Why does my Karaf container slowly overwhelm the server?

[Anonymous]

Hi, I'm running the Talend Runtime karaf container from Talend ESB r101800 v.5.3.0 with these features installed and running (see attached).
I'm running one ActiveMQ broker, 4 jobs that produce ActiveMQ messages and 4 that consume them.
No matter how large I make $JAVA_MAX_MEM, eventually the container occupies all the memory allotted to it, the web consoles become slow and eventually unresponsive. After that, processing of messages stops altogether and I have to force kill the container and restart it.
What I've noticed when checking the ActiveMQ queue stats (activemq:query -QQueue=*) is that some of my queues have a constantly growing ProducerCount number. I don't know if that's related.
Also attached is my ActiveMQ configuration XML.

To see the whole post, download it here

[Anonymous]

I performed a diagnostic dump and found that thousands of "NonBlockingInputStreamThread" threads have been spawned, and there are more all the time. I don't really understand what these are for, but it seems wrong to me that the software would create more and more of something without deleting the old, presumably unused, ones.
Is it possible something in one of my jobs is doing this? Is it a bug in Karaf, or something else?

[Anonymous]

just moved your topic to the relevant Forum to give it a better visibility with the right people.

[Anonymous]

created a bug report about it:
https://jira.talendforge.org/browse/TESB-12495

[Anonymous]

Thanks for the attention. I was able to resolve the issue by running the container under its own user instead of root.

[Anonymous]

Hi bthompson-bwb,
Could you please open a jira issue of ESB project on Talend Bug Tracker, our ESB developer will check if it is a bug.
Best regards
Sabrina