Re: ERROR 1033 (HY000): Incorrect information in file

[Stu Derby]

As best I can tell, this error message is produced whenever the DB is unable to perform INNODB recovery upon restart. The DB needs to assume on restart that it might have crashed, maybe power failure, and so there may be partially stored transactions in the log file that either need to be completed or unwound. If it can't do that, recovery fails and the tables are therefore inaccessible, with this rather cryptic message.

I ran into this problem when restoring a slave DB, using the backup of the master. Someone had cleverly (not) increased the innodb_log_file_size on the slave but not the master and I was using the master's data and log files with the slave's my.cnf The only clues I got were the 1033 error message and the startup error message (in /var/log/mysqld.log on my system):
InnoDB: Error: log file /var/lib/mysql/ib_logfile0 is of different size 0 268435456 bytes
InnoDB: than specified in the .cnf file 0 536870912 bytes!

Making the slave's log_file_size match the log file copied from the master made everything work (some 12 hours after I started on the problem).

I suspect there are several other ways to get this situation, probably corrupted or missing log files, path problems, etc. It would be nice if the DB would put out an error saying something like "recovery has failed, InnoDB tables not accessible" to help people focus on the earlier error (in my case the log file size issue).