[
https://issues.apache.org/jira/browse/QPIDJMS-230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15734923#comment-15734923
]
Robbie Gemmell commented on QPIDJMS-230:
----------------------------------------
Whats happening is that your usage is creating an 'anonymous producer' (no
destination specified at creation), and the server isn't advertising the
capability the client checks before using an 'anonymous sending link' for this
producer.
In AMQP 1.0, producer/consumer 'links' carry messages, and these are
established to a source/target with a specific address. That unfortunately
doesn't play very nice with JMS 'anonymous producers' which dont have a
specific address and each send could go to a different destiantion. An
extension is used by many AMQP 1.0 servers, of treating a target with null
address as an anonymous sending link, then inspecting each message for its 'to'
address and using this as the message destiantion rather than the link address,
with support advertised for this using a connection-capability of
ANONYMOUS-RELAY.
As the server is not offering this capability the client doesn't try using it,
and instead uses a fallback of opening (and by default, closing) links to the
destinations as required, which is what you are seeing. You can see this
mentioned at the start of your logs: "2016-12-09 07:16:34,937 | DEBUG |
Creating an AmqpAnonymousFallbackProducer ".
There was some limited support for caching these links, but I dont think it is
tested and I don't think anyones really used it before, we've considered
removing it previously. I'll need to take a look to see if its an option for
you.
> AmqpFixedProducer created for every message sent
> ------------------------------------------------
>
> Key: QPIDJMS-230
> URL: https://issues.apache.org/jira/browse/QPIDJMS-230
> Project: Qpid JMS
> Issue Type: Bug
> Components: qpid-jms-client
> Affects Versions: 0.11.1
> Reporter: Milan Nikl
> Priority: Minor
> Labels: performance
> Attachments: amqps_frames.log, apache.qpid.log
>
>
> *Configuration:* Hi, I'm using ActiveMQ 5.14.1 to connect from a device
> (running linux derivate OS) to Azure IoT Hub. With Qpid JMS client 0.11.1 and
> protonj 0.15.0 providing AMQPS based JmsConnection implementation. I'm using
> Destination bridges to connect my local queues to remote queues.
> *Problem description:* For every message I send to the server there is new
> connection attempt made. I can see those reconnections in Azure IoT Hub
> monitoring, in AMQPS frames logs, qpid log etc. And while I have the same
> JmsConnection instance active for the whole time, in reality the device keeps
> connecting and disconnecting for each message, which presents additional load
> for both IoT Hub and device. Message throughput is really affected by this
> behaviour.
> When I try running similar code from my desktop using Qpid JMS client 0.11.1
> to send messages, it creates single AmqpFixedProducer at start and
> disconnects when all messages are sent. I'm not aware of any special
> settings, both application use JmsConnectionFactory.setForceSyncSend(true).
> So ActiveMQ or some of its components could be blamed for this. But I would
> like to know if someone has any experience with similar problem and maybe
> some idea how to solve it.
> Thanks!
> Attachments: In logs you can see the device connecting, then some already
> enqueued messages are sent. Once the initial load is processed, one message
> per minute should be sent.
> Originaly filed in
> http://qpid.2158936.n2.nabble.com/AmqpFixedProducer-created-for-every-message-sent-td7655816.html
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
