When a replica is started it attempts to connect to other RS instances in the topology. This may lead to simultaneous cross connection attempts ultimately leading to one of the connections being terminated prematurely, since only one connection is needed.
- RS X initiates connection to RS Y
- RS Y initiates connection to RS X
- RS Y responds to RS X: connection initiation attempt from (1) successful
- RS X sees that there is already a connection from X to Y, so it aborts the connection initiation attempt from (2)
- RS Y is notified that its connection attempt has been aborted mid-negotiation and logs a severe error, even though nothing is fundamentally wrong
Here's the message:
[04/Jan/2012:15:29:32 +0100] category=SYNC severity=SEVERE_ERROR msgID=14942387 msg=Replication server XXXX was attempting to connect to replication server YYYY but has disconnected in handshake phase
Although there is no loss of functionality as a result of this issue, the error message is very misleading and users will think that their topology has broken.