[MDEV-20976] Undo log purge thread stops in replication slave Created: 2019-11-05  Updated: 2023-04-25  Resolved: 2023-04-25

Status: Closed
Project: MariaDB Server
Component/s: Replication
Affects Version/s: 10.3.10
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: naka sig Assignee: Andrei Elkin
Resolution: Won't Fix Votes: 0
Labels: replication
Environment:

OS:RHEL7.5

Attachments: File (master)my.cnf     File (slave)my.cnf     Text File SHOW_ENGINE_INNODB_STATUS.txt    

 Description   

Currently, it operates with the configuration of mariaDB (master / slave) + maxscale.

About once a week, the undo log purge thread stops on the slave side,
There is a phenomenon that ibdata1 is enlarged.
(the master side is normal)

Contents of SHOW ENGINE INNODB STATUS when the phenomenon occurs:
[Attached file]

History list length becomes large, and "1 read views open inside InnoDB" occurs.
It seems that the purge thread has stopped.

This phenomenon can be solved by stopping replication with "stop slave".

Where is the cause?
There may be a problem in the operation between master and slave.

 Comments   
Comment by naka sig [ 2019-11-22 ]

Added my.cnf file.
(The port number is masked)

Comment by Andrei Elkin [ 2023-04-17 ]

mis-naka, thank you for the report! While 10.3 is reaching EOL maybe you can tell us about 10.4 and up behaviour in this regard?

Comment by Andrei Elkin [ 2023-04-25 ]

Given that 10.3 is leaving the service and there's been no feedback, I am closing the ticket.

Generated at Thu Feb 08 09:03:39 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.