prevents the slave server from automatically connecting to the master server after starting.
5. modify server-id that is in my.cnf, for execution of STOP EVENT.
6. start slave server and executing START SLAVE SQL_THREAD.
7. executing STOP SLAVE SQL_THREAD.

A dead cycle occurs!

[21 Feb 2023 16:57] MySQL Verification Team

Hi,

Did you try this on later 8.0.31? I'm having issue reproducing this on 8.0.31.

Thanks

[22 Feb 2023 8:36] GONGTUI FU

I didn't.

[22 Feb 2023 13:55] MySQL Verification Team

Hi,

Having issue reproducing this with 8.0.32, please provide reproducible test case that works with 8.0.32