Bug #108278 MGR server hang when MGR is slave and relay_log_recovery=1 and group_replicati
Submitted: 25 Aug 2022 2:21 Modified: 26 Aug 2022 7:00
Reporter: miu 123 Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: Group Replication Severity:S2 (Serious)
Version:8.0.29 OS:Any
Assigned to: MySQL Verification Team CPU Architecture:Any
Tags: MGR hang

[25 Aug 2022 2:21] miu 123
Description:
The server cannot be started after it crashes. Hang stops during binlog recovery

How to repeat:
The server cannot be started in the following cases
(1)MGR,single primary,8.0.29
(2)MGR as slave,Configure master-slave replication using binlog position,for example:CHANGE MASTER TO MASTER_HOST='172.16.106.20',MASTER_USER='repl',MASTER_PASSWORD='repl',MASTER_PORT=3306,MASTER_LOG_FILE='mysql-bin.000001',MASTER_LOG_POS=154 FOR CHANNEL 'test';
(3)relay_log_recovery=1 and group_replication_start_on_boot=on

kill -9 this server,The server could not be started successfully. Hang stopped during binlog recovery. When relay_log_recovery = 0 or group_replication_start_on_boot = off, it can be started successfully
[25 Aug 2022 2:31] miu 123
The log finally stays in binlog recovery, and there is no output

Attachment: error.log (application/octet-stream, text), 50.38 KiB.

[25 Aug 2022 10:33] MySQL Verification Team
Hi,

Did this crash happen after upgrade to 8.0.29 or ?
What was changed before this crash?

Can you share your config, how many nodes you have in the group replication? 
What is the status of each node?

Have you contacted MySQL support team as this looks more like a support issue than a bug

thanks
[26 Aug 2022 6:15] miu 123
Hello,Thank you for your attention.

server crashed after executing a DDL statement. like 'alter table table_A add column_name varchar(500);'
After the database is started, any operation to query this table_A will cause a crash,I think the innodb file may be corrupted.

Before 8.0.29, I never encountered innodb corruption.
As far as I know, it may be caused by the new feature of 8.0.29,I think this is another bug.

On this issue,group replication have three nodes,the crashed node is primary,the status of the other two nodes is online.
After testing, this configuration is OK at 8.0.27,I think this problem is caused by the version upgrade, so I report the bug.
[26 Aug 2022 6:18] miu 123
my.cnf

Attachment: my.cnf (application/octet-stream, text), 4.35 KiB.

[26 Aug 2022 6:59] MySQL Verification Team
Duplicate of Bug#34233264
[26 Aug 2022 7:00] MySQL Verification Team
Hi,

This is a bug in 8.0.29 fixed in 8.0.30, please upgrade as soon as possible.

thanks for using MySQL and sorry for the inconvenience