[
https://issues.apache.org/jira/browse/ARTEMIS-4058?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17621334#comment-17621334
]
Justin Bertram commented on ARTEMIS-4058:
-----------------------------------------
bq. ...here is the correct architecture diagram. Microservice(kube) --> LB --->
QPID ---> Brokers(4-masters/4-slaves)
Thanks for clarifying. Is there a specific reason you're using 4 pairs?
bq. The issue we are seeing now is multiple sessions are getting created from
routers during the restart of qpid. Not for all queues but 40% of the queues.
where 1 qpid-router is having 3 sessions while other qpid is having another 3.
total 6 consumers for one queue.
If Qpid Dispatch Router is creating "multiple sessions" I'm not sure there's
anything the broker can do about that. Isn't Qpid Dispatch Router responsible
for the sessions that is creates? Can you inspect any metrics on Qpid Dispatch
Router to see how many sessions it thinks it is creating and then compare that
with the value reported by the broker? I would expect those values to be the
same.
bq. I seriously don't understand how this queue is created since all of the
brokers (master & slaves work in a pair to support HA. How this broker got out
of HA is a mystery.
Now that you've clarified that you actually have a cluster of 4 HA pairs (i.e.
8 brokers) the existence of this queue makes more sense. In a cluster of
multiple active brokers every broker will have queues like this. They are
created automatically by the broker itself for moving messages between the
cluster nodes. This particular queue is meant for sending messages to the
broker whose node ID is {{337d5748-6c39-11eb-9983-0ae398a5a5b7}}.
At this point I don't see any data indicating a problem with the ActiveMQ
Artemis.
> Seeing higher consumer count in AMQ CONSOLE than configured
> -----------------------------------------------------------
>
> Key: ARTEMIS-4058
> URL: https://issues.apache.org/jira/browse/ARTEMIS-4058
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: AMQP, Broker, JMX, Web Console
> Affects Versions: 2.18.0
> Environment: PROD
> Reporter: Ekta
> Priority: Major
>
> Hello,
> Recently we have been facing an issue with amq brokers where the consumer
> count that shows up in AMQ console for each queue is much higher than what is
> configured. It is coming as a different number for each queue. We believe
> that consumer count should always remain 2 across all of the queues since we
> have 2 qpid dispatch routers acting as consumers. Lately we have observed
> the consumer count is not consistent across all brokers. However there are no
> recent changes made to the application. It is all of the sudden acting up.
> Below is our architecture.
> Microservice(kube) --> LB ---> QPID ---> Brokers(master/slave)
> The queues that are having more than 2 consumers looks like they are holding
> up multiple sessions from each qpid dispatch router for an example: for one
> queue having 6 consumers there are 6 session ID's. all those 6 id's belong to
> those qpid's. 1-qpid has two consumers sessions while 2-qpid is having 4
> session. Just like that we are seeing numbers like 6,5,3 for other queues.
>
> As part of troubleshooting this issue I found a queue under addresses called
> $.artemis.internal.sf.artemis-cluster-prod.337d5748-6c39-11eb-9983-0ae398a5a5b7
> only on one of the broker. I don't recognize this queue nor we created it.
> Looks the system somehow created but not sure why. All of our queues are
> created on qpid, we are not creating any queues on brokers. Routing-type on
> that queue is coming as MULTICAST
> Would like to know if anyone knows how/where to fix these issue.
> Versions:
> activemq: 2.18
> qpid-dispatch router: 1.18
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
