MySQL Bugs: #111662: continuous replication delays occur.

Bug #111662	continuous replication delays occur.
Submitted:	5 Jul 2023 6:42	Modified:	18 Jul 2023 11:04
Reporter:	Hoseung LEE	Email Updates:
Status:	Not a Bug	Impact on me:	None
Category:	MySQL Server: Options	Severity:	S3 (Non-critical)
Version:	8.0.33	OS:	CentOS
Assigned to:	MySQL Verification Team	CPU Architecture:	x86

Description:
When MySQL 8.0.33 slave_parallel_workers is set to 1, replication delays occur continuously for 1 second.
If set to 0, replication delay is disappeared.

I have two questions.
In the reference manual, beginning with MySQL 8.0.30, setting variable to 0 is deprecated and 1 has the same effect.
However, it seems to behave differently when set to 0 and when set to 1.
And, when setting replica_parallel_workers to 1, there is coordinator thread.

How can I solve the replication delay when setting 1?

How to repeat:
when set the slave_parallel_workers 1 or high in MySQL 8.0.33

Hi,

I need you to clarify what you mean by:

> replication delays occur continuously for 1 second

Do you see constantly that slave is 1 second behind master?
Do you see constantly that slave behind master value is increasing every 1 second (so after 5 days of running you have many hours delay)?

If you run your system for 7 days, what is your slave behind master value on your slave(s)?

Thanks

I think it's the same bug
https://bugs.mysql.com/109133

@Max, thanks but there we have some "serious" delays, here reporter just confirmed he has a delay of 1 second so not the same

Hi Hoseung LEE,

There always need to exist some delay between master and slave. 1second delay is not a bug. It is just different measuring and rounding with and without parallel replication.

Thank you for using MySQL

Hi,

What I'm curious about is that the manual says that setting from MySQL 8.0.30, 1 is the same as setting to 0 and that there is no coordinator thread.
But there seems to be a coordinator thread.

When set to 1, it works as parallel replication, and the method of measuring replication delay is different, but is it the same as when it is 0?

Thanks