[MDEV-28212] Query Cache on Primary-Secondary-Secondary not invalidated on updates Created: 2022-04-01  Updated: 2022-05-03  Resolved: 2022-04-28

Status: Closed
Project: MariaDB Server
Component/s: Query Cache, Replication
Affects Version/s: 10.5.13
Fix Version/s: 10.5.14

Type: Bug Priority: Major
Reporter: Tim Bienias Assignee: Angelique Sklavounos (Inactive)
Resolution: Duplicate Votes: 0
Labels: None
Environment:

bitnami/mariadb Helm Chart on Kubernetes running on Linux

Issue Links:
Duplicate
duplicates MDEV-27047 replication does not remove affected ... Closed
Relates

 Description   

When running Primary-Secondary-Secondary setup DML operations on Primary do not trigger query caches on Secondaries to be invalidated when binary log is in ROW format.

Reproduce as follows (given a table ):

On Primary:

insert into test values (123);
 
select * from test; #=> gives 123

On Secondaries:

select * from test; => gives nothing
 
select * from test where 1=1; # or select SQL_NO_CACHE * from test; => gives 123

Excerpt from my.cnf for Primary and Secondaries:

binlog_format=row
 
query_cache_type=1
 
query_cache_limit=4M
 
query_cache_size=256M



 Comments   
Comment by Angelique Sklavounos (Inactive) [ 2022-04-12 ]

Have been able to reproduce on 10.5.13 with primary --> replica. Still need to check other versions and write up test case...

Comment by Angelique Sklavounos (Inactive) [ 2022-04-21 ]

I tested this with 10.5.15 and could not see the behavior with a primary server and two replicas (and could see it with 10.5.13). I later stumbled upon MDEV-27047, which I believe this bug duplicates. I made the assumption that the replication geometry was simple, with the primary replicating to two replicas who are unaware of each other, but I could test again if the geometry is something a bit more complicated, say one replica is in between the primary and second replica (however, I still think the issue would still be the same as MDEV-27047).

Comment by Angelique Sklavounos (Inactive) [ 2022-04-21 ]

tbienias could I please verify the exact setup of the primary and two replicas to test this and see if it is a duplicate bug or not? Thank you.

Comment by Marko Mäkelä [ 2022-04-28 ]

Could this be a duplicate of MDEV-27047, which affected 10.5.13 and 10.6.5 and was fixed in 10.5.14 and 10.6.6?

Comment by Angelique Sklavounos (Inactive) [ 2022-04-28 ]

Could this be a duplicate of MDEV-27047, which affected 10.5.13 and 10.6.5 and was fixed in 10.5.14 and 10.6.6?

Yes, I wrote this above.

Comment by Angelique Sklavounos (Inactive) [ 2022-04-28 ]

tbienias please try upgrading to 10.5.14. I will resolve this bug as a Duplicate for now.

Generated at Thu Feb 08 09:58:57 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.