HA_ERR_KEY_NOT_FOUND on both master at the same time (no replies)

Hi everyone!

It's my first post on this forum so I'll try to give all the details you need to help me.

I have a Master/Master ROW-based replication working with both servers running 5.1.47. There is also a slave connected to db1 just for the daily backups.

Here are the relevant parts from my.cnf on the masters:

transaction_isolation = REPEATABLE-READ
binlog_format=row
sync_binlog=1
log_slave_updates
replicate-same-server-id=0
server-id = 2
auto_increment_increment=10
auto_increment_offset=2
replicate-ignore-db=phpmyadmin
replicate-ignore-db=mysql
slave-net-timeout = 5
master-connect-retry = 5

The config on both servers are the same except for the server id and the auto_increment_offset.

It's working nice and fast. The problem is that I seldom gets HA_ERR_KEY_NOT_FOUND on both servers at the same time. I understand this is kind of the classic replication error, but what I don't understand is how I can get it on both servers at the same time when there's no write on db2. At the moment, db2 is only a replicated MySQL and nothing writes to it (but it's not read-only).

I would totally understand to get this error on db2 if there has been an issue with replication. What's puzzling me is how can I get this error "loops back" to db1. Everytime I get a replication error on this setup, both SQL slave thread stops with the same error (the log positions are different of course). How is this possible?

Thanks in advance for your insights,

Martin