[MDEV-8959] change master to master_use_gtid= purges relay log and fetchs binlog from master Created: 2015-10-17  Updated: 2022-11-10  Resolved: 2022-11-10

Status: Closed
Project: MariaDB Server
Component/s: Replication
Affects Version/s: 10.0.21
Fix Version/s: N/A

Type: Bug Priority: Major
Reporter: Daniel Black Assignee: Kristian Nielsen
Resolution: Duplicate Votes: 0
Labels: None

Issue Links:
Relates
relates to MDEV-4698 With GTID replication, relay logs can... Open
relates to MDEV-8945 Avoid overloading the master NIC on r... Open

 Description   

when STOP SLAVE; CHANGE MASTER TO MASTER_USE_GTID=

{slave_pos|current_pos}

; START SLAVE is used, the relay logs are purged and refetched from the master. In this case, or when the gtid_slave_pos is advanced, the relay logs will contain all the transactions required to continue.

As I discovered when doing when a replication delay exceeded expire_log_days of the master was that I was left in a state with no relay logs and a master file:position that wasn't able to be fetched.

An ideal implementation would examine to see at what point the existing relay logs can be used to continue replication.



 Comments   
Comment by Daniel Black [ 2015-10-17 ]

possibly a specific case of MDEV-8945. I'm unsure if just doing a stop slave SQL_THREAD would of helped in this scenario.

Comment by Elena Stepanova [ 2015-10-17 ]

See also MDEV-4698.

Comment by Elena Stepanova [ 2022-11-10 ]

Closing as a duplicate of MDEV-4698

Generated at Thu Feb 08 07:31:04 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.