my problem is the following. We are running a MySql server on AWS RDS and are about to upgrade from MySql 5.7 to 8.0.34. When testing this on a restored snapshot I discovered that there is a "dummy"-schema, without any tables in it, not used (and never have been used) that is saved on disc with upper case letter in the name. This is causing the error MY-013520 when trying to upgrade to 8.0.34 since we have lower_case_table_names set to 1.
I know that changing the setting to 0 allows me to drop the schema, but this would involve downtime and if possible I want to avoid this.
I tried setting up a read-replica, changing to lower_case_table_names = 0 on the replica, dropping the schema on the replica and then changing back to lower_case_table_names = 1. This works almost. The replication stops at first but then resumes without error after the second reboot. For some reason however, the replication is not keeping up with the master after the replication start up again. I.e. before dropping the schema the replication lag decreased with time but after the schema is removed the replication lag increases slowly with time. I.e. ~ 1000 seconds per hour. No errors are written to the error log, it just seems there is some overhead after the schema is removed.
So basically my question is two folded
- Is there a way to force a DROP DATABASE to be case-sensitive although having lower_case_table_names set to 1?
- Anyone knows why the replication is slow after removing the unused schema on the slave?
Thanks in advance!