We have a 3 node cluster that runs mostly InnoDB tables. We have MyISAM replication enabled in my.cnf. One database has all InnoDB tables aside from a brand new table that was set to MyISAM accidentally. After this table got an insert for the first time, it crashed the database.
OK, so PXC node 1 went down, PXC node 2 picked up. The developer saw a “mysql connection disappeared” or something similar after his attempt to do the INSERT. So he then went to PXC node 2 and did the same insert. This crashed node 2. This then left node 3 in “initialized” state and data could not be read or written to.
Questions:
- Node 3 is not going to do anything until node 1 and/or 2 are fixed, correct? Should node 3 still operate?
- Does all of this sounds like it is the proper outcome to accidentally creating a MyISAM table in a primarily InnoDB setup?
- Can I avoid these issues? We need to convert a another database to InnoDB from MyISAM and we know it will crash everything until we do a restart.