Hi, I am experimenting how network configuration impacts BDR replication, ran into something that I can’t explain, and wonder if someone can shed light. Here it goes:
With a four node BDR group configured and running (all using default port 5432), I purposely blocked port 5432 on one of the node in the group, and was expecting to see changes on other nodes stop being replicated to this node, but that’s not what happened.
Shell commands show that the port was indeed blocked (In the following example session, the port 5432 is blocked on 10.3.122.31, but open on 10.3.122.21):
% nc -v --send-only 10.3.122.21 5432 </dev/null
Ncat: Version 6.40 ( http://nmap.org/ncat )
Ncat: Connected to 10.3.122.21:5432.
Ncat: 0 bytes sent, 0 bytes received in 0.00 seconds.
% nc -v --send-only 10.3.122.31 5432 </dev/null
Ncat: Version 6.40 ( http://nmap.org/ncat )
Ncat: Connection timed out.
% psql -h 10.3.122.21 mydb
psql (9.4.10)
Type "help" for help.
mydb=#
% psql -h 10.3.122.31 mydb
psql: could not connect to server: Connection timed out
Is the server running on host "10.3.122.31" and accepting
TCP/IP connections on port 5432?
At this state, I tried insertion and update on node 10.3.122.21, and all of which were replicated to node 10.3.122.31. However, attempt to create a new table on node 10.3.122.21 was stuck (as expected) until the port 5432 on 10.3.122.31 opened again.
So my question is, is there another port other than port 5432 that BDR uses for replication? If not, how could changes be replicated to 10.3.122.31 when its port 5432 was blocked?
Thanks,