[jira] [Commented] (ARTEMIS-3517) Memory leak when using AMQP

["Robbie Gemmell (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARTEMIS-3517?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17425095#comment-17425095 ] 

Robbie Gemmell commented on ARTEMIS-3517:
-----------------------------------------

Perhaps give the main branch a test out as comparison, since 2.19.0 is due soon (was expected to already be done by now) and is where any change will be applied. https://github.com/apache/activemq-artemis/

Does it actually OOM? If you already have a system showing memory use, taking a heap dump and having a look at what is using the memory would be good, might point to something obvious/specific to then be investigated rather than largely guessing and trying to reproduce what you can already see.

Can you also give some more details of what is happening in the system. Are the connections, or sessions, or consumers long-lived or do they come and go (or some mixture) ? What size are the messages? What is happening on the queues not shown (those on the second page), etc. 

 

> Memory leak when using AMQP 
> ----------------------------
>
>                 Key: ARTEMIS-3517
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-3517
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>          Components: AMQP
>    Affects Versions: 2.18.0
>         Environment: Artemis mq running as a docker container. The image was built as debian.sh 
> Consumers are working on .Net Core 3.1 on AMQPNetLite.Core version 2.4
>  
> {color:#444444}Java info: {color}
> {color:#444444}openjdk version "1.8.0_302"{color}
> {color:#444444}OpenJDK Runtime Environment (build 1.8.0_302-b08){color}
> {color:#444444}OpenJDK 64-Bit Server VM (build 25.302-b08, mixed mode){color}
>  
> Artgument in artemis.profile:  -XX:+UseG1GC -XX:+UseStringDeduplication -Xms512M -Xmx2G
>            Reporter: Paweł Trynkiewicz
>            Priority: Major
>             Fix For: 2.19.0
>
>         Attachments: memory_usage.PNG, ques.PNG
>
>
> In our Docker, swarm Artemis run in a cluster where two nodes are master and two slaves. 
> The cluster is practically not used, but I have noticed a very rapid increase in RAM consumption on the node to which AMQP consumers are connecting. On the remaining nodes, I did not notice any significant memory consumption. Therefore, I believe that consumers of AMQP messages are causing the problem on the server side.  The problem appeared after switching from vromero/activemq-artemis-docker from version 2.15 to 2.18.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)