Description:
Seen on a server recently:
| 792552 | system user | | NULL | Query | 71991 | Waiting for dependent transaction to commit | NULL |
| 792553 | system user | | emk | Query | 71992 | Waiting for backup lock | CREATE TABLE IF NOT EXISTS SomeTable (
pk_id INT UNSIGNED PRIMARY KEY NOT NULL,
|
This server was running 8.0.28
It was also an intermediate master, eg. master -> intermediate master -> slave
The SQL thread was blocked.
https://dev.mysql.com/doc/refman/8.0/en/clone-plugin-options-variables.html#sysvar_clone_b... implies this should no longer happen since 8.0.27 if clone_block_ddl = 0 which is the default and was the setting on this host.
Impact: several replicas under this intermediate master were not getting updated as expected.
How to repeat:
In theory see above.
1. Setup master --> intermediate --> master with 8.0.28
2. have the master push out CREATE TABLE IF NOT EXISTS <some_name>; statements
3. Do a native clone from the intermediate server
4. check to see if replication is blocked
I'll upload privately the configuration of the master and intermediate master.
Suggested fix:
The documentation says this should be resolved now. It appears this is not the case under some circumstances. Is this type of statement not considered a DDL statement? If so it should be.
Please ensure that as described by documentation the DDL statement can pass through and replication can continue in such conditions.
