[
https://issues.apache.org/jira/browse/AMQ-6492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15625949#comment-15625949
]
Ben Manes commented on AMQ-6492:
--------------------------------
Fixed in CAMEL-8994. It appears that current versions of ActiveMQ (e.g.
v5.14.0) use Camel v2.16.3 which includes this. Subsequent versions switched to
the Java 8 rewrite which allocates lazily based on contention, rather than
upfront, which should further avoid this type of problem.
> concurrentlinkedhashmap used by LRU cache consumes 100MB memory on 1024 core
> Solaris T5
> ---------------------------------------------------------------------------------------
>
> Key: AMQ-6492
> URL: https://issues.apache.org/jira/browse/AMQ-6492
> Project: ActiveMQ
> Issue Type: Bug
> Components: activemq-camel
> Affects Versions: 5.14.1
> Environment: Solaris T5 with 1024 core
> java 1.7.045
> Reporter: Matt Baker
> Attachments: screenshot-1.png, screenshot-2.png, screenshot-3.png
>
>
> A basic setup of activemq-all.jar setup on a high CPU count machine ends up
> consuming 100's of MB due to the poor memory allocation used by the
> ConcurrentLinkedHashMap implementation that is being used by LRU Cache.
> The ConcurrentLinkedHashMap allocates concurrency for its lock table based
> on the number of CPUs with no upper limit. Apache Camel uses multiple copies
> of LRUCache each of which consume ~20MB of memory. Due to this my install
> ends up consuming almost 400MB of memory.
> Will post screenshot from yourkit.
> Have logged this with owner of this code.
> https://github.com/ben-manes/concurrentlinkedhashmap/issues/49
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
