I have two MySQL Clusters which normally operate with Master - Master Replication between two of their nodes.
Site 1 gets the majority of the write traffic, with Site 2 writing small amounts of isolated tables. The Auto Inc's are configured differently, so they can not clash.
Site 1 App 1 -> Site 2 App 1 replication is working fine, and was started using the recommended method on MySQL's website.
I had Site 2 App 1 replicating to Site 1 App 1 using the SHOW MASTER STATUS
method, and starting the replication from that position.
Last week, Site 2 App 1 was restarted, and Site 1 -> Site 2 replication restarted fine. Site 2 App 1 -> Site 1 App 1 was attempted to restart using SKIP ERRORS
to ignore the cluster log file break error (I can't remember the name).
This replication keeps breaking with:
Error 'Can't write; duplicate key in table ''' on query. Default database: ''. Query: 'COMMIT'
I can't do the normal method (getting the epoch from ndb_binlog_index
), because it doesn't return any files. Any time I try the SHOW MASTER STATUS
method, it brings up the error above. I've even tried doing it via Site 2 App 2 -> Site 1 App 2, and it still fails.
Any ideas?