I cant see how this caan happen, your sure your backup is actually a backup and you arent seeing the message re distributed, ive seen cases where this happens with 2 live servers and the message ends up back where it started.
what TX Mode are you using?
once the duplicate is detected the whole transaction will be ignored.
Replying to both of you --
Thanks, Andy. I did check and confirm that my backup is a backup. I also looked to see if any failover happened and none has.
Clebert, I created my session with "false" for the transacted boolean and AUTO_ACKNOWLEDGE for the acknowledgement mode. I've also set the session to use a custom send acknowledgement handler. The sendAcknowledged method is empty on the handler.
Thanks for helping
Sorry, more info that I realize I wasn't clear on.
I'm running on 2 servers. Each server has
- Primary HornetQ
- Backup HornetQ
- Producer (Java service of my own)
- Consumer (Java service of my own)
The HornetQ primary and backup use a broadcast group and a discovery group. The group-address and group-port is the same for broadcast and discovery and the same in all 4 configurations (both primary and both backup). Some other configuration notes
You only have a primary and backup server? or you have other nodes?
if you have more than one node, there's an issue with redistribution when using duplicate detection which I'm by coincidence (not related to this issue) working on at the moment:
(post edited by Clebert.. had a wrong link here before)