Have you heard this:
master to master replication replicated its own queries (replicated to another server and
replicated back)? Although both two servers has own server id. For example, master A
reads binlog on master B, and executed, got duplicate key error. master B reads binlog
from master A, and got duplicate key error. (these two queries are different). The
condition is, On master B, there are no application setup to access database, it only
replicates from master A. No updates on server B.
MySQL used v5.037.