I set up a primary-replica replication on AWS-managed MariaDB instances, I let the databases run uninterrupted for 24hours while I monitored and ran consistency tests on the replica database to confirm replication was active.
After activating and allowing frappe’s read-write feature to run for a few hours on a production site, I observed that a duplicate error 1062 is raised on the replica database, this stops the replication, and users on the production site are not able to fetch the most recent data.
I ran the sequence stated above twice and experienced the same error each time I activate read-write on Bench.
Could the read-write feature be responsible for writing directly to the replica database? because the error is not generated when the read-write feature is deactivated.
Could it be that the function is also switched and used for write operations?
In the interim, I have set the parameter group option for read_only to True if the database is a replica to restrict the database to only read operations.
I am aware that I can skip the error and move on but skipping errors could lead to database divergence.
Any information regarding this will be appreciated.