We ran into an interesting problem the other day, when our production database stopped inserting row into a given table even though from the clients perspective they were inserted. What makes this even stranger is that the binlog recorded these “phantom” rows which then ended up on the slave. So the Slave actually have the valid data which was missing from the Master.
After playing for a bit, I notice that the given table was not locked, i.e. I could insert row into it. However, these rows were inserted into the next available row in the Master database, i.e. it did not skip over these phantom row. So this leads me to believe that it is some problem with InnoDB but I cannot figure out how this could have happened. Below is the table status for the table that locked up. Thanks in advance for any help.
Create_time: 2009-09-04 19:27:39
Comment: InnoDB free: 4096 kB; (