With replication setup as A => B => C, and a MyISAM table with only an id field as a primary key, executing these statements on A
of course results in a duplicate key error on the second statement, but not until after keys 6 and 5 are already written. The two statements are then sent to Slave B, with the same result; I.E., the tables on A and B now each have the same 6 rows.
However, Slave B does not pass the second statement on to Slave C at all (never hits the Slave B binlog). So Slave C only has 4 rows. Additionally, I see no errors or messages in the Slave B log to indicate that the statement resulted in duplicate key errors, nor that the statement was not replicated.
sql_mode is blank on all three servers, and slave_skip_errors is OFF. Also, binlog_format is MIXED.
Is this expected/intended behavior, and is it documented somewhere that I could take a look?