0

I recently had to rollback from WF13 to WF11 due to a regression in one of the dependencies.

Now I am trying to get the AMQP protocol to work on WildFly 11's messaging system. I am running a high availability setup with two nodes. Each of the node has a message consumer locally. This message consumer connects through AMQP1. I've added io.netty as a dependency to the org/apache/activemq/artemis/protocol/amqp module and updated org/apache/qpid to get the AMQP protocol to work (see also WFLY-7823). Now my AMQP message consumer works fine, but it seems to receive messages always exactly twice, and it appears to be even in the same frame. This happens on the same node (the other node receives messages through the bridge if the message isn't handled locally in the first place). So on one node and one queue consumer, I receive every message exactly twice, at the very same instant, before I even got to send an ACK/NACK for the first message I received.

I don't remember seeing this issue on WildFly 13.

Are there any known regressions regarding how messages are sent through the remote connectors? Perhaps an issue in the AMQP protocol? Or could it be a compatibility issue with the updated version of qtip?

Jan-Willem Gmelig Meyling
  • 3,685
  • 2
  • 21
  • 43

1 Answers1

0

For what it's worth Wildfly only uses ActiveMQ Artemis to fulfill their need for a JMS implementation (i.e. a traditional part of Java EE). I don't believe Wildfly has any real interest in other protocols or APIs aside from JMS.

My understanding here is that if you need to support multi-protocol message use-cases you should likely be using a standalone broker. I believe this is why none of the other protocols which Artemis supports (e.g. AMQP, STOMP, MQTT, OpenWire) are enabled in Wildfly or have documentation on how to enable them (aside from the occasional forum post).

It's also worth noting that Wildfly 11 has Artemis 1.5.5 in it which is 10 releases or so behind the latest version (i.e. 2.6.2). Much work has been done on the AMQP implementation in those releases so I think you're best served by using a standalone version of Artemis rather than Artemis embedded in Wildfly.

Justin Bertram
  • 29,372
  • 4
  • 21
  • 43
  • Ok, I was hoping to find a concrete issue ID and perhaps commit that I could patch on top of my WildFly 11 as a temporary solution. Based on your answer it seems logical to wait for WF14 that ships with ActiveMQ 2.x AFAIK. Hopefully that also comes with fixes for the other issues I had. Else I will probably end up setting up a separate ActiveMQ cluster. – Jan-Willem Gmelig Meyling Aug 13 '18 at 11:02