I’m not used to support duties, so every time I get a call like that it’s a bit frightening: replication doesn’t work. A connect to master gets established OK, the server receives updates, but the replica… uhm… sits on LSN #1.

Hm.. what could be causing this?-)

Tarantool has a master port, which speaks binary request/response protocol, and a separate replication port, which simply expects a binlog position(64-bit integer) on connect and then keeps feeding the replica WAL updates. Now, case in question, the replica would connect to the master port and begin the infinite wait…

The problem was easy to detect since the master didn’t even listen on its replication_port.

Now, I’m thinking how to fix this. One solution is to simply merge two ports and agree on protocol type during handshake.

Another is to leave things as is but add handshake to replication.

Once (if) we get authentication support, we’ll need to merge all ports into one anyway.