I am investigating a problem encountered by an application using BDB JE (HA) 5.0.104 deployed across a network whose nodes have 'internal' IP addresses that not do route from the other nodes in the network. All the configuration being passed to JE (helper, nodeHostPort) is in the form of DNS resolvable addresses that are resolvable at every node of the network.
When a group is formed, the first node is created successfully. The problem occurs when the second node is created. Its creation hangs within the ReplicatedEnvironment constructor.
Turning up the logging, I see that it has sent a Master Query request but the Master Query response that comes back contains a resolved IP address, rather than node host/port known to the first node.
Looking at the code I see that MasterSuggestionGenerator changed between 6.4.9 and 6.4.25 from generating a MasterValue built from the socket's resolved IP to one using the host/port known to RepNode. Repeating the same test with 7.4.5 showed that the problem disappears.
My question: I don't see this defect referenced in the change log. Would you be able to summarise the defect and its scope? The application will be upgrading to 7.4.5 soon, but I am interested to know if there was a known workaround for 5.0.104.
Thanks in advance
2017-11-06 18:56:56.377 UTC FINE [NODE_1] Learner started
2017-11-06 18:56:56.386 UTC FINE [NODE_1] Sent master request MQ 2.0|testgroup|-2147483648 2.0|testgroup|-2147483648|MQ to [xxxxxxx.xxxx.xxxxxx.xxxxx.xxxx.xxxx.xxx.xxx/10.104.51.49:5700]
2017-11-06 18:56:56.404 UTC FINE [NODE_1] Learner notified. Proposal:Proposal(0000015f92b05321:0000000000000000000000000a13960b:00000001) Value: Value:10.19.150.11$$$5700$$$NODE_0$$$1
2017-11-06 18:56:56.405 UTC FINE [NODE_1] Master change listener notified. Proposal:Proposal(0000015f92b05321:0000000000000000000000000a13960b:00000001) Value: Value:10.19.150.11$$$5700$$$NODE_0$$$1
2017-11-06 18:56:56.405 UTC INFO [NODE_1] Master changed to NODE_0